US20100063870A1 - Methods and apparatus to determine the effects of trade promotions on competitive stores - Google Patents

Abstract

Example methods and apparatus to determine the effects of trade promotions on competitive stores are disclosed. A disclosed example method includes identifying a target product sold at the target store associated with the trade promotion, and identifying a plurality of competitive products having a degree of similarity to the target product. The example method also includes retrieving all commodities volume (ACV) data from a plurality of stores in a geography of interest that sell the plurality of competitive products, calculating a percentage overlap of the ACV data to generate a competitive trading area table of store pairs within a threshold value of the overlap, calculating a brand store switching value by dividing a brand incremental volume by a base equivalent volume for the store pairs, the brand store switching value indicative of the trade promotion store effect; and generating a report including the brand store switching value.

Description

RELATED APPLICATION
• This patent claims the benefit of U.S. provisional application Ser. No. 61/094,691, filed on Sep. 5, 2008, which is hereby incorporated by reference herein in its entirety.
FIELD OF THE DISCLOSURE
• The present disclosure relates generally to monitoring sales and, more particularly, to methods and apparatus to determine the effects of trade promotions on competitive stores.
BACKGROUND
• Retail establishments and product manufacturers are often interested in the shopping activities, behaviors, and/or habits of consumers. Consumer activity related to shopping can be used to correlate product sales with particular shopping behaviors and/or to improve timings or placements of product offerings, product promotions, and/or advertisements. Known techniques for monitoring consumer shopping activities include conducting surveys, counting patrons, and/or conducting visual inspections of shoppers or patrons in the retail establishments.
• Trade promotions for consumer goods often result in additional or incremental unit sales of a product in the week the trade promotion is run. This effect can be referred to as the pantry loading effect, which is associated with the tendency of consumers to stock up their households with those products to which trade promotions apply. Although the pantry loading effect may result from a sudden increase in product sales at a particular time when a promotion on that product was run, this effect often causes consumers to delay subsequent purchases of that same product. Thus, although offering trade promotions provides retailers and manufacturers a means by which to move large volumes of stock in a short amount of time, such promotions often result in a reduction of later purchases leading to lowering sales trends in subsequent weeks. Reduced sales volumes are often attributed to numerous factors. That is, the pantry loading effect may not be the only factor leading to reduced sales.
BRIEF DESCRIPTION OF THE DRAWINGS
• FIG. 1 depicts an example system configured to implement a syndicated sales analysis service for a plurality of retail establishments to analyze sales volumes information and causal factors related thereto using the example methods and apparatus described herein.
• FIG. 2 is a block diagram of an example apparatus that can be used to analyze sales volumes information in connection with the example system ofFIG. 1.
• FIG. 3 is a block diagram of an example apparatus that can be used to generate weighting values in connection with the example system ofFIGS. 1 and 2.
• FIG. 4 is a flow diagram representative of machine readable instructions that may be executed to analyze sales volumes information and causal factors related thereto for retail establishments under analyses in connection with the example system ofFIG. 1.
• FIG. 5 illustrates a flow diagram representative of machine readable instructions that may be executed to acquire and prepare data for modeling in connection with the flow diagram ofFIG. 4.
• FIG. 6 illustrates a flow diagram representative of machine readable instructions that may be executed to create product pairings in connection with the flow diagram ofFIG. 5.
• FIGS. 7A and 7B illustrate a flow diagram representative of machine readable instructions that may be executed to prepare data for modeling in connection with the flow diagram ofFIG. 5.
• FIG. 8 illustrates a flow diagram representative of machine readable instructions that may be executed to determine causal facts in connection with the flow diagram ofFIGS. 7A and 7B.
• FIG. 9 illustrates a flow diagram representative of machine readable instructions that may be executed to calculate lift data in connection with the flow diagram ofFIGS. 7A and 7B.
• FIG. 10 illustrates a flow diagram representative of machine readable instructions that may be executed to calculate competitive trade data in connection with the flow diagram ofFIGS. 7A and 7B.
• FIG. 11 illustrates a flow diagram representative of machine readable instructions that may be executed to calculate competitive regular price data in connection with the flow diagram ofFIGS. 7A and 7B.
• FIG. 12 illustrates a flow diagram representative of machine readable instructions that may be executed to calculate competitive distribution data in connection with the flow diagram ofFIGS. 7A and 7B.
• FIG. 13 illustrates a flow diagram representative of machine readable instructions that may be executed to calculate trading area data in connection with the flow diagram ofFIG. 5.
• FIGS. 14A and 14B illustrate a flow diagram representative of machine readable instructions that may be executed to determine base driver volume data in connection with the flow diagram ofFIG. 4.
• FIGS. 15A and 15B illustrate a flow diagram representative of machine readable instructions that may be executed to determine incremental driver volume data in connection with the flow diagram ofFIG. 4.
• FIG. 16 is a block diagram of an example processor system that may be used to implement some or all of the example methods and apparatus described herein.
• FIGS. 17A-17E illustrate example pseudo code to prepare data for modeling in connection with the example system ofFIG. 1.
• FIGS. 18A and 18B illustrate example pseudo code to prepare data for store switching calculations in connection with the example system ofFIG. 1.
• FIGS. 19A and 19B illustrate example pseudo code to specify model parameters in connection with the example system ofFIG. 1.
• FIGS. 20A and 20B illustrate example pseudo code to perform base driver scoring calculations in connection with the example system ofFIG. 1.
• FIGS. 21A-21C illustrate example pseudo code to perform incremental trade driver scoring calculations in connection with the example system ofFIG. 1.
DETAILED DESCRIPTION
• Although the following discloses example methods and apparatus including, among other components, software executed on hardware, it should be noted that such methods and apparatus are merely illustrative and should not be considered as limiting. For example, it is contemplated that any or all of these hardware and software components could be embodied exclusively in hardware, exclusively in software, or in any combination of hardware and software. Accordingly, while the following describes example methods and apparatus, persons having ordinary skill in the art will readily appreciate that the examples provided are not the only way to implement such methods and apparatus.
• Example methods and apparatus to determine the effects of trade promotions on competitive stores are disclosed. A disclosed example method includes identifying a target product sold at the target store associated with the trade promotion, and identifying a plurality of competitive products having a degree of similarity to the target product. The example method also includes retrieving all commodities volume (ACV) data from a plurality of stores in a geography of interest that sell the plurality of competitive products, calculating a percentage overlap of the ACV data to generate a competitive trading area table of store pairs within a threshold value of the overlap, calculating a brand store switching value by dividing a brand incremental volume by a base equivalent volume for the store pairs, the brand store switching value indicative of the trade promotion store effect; and generating a report including the brand store switching value.
• A disclosed example apparatus includes a product reference interface to identify a target product sold at a target store, the target product associated with the trade promotion, and a target product comparator to identify a plurality of competitive products having a degree of similarity to the target product. The example apparatus also includes a promotion analysis engine to retrieve all commodities volume (ACV) data from a plurality of stores in a geography of interest, a trading area competitive activity assessor to calculate a percentage overlap of the ACV data and to generate a competitive trading area table of store pairs within a threshold percentage value, the store pairs containing the identified plurality of competitive products, and a pre-model data generator to calculate a brand store switching value by dividing a brand incremental volume by a base equivalent volume for the store pairs, the brand store switching value indicative of the trade promotion store effect.
• The example methods and apparatus described herein may be implemented by a consumer metering entity, by retail businesses, or by any other entity interested in collecting and/or analyzing the effects of different causal factors on the performance of product sales for retail entities. The example methods and apparatus may be used to help marketing professionals better understand consumers and how to reach and influence consumers that buy goods. Retailers and manufacturers are often interested in assessing the contributory effect of each factor on sales. One such assessment involves determining whether a future trade promotion and its terms will have an overly negative affect on future sales. Another assessment involves determining whether previously offered trade promotions were a significant cause of subsequently observed lower sales. For example, by analyzing different causal factors that could potentially influence product sales performance, the example methods and apparatus described herein can be used to determine whether those factors were or will be influential and the extent to which they affected or will likely affect product sales performance. Such information can help retail management to better understand the impact of different effects or factors on growth of sales volumes. In some cases, retail managers may be able to control the presence of such effects or factors to influence sales performance. In other instances, although the existence or presence of contributing effects or factors cannot be controlled, the extent to which such effects or factors are allowed to influence consumer purchases may be controlled. For example, knowing that a competing store's product promotion will take sales (e.g., share of the relevant market), a store may counteract by offering its own promotion.
• Causal relationships between factors and effects could include the effect that running a trade promotion on a particular product during a particular point in time has on sales of that same product subsequent to the termination of the trade promotion. Such an analysis could be performed for the effects on sales within the same store that ran the promotion. Additionally, analyses could be performed for the effects that a promotion in one store had or will have on sales in other stores. While the example methods and apparatus can be used to analyze the effects on the same product for which a promotion was run, the methods and apparatus can more generally be used to determine the affect of a trade promotion on other products. For example, while a trade promotion on brand-A beverage may affect current and subsequent sales of that brand of beverage, the example methods and apparatus described herein can also be used to determine how such a promotion affects sales of beverages associated with other brands. That is, the methods and apparatus described herein can be used to analyze relationships between different products and analyze cross-brand or cross-product sales effects.
• In a particular example, running a trade promotion on a product will result in consumers stocking up on that particular product during the promotion period. This pantry loading effect will influence sales of that product during subsequent weeks after the trade promotion has ended. However, the actual affect of pantry loading on subsequent sales is not readily apparent as there are numerous other factors that could also influence those subsequent sales. Such other factors could include competitor trade promotions, temporary or permanent price reductions for similar products, consumer affinity or loyalty to certain brands, consumers' perceived similarities between different brands, etc.
• FIG. 1 depicts anexample system100 configured to implement a syndicated sales analysis service for a plurality of retail establishments to analyze sales volumes information and causal factors related thereto using the example methods and apparatus described herein. In the illustrated example, a centraldata processing server102 receives and processes information associated with a plurality of retail establishments and an examplepromotion analysis engine103 is configured to implement a syndicated service that can centralize the processing of information from different retail establishments of retail chains across a geographic region in a substantially similar manner. In this manner, thesystem100 enables owners and/or operators of the retail chains to retrieve analysis results related to each of their retail establishments from one location (i.e., the central data processing server102). Specifically, the centraldata processing server102, via the examplepromotion analysis engine103, receives information (e.g., household purchase data, store point-of-sale volume data, retail establishment demographic data, product universal product code data (UPC data), etc.) from a plurality of data sources and analyzes the information to generate sales volume data related to different retail establishments and/or wholesalers (hereinafter individually and collectively referred to as merchants), such asmerchants104aand104b, to use in determining the extent to which different effects or factors contributed to past sales for those merchants. The sales volume data generated by the centraldata processing server102 can also be used to help determine the extent to which different effects or factors may affect future sales.
• To collect demographics data of at least some of the shoppers that shop with themerchants104aand104bor other monitored merchants, the centraldata processing server102 is communicatively coupled to ademographics panelist server106. An example demographics panelist server service includes, but is not limited to the Scantron® service and/or the Homescan® service, each provided by A.C. Nielsen. The exampleHomescan® server106 is coupled to a plurality of Homescan® home units (not shown) provided at panel member homes. Each of the Homescan® home units is provided with a scanner interface to scan universal product codes (UPCs) of products purchased by household members during different shopping trips. In this manner, theHomescan® server106 can collect specific purchase data describing the quantities and brands of products and/or services purchased in each household. In addition, theHomescan® server106 sends surveys to household panel members via the panel members' Homescan® home units. The surveys are designed to collect demographic type information and other information describing shopping trips of the panel members. An example survey question may ask a panel member to identify, from a pre-populated list, all of the merchants that the panel member has visited within the past seven days and the days and times of day during which the shopping trips were made. The surveys may also request the panel members to provide the purpose of their shopping trips and how many people participated in the trip. In addition, the surveys may request the panel members to provide their gender and age. TheHomescan® server106 is configured to collect the survey responses from the different Homescan® home units and to collate and organize the survey data. In the illustrated example, theHomescan® server106 is configured to communicate the survey response information to the centraldata processing server102.
• To receive demographics data related to respective monitored merchants, the centraldata processing server102 is communicatively coupled to amerchant demographics server108. Themerchant demographics server108 is provided with merchant demographics information describing respective monitored merchants and is configured to communicate the demographics information to the centraldata processing server102. In the illustrated example, merchant demographic information includes demographic information about typical shoppers of corresponding merchants, sizes of merchant stores, merchant categories (e.g., supercenter store, supermarket store, discount store, drug store, etc.), percentages of consumers drawn from geographic zones surrounding each merchant store, etc.
• To collect product reference data indicative of all products and/or services offered for sale by respective merchants, the centraldata processing server102 is communicatively coupled to aproduct reference server110. For each of themerchant stores104aand104bor other monitored merchant stores, theproduct reference server110 is configured to store a UPC corresponding to each product and/or service offered for sale by that merchant. Each UPC is stored in association with the name of the product it represents including brand, unit size, and price. A UPC may be associated with several prices, each corresponding to a different offering time (e.g., a different week) at which the product's price was lower or higher than usual. For example, a week-long product promotion involving a reduced price will be reflected in a week-price data set stored in association with the UPC of that product.
• To collect point-of-sale (POS) sales data indicative of all products and/or services sold in respective merchant stores, the centraldata processing server102 is communicatively coupled to asales data repository112. In the illustrated example, thesales data repository112 is communicatively coupled to the merchant stores104a-b. Thesales data repository112 may further be communicatively coupled to any other monitored merchant stores to collect sales data therefrom. Each of the merchant stores104a-bmay store sales data in local servers (not shown) that are communicatively coupled to POS terminals at checkout counters used to perform sales transactions. In this manner, product/service sales data from all transaction logs can be communicated to thesales data repository112 in an automated manner with minimal or no human interaction.
• FIG. 2 is a block diagram of the examplepromotion analysis engine103 that can be used to analyze sales volumes information in connection with theexample system100 ofFIG. 1. In the illustrated example, the examplepromotion analysis engine103 includes aweighting value generator202, abaseline volume manager204, a trading areacompetitive activity assessor206, apre-model data generator208, amodel coefficient estimator210, a scoringprocessor212, a panelistpurchase data store214, aproduct reference database216, a trading areasales data store218, asales data repository220, a transformeddata repository222, amodel coefficients repository224, and avolume driver repository226.
• The examplepromotion analysis engine103 may be implemented using any desired combination of hardware, firmware, and/or software. For example, one or more integrated circuits, discrete semiconductor components, and/or passive electronic components may be used. Thus, for example, any of theweighting value generator202, thebaseline volume manager204, the trading areacompetitive activity assessor206, thepre-model data generator208, themodel coefficient estimator210, the scoringprocessor212, the panelistpurchase data store214, theproduct reference database216, the trading areasales data store218, thesales data repository220, the transformeddata repository222, themodel coefficients repository224, and/or thevolume driver repository226, or parts thereof, could be implemented using one or more circuit(s), programmable processor(s), application specific integrated circuit(s) (ASIC(s)), programmable logic device(s) (PLD(s)), field programmable logic device(s) (FPLD(s)), etc.
• Some or all of theweighting value generator202, thebaseline volume manager204, the trading areacompetitive activity assessor206, thepre-model data generator208, themodel coefficient estimator210, the scoringprocessor212, the panelistpurchase data store214, theproduct reference database216, the trading areasales data store218, thesales data repository220, the transformeddata repository222, themodel coefficients repository224, and/or thevolume driver repository226, or parts thereof, may be implemented using instructions, code, and/or other software and/or firmware, etc. stored on a machine accessible or readable medium and executable by, for example, a processor system (e.g., theexample processor system1610 ofFIG. 16). When any of the appended claims are read to cover a purely software and/or firmware implementation, at least one of theweighting value generator202, thebaseline volume manager204, the trading areacompetitive activity assessor206, thepre-model data generator208, themodel coefficient estimator210, the scoringprocessor212, the panelistpurchase data store214, theproduct reference database216, the trading areasales data store218, thesales data repository220, the transformeddata repository222, themodel coefficients repository224, and/or thevolume driver repository226 is hereby expressly defined to include a tangible medium such as a memory, DVD, CD, etc. storing the software and/or firmware.
• Turning in detail toFIG. 2, theweighting value generator202 is configured to generate weighting values and/or a weighting similarity index value (hereinafter referred to as a weighting similarity index) for product pairs to indicate the similarities or diversities between different products based on particular criteria. For example, a weighting index value between brand A soda and brand B soda (i.e., an A:B weighting index) based on price will indicate the consumer-perceived similarity between the brands and the propensity of consumers to purchase either product based on price differences. In this instance, a weighting similarity index closer to one may indicate that consumers consider the two soda brands as closely comparable to one another such that price (as opposed to taste or brand loyalty) may be the most significant factor contributing to purchasing brand A over brand B. In the example implementations described herein, theweighting value generator202 is implemented using Jaccard index values or Jaccard similarity coefficients, which are values based on the size of an intersection of product characteristics divided by the size of the union of sample sets. However, other weighting techniques may additionally or alternatively be used to generate weighting similarity index values to represent similarities or differences between different products.
• Generally speaking, the examplepromotion analysis engine103 ofFIG. 2 collects transaction data related to panelist purchase data and identifies pairings of target product(s) of interest and one or more competitive product(s) that are deemed to be competitively related to the target product(s). One or more criteria and/or thresholds may be employed to calculate which competitive products should be paired with the target product(s) of interest based on, for example, similarity in price, similarity in features, similarity in one or more categorical product types, and/or the frequency at which a panelist purchases the competitive product instead of the target product(s). To minimize and/or eliminate bias effects that typically occur when an analyst attempts to select one or more competitive products to compare against the target product(s) of interest, the examplepromotion analysis engine103 ofFIG. 2 saves the list of calculated competitive products for subsequent analysis. As discussed in further detail below, employing such a calculated list operates as a model to effectively restrict the user's scope of competitive products used during analysis of target product(s) of interest being promoted.
• Additionally, the examplepromotion analysis engine103 ofFIG. 2 calculates, retrieves, and/or receives a baseline volume of sales, which refers to the volume of sales that would occur absent promotions of the target product of interest. One or more working variables are calculated and input into a regression model, and additional data related to the merchants by which the target and competitive products are sold is acquired by the examplepromotion analysis engine103. One or more model coefficients are calculated before the examplepromotion analysis engine103 ofFIG. 2 facilitates a calculation of the volume sales for a target product in view of a promotion. Further, the examplepromotion analysis engine103 facilitates a calculation of the volume sales in competitive stores that result from the promotion in a target store. Without limitation, the examplepromotion analysis engine103 ofFIG. 2 facilitates a calculation of expected sales volumes in response to consumer pantry loading after a promotion on the target product has run.
• In the example implementations described herein, the weighting similarity index values are used to determine inter-brand weighting values and intra-brand weighting values. Inter-brand weighting values are used to quantify the similarities or differences between products of different brands as in the example above involving brand A soda and brand B soda. Intra-brand weighting values are used to quantify the similarities and differences between products within the same brand such as brand A cola and brand A root beer. In this manner, inter-brand weighting values can be used to quantify the effects of market erosion on competitor products in which sales of a target product under analysis are lost to competitor products due to one or more factors including, for example, consumer taste preference, price, promotion, brand loyalty, etc. Intra-brand weighting values can be used to quantify the effects of cannibalization within the same brand. Cannibalization occurs when a company offers a new product or service that takes sales or market share from one or more of its already existing products or services. Cannibalization may be due to one or more factors including, for example, consumer taste preference, price, promotion, novelty, etc.
• Turning briefly toFIG. 3, the exampleweighting value generator202 is shown in greater detail. In the illustrated example ofFIG. 3, theweighting value generator202 includes apanelist interface302 to communicatively connect with the panelistpurchase data store214, and aproduct reference interface304 to communicatively connect with theproduct reference database216. Based on an identified target product of interest, an exampletarget product comparator306 identifies one or more competitive product candidates via theproduct reference interface304 that are within a similar product category (e.g., carbonated beverages, snack foods, cold breakfast cereals, etc.). Additionally or alternatively, the exampletarget product comparator306 may further narrow any resulting list of competitive product candidates down by applying one or more sub-category constraints on the candidate list. For example, a general category of carbonated beverages may further contain soda sub-categories of cola, citrus, and fruit. Similarly, a general category of cold breakfast cereals may further contain cereal sub-categories of health-cereals, sugared-cereals, and children's cereals.
• The exampletarget product comparator306 also refers to thepanelist interface302 to determine which identified competitive product candidates were purchased by panelists at or near a selected time-period as the target product of interest. One or more thresholds may be employed by the exampletarget product comparator306 to determine which competitive product candidates should be added to a working list of competitive products that correspond to the target product of interest. For example, if the target product of interest is Coke®, which is included in the general category of soda and the sub-category of cola, then the exampletarget product comparator306 searches the panelist data store to determine how many instances occurred in which a consumer purchased Coke® on one occasion, but also purchased a competing cola brand instead of Coke® on one or more alternate occasions. In this example, if the consumer is found to have purchased Pepsi® on a substantially greater number of occasions, but only purchased RC Cola® on relatively infrequent occasions, then Pepsi® may be deemed to be an appropriate competitive product to Coke® for purposes of competitive analysis.
• The exampletarget product comparator306 is also coupled to aweighting assignor308 to apply a similarity calculation (weighting similarity index value) to each pair of identified target product and competitive product. As described above, the Jaccard index may be employed by theweighting assignor308 for each pair to yield a unitless numerical identifier (weighting similarity index value) indicative of the similarity between products (e.g., inter-brand weighting values or intra-brand weighting values). The assigned weight associated with the Jaccard index represents the inter-brand weighting value β_Inter(i,k), where the variable i represents the target product of interest and variable k represents a competitive product. Similarly, a Jaccard index may be calculated for different UPCs by the same manufacturer to represent potential effects of cannibalization. The assigned weight associated with the Jaccard index in view of two products of the same brand represents the intra-brand weighting value β_Intra(i,k), where variable i represents the target product of interest and variable k represents another product by that same manufacturer.
• Returning toFIG. 2, thebaseline volume manager204 is also configured to generate, calculate, retrieve, and/or otherwise receive baseline volumes for different target products under analysis. A baseline volume is indicative of the normal expected sales volume for a product absent any trade promotion for that product. Trade promotions can include feature advertisements (e.g., coupons or advertisements in newspapers, magazines, etc.), in-store displays (e.g., in-store posters, shelf talkers, coupon dispensers), or price reductions. In the presence of trade promotions, promoted products typically undergo an increase in sales. To determine the effects of a promotion on the increased sales for a particular product, a baseline volume can be used to determine what the normal sales volume for that product would have been absent the promotion and determining the difference in volume (known as the incremental volume) observed between the baseline volume and the actual sales volume during the promotion period. This difference in volume or incremental volume is the volume attributable to the effects of the trade promotion. Different techniques are available for determining baseline volumes. On such known technique is the Box-Jenkins time series calculation. In the example implementations described herein, baseline volumes are based on a store-week-UPC level, meaning that a baseline volume represents the sales in a particular store, during a particular week, of a particular product identified by its unique UPC.
• The illustrated example trading areacompetitive activity assessor206 ofFIG. 2 is configured to receive census blocks for each merchant (e.g., retailer, wholesaler, and/or other business entity, and hereinafter generally referred to as “retailers”). A census block is a delineated portion of purchasers that spend money with the retailer and may be defined based on geographic proximity to the retailer. For example, a single retailer may include four census blocks that each contain an equal number of shoppers located to the North, South, East, and West of the retailer. Additionally, each census block is typically associated with a percentage value to represent how much each block purchases (e.g., the North census block accounts for 35% of the retailer's weekly sales, while the South and East account for 25%, and the West accounts for only 15% of the retailer's weekly sales).
• Each census block typically includes shoppers that shop with more than one merchant. Thus each census block includes some amount of overlap with one or more other merchants. Overlap of the census blocks may also produce an effect of greater or lesser sales at a first merchant based on promotions run by a second merchant. For instance, if the second merchant runs a promotion for sales of Coke® 12-ounce cans, then a corresponding effect of lesser sales of Coke® of the first merchant (that has a common census block) may occur. As described in further detail below, the example trading areacompetitive activity assessor206 acquires merchant data from the trading areasales data store218 to facilitate one or more volume calculations at competitive stores of a product in response to a promotion of that sold product of interest.
• The illustrated example ofFIG. 2 also includes apre-model data generator208 to generate variables to be used in a model prior to being used in one or more scoring algorithms that ultimately yield volume estimates as a result of target product promotional activity. As discussed in further detail below, the examplepre-model data generator208 ofFIG. 2 uses point-of-sale (POS) input data, calculated competitive product pairing data, panelist data, and trading area data to calculate an estimated volume of products sold in response to one or more causal factors. Causal factors may include, but are not limited to, a temporary price reduction (TPR), a feature (e.g., flyers describing the target product), a display (e.g., an in-store display of the target product, such as at the end of an aisle), and/or a combination of a feature and a display.
• The examplepre-model data generator208 also minimizes and/or eliminates data anomalies due to seasonality effects. For example, in the event that the sale of chocolates is particularly high during the months of December and February (e.g., Christmas holiday and Valentines Day), corresponding sales peaks of such products may occur absent a promotion. As such, the examplepre-model data generator208 calculates a dimensionless index to minimize such expected peaks of chocolate sales so that corresponding volume data calculations are not artificially high.
• The examplepre-model data generator208 also calculates equivalent volumes and base equivalent volumes for products to eliminate, for example, product packaging disparities. Such disparities occur when a product manufacturer sells products with varying pre-packaged quantities, which may prevent direct comparisons between the same brands on a per-unit sale basis. For example, Coke® sells 12-ounce cans of cola as well as 2-liter bottles of cola. On a per-unit sale of a 6-pack UPC and a 2-liter UPC sale, direct comparison is not representative of a true measure of the ultimate amount of cola sold (i.e., a 2-liter bottle is approximately 68 ounces). As such, thepre-model data generator208 converts one or more target products of interest into a corresponding equivalent volume unit to, in effect, implement a common base-comparison unit for all brand package configurations. In the illustrated example above, a common base-comparison unit for Coke® cola products (whether such products are sold in cans or bottles) is units of ounces.
• One or more competitive effects variables are also calculated by the examplepre-model data generator208. Generally speaking, there are three actions that a competitor can take to affect sales of the target product. In particular, the competitor can increase or decrease the amount of trade promotion activity, the competitor can change their regular price, and/or the competitor can put new UPCs (i.e., new or additional products) on store shelves. Increasing or decreasing the amount of trade promotion activity may include running features, displays, temporary price reductions, and/or any combination thereof. Thepre-model data generator208 calculates a competitive trade index, such as an example Jaccard weighted category index, as a measure of the weighted aggregate level of trade promotion activity present in the competitive brands relative to the target brand of interest. The competitive trade index allows a determination/estimation of the degree to which another trade promotion may be affecting the sales of the target product (inter-brand trading index), as well as the degree to which the trade promotion of the target product may be affecting the sales of competitive products. In the event that the target product trade promotion affects sales of its own products, the examplepre-model data generator208 also calculates a brand-trade index (intra-brand trading index) to identify cannibalization effects.
• The second action that a competitor may take that affects sales of the target product of interest is to change the product regular price. Generally, if a competitor lowers a shelf price of the competitive product, a corresponding decrease in sales is expected of the target product of interest. As such, the examplepre-model data generator208 examines the aggregated price changes in items/products that are considered to be competitive, weights such changes based on trade weights (e.g., a Jaccard weight) and creates an inter-brand price weight and/or an intra-brand price weight to represent a measure and/or a magnitude of the degree to which competitors take price action from the target product of interest.
• The third action that a competitor may take that affects sales of the target product of interest is to place new UPCs on retail shelves. Generally, as the competitor increases the number of items in distribution, a corresponding negative affect on sales volumes of the target product is expected. As such, the examplepre-model data generator208 calculates inter-brand and intra-brand weighted distribution factors to represent the effects on sales in response to new product UPCs (i.e., both competitive UPCs and UPCs introduced by the same manufacturer as the target product of interest).
• In view of the fact that statistical analysis techniques may include the phenomenon in which there are multiple sources of randomness in the data, the examplepre-model data generator208 minimizes and/or eliminates such effects. Sources of variation include cross-sectional variants and randomness that occur over time. Removal of cross-store effect from a model may be further accomplished by, for example, a mean-centering process executed by thepre-model data generator208. Mean centering prevents potential bias and/or nuisance variants in the data.
• While calculation of one or more corresponding effects on product sales volumes may be conducted in view of a current week for a particular store, the methods and apparatus described herein also examine cross-store promotional effects. In particular, to further examine cross-store promotional effects and/or to numerically describe a store pattern of sales, the examplepre-model data generator208 prepares one or more interaction variables related to store categories (e.g., store interaction variables). Store categories typically exhibit characteristic patterns due to, in part, varying degrees of homogeneity with similar stores. For example, store characteristics related to a percent of Hispanic shoppers, African-American shoppers, and/or stores within certain latitudes/longitudes (e.g., particularly poor neighborhoods, particularly affluent neighborhoods, etc.) may have corresponding sales expectations that are statistically relevant and assist the process of projecting sales volume estimations. As discussed in further detail below, the examplepre-model data generator208 incorporates such store characteristics when calculating effects due to promotional activity.
• Estimations derived from a model-based approach typically differ from estimations derived from an algorithm-based approach. As such, the examplepre-model data generator208 also calibrates regression-based incremental volume estimates for trade effects in a manner that aligns with estimates derived from an algorithmic baseline approach. As described above, the baseline is an estimate of sales in the absence of a feature, a display, or a price cut. Any difference between the baseline and the total unit sales is referred to as an incremental volume, which is attributed to the presence of causal factors of a corresponding feature, display, or price cut. However, similar estimates via the model-based approach introduce differences that are calibrated by the examplepre-model data generator208 by calculating a correction factor, as discussed in further detail below.
• Themodel coefficient estimator210 is configured to generate coefficient values that may be used in one or more scoring algorithms to estimate one or more of volumetric sales effects of products in view of causal factors (e.g., promotional activity) in a target store, volumetric sales effects of products at competitor stores in view of such causal factors, and volumetric sales effects of products in view of consumer pantry loading behavior. Some modeling techniques view the target products and competitive products as nested within one store and/or multiple competitive stores. Additionally, the one or more competitive stores are further nested within geographic localities, which reside within delineated marketing target areas. In the illustrated example ofFIG. 2, themodel coefficient estimator210 employs a hierarchical linear modeling (HLM) technique. However, any other modeling technique may be used including, but not limited to, linear regression and/or multiple linear regression techniques.
• The coefficient values generated by the examplemodel coefficient estimator210 ofFIG. 2 include fixed effects and random effects to allow an example basedriver data generator228 to calculate net store-UPC coefficients, as described in further detail below. Generally speaking, the examplemodel coefficient estimator210 ofFIG. 2 initially sets all coefficients to zero unless and until the received data includes enough weighted evidence to justify an alternate value. Such evidence may be revealed via the aforementioned HLM and/or regression technique(s).
• Theexample scoring processor212 ofFIG. 2 is configured to employ one or more scoring algorithms that utilize one or more variables prepared by the examplepre-model data generator208 and one or more model coefficients from the examplemodel coefficient estimator210. As discussed in further detail below, the example basedriver data generator228 of the scoringprocessor212 calculates, in part, a volume due to intra-brand and inter-brand pricing, intra-brand and inter-brand distribution, volumes due to pantry loading, and volumes due to intra-brand and inter-brand trading. Additionally, an example incrementaldriver data generator230 of the scoringprocessor212 calculates, in part, raw driver volumes that are specific to one or more causal types. Causal types include, but are not limited to, corresponding volumes based on a price-cut, a feature, a display, and/or one or more combinations thereof.
• FIGS. 4 to 15B are flow diagrams representative of machine readable and executable instructions or processes that can be executed to implement the examplepromotion analysis engine103 ofFIG. 2. The example processes ofFIGS. 4 to 15B may be performed using a processor, a controller and/or any other suitable processing device. For example, the example processes ofFIGS. 4 to 15B may be implemented in coded instructions stored on a tangible medium such as a flash memory, a read-only memory (ROM) and/or random-access memory (RAM) associated with a processor (e.g., theprocessor1612 ofFIG. 16). Alternatively, some or all of the example processes ofFIGS. 4 to 15B may be implemented using any combination(s) of application specific integrated circuit(s) (ASIC(s)), programmable logic device(s) (PLD(s)), field programmable logic device(s) (FPLD(s)), discrete logic, hardware, firmware, etc. Also, one or more operations of the example processes ofFIGS. 4 to 15B may be implemented manually or as any combination(s) of any of the foregoing techniques, for example, any combination of firmware, software, discrete logic and/or hardware. Further, although the example processes ofFIGS. 4 to 15B are described with reference to the flow diagrams ofFIGS. 4 to 15B, other methods of implementing the processes ofFIGS. 4 to 15B may be employed. For example, the order of execution of the blocks may be changed, and/or some of the blocks described may be changed, eliminated, sub-divided, or combined. Additionally, any or all of the example processes ofFIGS. 4 to 15B may be performed sequentially and/or in parallel by, for example, separate processing threads, processors, devices, discrete logic, circuits, etc.
• Turning toFIG. 4, initially, the examplepromotion analysis engine103 ofFIG. 2 acquires sales data of products from multiple sources (block402). In the illustrated example ofFIG. 4, point-of-sale (POS) data is retrieved, which is indicative of sales for product UPCs, merchant location(s) in which the product(s) were sold, sale dates, and/or volumes for each UPC sold in a given week. Retailers and/or merchants may include traditional channels, such as grocery stores, community hardware stores, and/or convenience stores, but specialty channels may further include discount stores, specialty food stores, large-scale hardware stores, and/or office supply stores. Additionally, the retrieved POS data includes information related to an amount of money paid for the item(s) and whether the item(s) were associated with, or the subject of a promotion. Promotion information may include, but is not limited to, a feature (e.g., a flyer, coupon, advertisement), a display (e.g., an in-store display), and a location and/or size of the display (e.g., end of aisle, large placard, etc.).
• The POS data may be collected, sold, and/or otherwise distributed by any third-party responsible for data acquisition, such as ACNielsen®. For example, ACNielsen® has long compiled data via its Scantrack® system. With the Scantrack® system, merchants install equipment at the POS that records the UPC of every sold product(s), the quantity sold, the sale price(s), and the date(s) on which the sale(s) occurred. The POS data collected at the one or more stores is periodically exported to ACNielsen® where it is compiled into one or more databases, such as the examplesales data repository220 ofFIG. 2. Additionally, ACNielsen® employs a Homescan® system, which is a panelist-based system to measure consumer behavior and identify sales trends. With the Homescan® system, households are selected to be statistically representative of one or more demographic components of the population to be measured. The panelists of these selected households are provided with home and/or mobile scanning equipment and agree to use that equipment to identify, and/or otherwise scan the UPC of every product that they purchase and to note the identity of the retailer or wholesaler (collectively or individually referred to as “merchant”) from which the corresponding purchase was made. The data collected via this scanning process is periodically exported to ACNielsen®, where it is compiled into one or more databases, such as the example panelistpurchase data store214 ofFIG. 2.
• In the illustrated example ofFIG. 4, acquired sales data (block402) also includes trading area merchant data. Each merchant store includes characteristic data related to its size, number of employees, all commodity volume (ACV) sales, location, and/or one or more census blocks that surround the merchant. ACNielsen® also categorizes such merchants (e.g., retailers and/or wholesalers) and/or compiles data related to the store characteristics via its TDLinx® system. In the TDLinx® system, data is tracked and stored that is related to, in part, a merchant store parent company, the parent company marketing group(s), the number of stores in operation, the number of employee(s) per store, the geographic address and/or phone number of the store(s), and/or the channel(s) serviced by the store(s). Data indicative of the merchant store(s) may be stored in the example trading areasales data store218 ofFIG. 2.
• Returning toFIG. 4, model coefficients are created (block404) in a manner consistent with the modeling technique(s) selected by one or more analysts. As described above, the selected modeling technique(s) may include, but are not limited to linear regression, multiple linear regression, and/or hierarchical linear modeling to set one or more coefficient values. After setting the one or more coefficient values, such as the fixed-effect coefficients and/or the random-effect coefficients to a non-default zero value (block404), the examplepromotion analysis engine103 applies such coefficients, acquired data, and prepared variables to one or more scoring algorithms (block406), as described in further detail below.
• FIG. 5 is a flowchart representative of example machine readable instructions that may be executed to acquire and prepare data for modeling atblock402 ofFIG. 4. Atblock502, the exampleweighting value generator202 creates one or more pairings of competitive products for each target product of interest. From each product pairing created, the exampleweighting value generator202 calculates a weighting similarity index value, which may be represented as a unitless numerical value to indicate a degree of similarity between the target product of interest and the competitive product. In the illustrated examples described herein, the calculated weighting similarity index value is implemented via the Jaccard similarity methodology, but other weighting techniques may additionally or alternatively be used to generate the weighting similarity index values.
• To allow an understanding of how causal factors, such as promotional prices (e.g., price reductions), displays, and/or features affect sales volumes, the examplebaseline volume manager204 calculates a baseline volume sales estimate (block504) that results in the absence of any promotional activity. The difference between the calculated baseline volume (block504) and product(s) sold in excess of that baseline are referred to as incremental sales. Acquired data and product pairs are further used by the examplepre-model data generator208 to, in part, generate working variables (block506) to facilitate base-driver and incremental-driver scoring calculations. Additionally, the example trading areacompetitive activity assessor206 includes trading-area data to identify one or more census blocks for each merchant (block508). In the illustrated example ofFIG. 5, trading area data is retrieved from Spectra® and/or TDLinx® services, but any other trading area data services may also be used.
• FIG. 6 is a flowchart representative of example machine readable instructions that may be executed to create product pairings between a target product of interest and one or more competitive products atblock502 ofFIG. 5. Atblock602, the exampleweighting value generator202 receives a target product of interest. The target product of interest includes a corresponding universal product code (UPC) that, when referenced against theproduct reference database216, reveals one or more specific characteristics related to the target product of interest. The exampleproduct reference database216 ofFIG. 2 may be implemented as a third-party global database in which manufacturers purchase the rights to exclusively print and use a unique UPC indicia. Additionally or alternatively, product reference information is also maintained by ACNielsen® as a Product Reference Library (PRL) that codes more than 700,000 items, in which each item includes an average of forty (40) descriptive product characteristics. The characteristics for each item may include, but are not limited to, manufacturer name, product size, brand, flavor, lot number, serial number, package type, and/or nutritional information.
• As described above in connection with ofFIG. 3, the exampleproduct reference interface304 of theweighting value generator202 searches the productreference data store216, such as the ACNielsen® PRL, for candidate competitive product(s) that may have one or more characteristic degrees of similarity to the target product of interest (block604). For example, if the target product of interest is a brand of soda, theproduct reference interface304 queries the productreference data store216 for all UPCs having a general category identifier of soda and adds such category matches to a candidate list. Theproduct reference interface304 may further constrain competitive UPC/product candidates in the list to one or more sub-categories of the target product, such as a sub-category of cola, a sub-category of 6-packs, and/or a sub-category of 12-ounce bottle 6-packs (e.g., as opposed to 12-ounce can 6-packs).
• Theexample panelist interface302 of theweighting value generator202 further queries the panelistpurchase data store214 for instances of the target product purchase and one or more purchases of a competitive product from the candidate list (block606). While the resulting candidate list may be very large depending on the product type(s), the exampletarget product comparator306 of theweighting value generator202 further constrains the candidate list by comparing matching purchase instances against one or more threshold(s) (block608). For example, if the target product of interest is a Coke® 2-liter bottle, theproduct reference interface304 may have generated a candidate competitive product list that includes any number of competitive 2-liter soda manufacturers. However, theexample panelist interface302 facilitates a method by which that candidate list may be reduced to a manageable size by identifying instances when such competitive products were actually purchased by consumers. In effect, only competitive products that actually sell will be retrieved by thepanelist interface302 for a given period of time, while other less popular soda brands that are deemed relatively obscure will not be considered.
• Thresholds employed by the exampletarget product comparator306 may include, but are not limited to, a maximum percent occurrence within a two-week time span. For example, thetarget product comparator306 may reduce the competitive product candidate list and retain only competitive products that were most frequently purchased by the consumer within a two-week time-span of purchasing the target product of interest. Employing such threshold test(s) may reveal, for example, that a consumer that purchases a 2-liter bottle of Coke® inweek1 is also observed to frequently purchase a 2-liter bottle of Pepsi®, but is rarely observed to also purchase a 2-liter bottle of orange Crush® instead of the Coke® purchase. As such, a more relevant list of competitive products may be realized and used for trend analysis.
• For each of the pairs (i.e., the target product of interest paired with one of the competitive products), theexample weighting assignor308 calculates a corresponding weighting similarity value/index (block610). For example, a Jaccard weighting index calculated by theweighting assignor308 may determine that Coke® and Pepsi® have a weighted index of 0.90, which indicates a relatively significant similarity between the two. On the other hand, theweighting assignor308 may determine via the Jaccard weighting methodology that Coke® and orange Crush® have a weighted index of 0.45, which indicates a much lower relative similarity between those two products. In operation, the exampletarget product comparator306 may further employ one or more thresholds to retain only those weighting similarity index values meeting such threshold parameter(s) and/or matching a target index value. The one or more thresholds allow a competitive product subset to be created, in which the products within the subset exhibit similar weighting similarity index values and/or weighting similarity index values that meet one or more target values and/or threshold values.
• Each target product of interest may have any number of candidate competitive products from the candidate list. Thus, if there are additional competitive products to analyze (block612), then control passes to block606. Additionally, if there are more target products of interest for which corresponding weights are to be calculated (block614), the exampleweighting value generator202 may repeat the example flowchart ofFIG. 5 beginning atblock602.
• FIGS. 7A and 7B are representative of example machine readable instructions that may be executed to prepare pre-model data atblock506 ofFIG. 5. Atblock702, the examplepre-model data generator208 identifies at least one store, week, and/or product to be analyzed (store-week-UPC). One or more unique stores of interest may be identified by a store record key, and each week of interest may be identified by a record key indicative of any seven day span of store data. As such, the examplepre-model data generator208 may identify one or more store-week-UPC data sets for which to calculate working variables that facilitate further volume calculations. While the illustrated examples described herein include a lowest level of time detail from each store as a week in duration, any other duration of interest may also be applied to the methods and apparatus described herein. Additionally or alternatively, thepre-model data generator208 may identify a product-centric week-UPC to analyze volume effects of product trade promotions on other products.
• Thepre-model data generator208 retrieves one of the store-week-UPC POS sales volume data sets (block704) from one or more of the panelistpurchase data store214, the productreference data store216, and thesales data repository220. In some example implementations, the sales volume may correspond to the sales of a product sold in one packaging size (e.g., two-liter soda), in which case the sales volume data is retrieved based on only one unique UPC. In other instances, the sales volume may correspond to the sales of a product sold in two or more packaging sizes (e.g., two-liter bottle, 12-ounce cans, 6-ounce cans, 24-ounce bottles), in which case the sales volume data is retrieved based on two or more unique UPCs. In either case, the sales volume analyses described herein can be performed using base equivalent units by representing the sales volumes using common, standard volumetric units (or piece-count units). For example, if a soda product is sold in different packaging sizes, the sales volume analyses can convert the POS sales volume data for each packaging type into ounces such that a sale of a two-liter bottle represents a sale of sixty-eight ounces of soda based on a base-equivalent measure, while a sale of a six-pack of twelve-ounce cans represents a sale of seventy-two ounces of soda based on the same base-equivalent measure.
• A dependent variable of the sales volume for each product is calculated as the log of the number of units sold, and independent variables of the log of the regular price and number of UPCs selling are calculated for future use (block706). Causal factors associated with the store-week-UPC data are identified (block708), which may include temporary price reductions (TPRs), features (e.g., flyers), displays (e.g., in-store displays in an aisle), and/or any combination of features and displays. However, despite the presence of causal facts, the examplepre-model data generator208 calculates seasonality and trend index values to prevent data influences that are generally expected based on known seasonal effects (block710). In other words, the seasonality index values and/or the trend index values normalize sales volume data. For example, seasonal effects may be apparent in view of barbeque sauce sales in summer months versus winter months. Additionally, the examplepre-model data generator208 may employ trading area sales data to further adjust seasonality index values based on geographic factors, such as a tendency for higher barbeque sauce sales in southern states versus northern states during the summer months.
• The examplepre-model data generator208 also calculates one or more lift variables that span one or more time periods to facilitate volume calculations in accordance with cannibalization (block712). As described above, cannibalization may occur when a manufacturer offers a new product or service, or promotes an existing product or service that takes sales or market share away from one or more of its own other products or services. As described in further detail below, thepre-model data generator208 retrieves inter-brand and intra-brand weights (block714), which are generated by the exampleweighting value generator202, and calculates equivalent volumes and base equivalent volumes for competitive products (block716). Based on competitor activity related to running a promotion, changing a product regular price, and/or introducing new UPCs on a store shelf, all of which tend to cause other competitive product sales to decrease, the examplepre-model data generator208 calculates corresponding competitive trade index values (block718), calculates competitive regular price variables (block720), and calculates competitive distribution variables (block722), all of which are described in further detail below.
• FIG. 8 is a flowchart representative of example machine readable instructions that may be executed to identify one or more causal facts associated with store-week-UPC data atblock708 ofFIG. 7A. Atblock802, the actual sale price (e.g., the current sale price) and the regular sale price of a product is retrieved. The examplepre-model data generator208 calculates the ratio/index of the actual sale price to the regular sale price and compares the result to a promoted price threshold (block804). For example, a threshold of 0.96 may be selected by a user to identify instances where a promotion has actually occurred, such that any calculated ratio less than 0.96 is deemed to indicate that a promotion has occurred (e.g., actual price ($5.00)/regular price ($5.50)=0.90). A ratio/index value below the promoted price threshold is, thus, indicative of a valid trade promotion having a potential to affect sales of the target product of interest and/or one or more competitive products. If the ratio is not less than the promoted price threshold, then the promoted price index is deemed to be zero and the log of the promoted price index (LPPI) is set to zero (block806). However, if the ratio is less than the threshold, then the LPPI is calculated (block808) as set forth inEquation 1 below:
• $\begin{array}{cc}LPPI=\log \frac{\left(\mathrm{Actual}\mathrm{Price}\right)}{\left(\mathrm{Regular}\mathrm{Price}\right)}& \mathrm{Equation}1\end{array}$
• To further distinguish one or more causal intercepts, the examplepre-model data generator208 queries the retrieved POS data for one or more causal parameters indicative of a feature, a display, a price reduction, or any combination thereof (block810). In the event that there are no causal parameters associated with the product purchase, then no temporary price reduction has occurred. However, if the causal parameters have one or more non-zero mapping codes, then thepre-model data generator208 identifies a valid trade promotion occurrence, such as a corresponding feature, display, or combination thereof (block812).
• FIG. 9 is a flowchart representative of example machine readable instructions that may be executed to calculate forward buy cannibalization lift terms atblock712 ofFIG. 7A. In other words, the calculation of lift terms represents a lagged effect over time, which facilitates volume calculations indicative of cannibalization effects. Atblock902, a lag-week duration is selected. Any value for lag-week may be selected, the size of which determines how large a window of time is used to ascertain one or more lagged effects that may occur in response to one or more promotions. A working variable y is initially set to zero (block904) to assist week-to-week looping of the retrieved store-week-UPC sales data. If the received data for any particular week of the loop is associated with a temporary price reduction or a causal parameter indicative of a promotion (block906), as determined earlier byblock708, then lift is calculated for that week (block908) as set forth inEquation 2 below.
• $\begin{array}{cc}\mathrm{Lift}=\log \frac{\mathrm{Units}\mathrm{Sold}}{\mathrm{Base}\mathrm{Units}}& \mathrm{Equation}2\end{array}$
• InEquation 2 above, UnitsSold represents all sales while BaseUnits represents sales that would have occurred without any promotional activity. Additionally, the BaseUnits value includes any effects due to seasonality and trending. However, if the data associated with week y in the loop is not associated with a TPR or a causal parameter indicative of a promotion (block906), then the lift is set to zero for week y (block910). To accommodate all weeks for which the lift calculation is desired, the examplepre-model data generator208 determines if the loop variable y is equal to the selected lag-week duration in block902 (block912). If not, then the loop variable y is incremented by one (block914) and control returns to block906 to determine the lift variable value for the next week.
• FIG. 10 is representative of example machine readable instructions that may be executed to calculate competitive trade index values atblock718. Atblock1002, a target item of interest is denoted by variable i and a competitive item is denoted by variable k. Theexample flowchart718 illustrates a loop-operation to calculate corresponding trade index values for each item/product of interest and all corresponding competitive items of interest. Accordingly, for any given store-week of data, an equivalent volume and a base-equivalent volume are employed to accommodate for UPCs having disparate quantities (e.g., one UPC for Coke® sold in liters and another UPC for Coke® sold in ounces).
• For each competitive product k, the examplepre-model data generator208 determines whether a temporary price reduction (TPR) has occurred or whether a causal condition is present that indicates promotional activity of the competitive product (block1004). In the event that neither a TPR nor a causal condition is present (block1004), then the variable k is incremented by one to evaluate another competitive product (block1006). On the other hand, if a TPR or a causal condition is present (block1004), then the examplepre-model data generator208 determines whether the equivalent volume is greater than the base equivalent volume (block1008). In other words, if the equivalent volume is greater than the volume that would have occurred absent any promotion (i.e., the base equivalent volume), then the data associated with the competitive product related to variable k is likely to have experienced an effect associated with promotional activity. Additionally, the pre-model data generator calculates a magnitude of the volume difference between the equivalent volume and the base equivalent volume.
• The raw inter-brand volume is calculated based on the inter-brand weighting values determined above (block1010) as set forth inEquation 3 below.
• 
  Raw Inter-Brand Volume(i,k)=β_Inter(i,k)*(EqVol(k)−BeqVol(k))  Equation 3
• InEquation 3 above, β_Inter(i,k)represents the Inter-Brand weight for the product of interest i as compared to the competitive product k. Additionally, EqVol(k) represents the equivalent volume of competitive product k, and BeqVol(k) represents the corresponding equivalent volume of product k absent any promotional activity.
• The raw intra-brand volume is calculated based on the intra-brand weighting values determined above (block1012) as set forth inEquation 4 below.
• 
  Raw Intra-Brand Volume(i,k)=β_Intra(i,k)*(EqVol(k)−BeqVol(k))  Equation 4
• InEquation 4 above, β_Intra(i,k)represents the Intra-Brand weight for the product of interest i as compared to another product k by that same manufacturer. In this case, EqVol(k) represents the equivalent volume of the manufacturer's product k, and BeqVol(k) represents the corresponding equivalent volume of the manufacturer's product k absent any promotional activity.
• The value of variable k is compared against a maximum value as a condition to allow the loop to continue evaluating competitive products, if any (block1014). However, once all competitive products and/or additional/alternate products by the same manufacturer have been evaluated for purposes of calculating the Raw Inter-Brand/Intra-Brand Volumes, the total Raw Inter-Brand Volume is calculated as the sum over all values of k (block1016), and the total Raw Intra-Brand Volume is calculated as the sum over all values of k (block1018).
• As described above, one of the three things that a competitor can do to affect sales of the target product of interest is to increase or decrease the amount of trade promotion activity. To quantify that effect, the examplepre-model data generator208 calculates the Inter-Brand Weighted Trade Index for each target product of interest (block1020) as set forth inEquation 5 below.
• $\begin{array}{cc}\mathrm{Inter}\text{-}\mathrm{Brand}\mathrm{Weighted}\mathrm{Trade}=\tanh \left(\frac{\mathrm{Total}\mathrm{Raw}\mathrm{Inter}\mathrm{Brand}\mathrm{Vol}}{\max \left(\mathrm{Beq}\mathrm{Vol}\left(i\right)\right)}\right)& \mathrm{Equation}5\end{array}$
• InEquation 5 above, BeqVol(i) represents the base equivalent volume for the target product associated with variable i.
• Additionally, to determine potential cannibalization effects, the examplepre-model data generator208 calculates the Intra-Brand Weighted Trade Index for each target product of interest (block1022) as set forth inEquation 6 below.
• $\begin{array}{cc}\mathrm{Intra}\text{-}\mathrm{Brand}\mathrm{Weighted}\mathrm{Trade}=\tanh \left(\frac{\mathrm{Total}\mathrm{Raw}\mathrm{Intra}\mathrm{Brand}\mathrm{Vol}}{\max \left(\mathrm{Beq}\mathrm{Vol}\left(i\right)\right)}\right)& \mathrm{Equation}6\end{array}$
• In the event that there are additional target products of interest as determined by monitoring the variable i (block1024), control advances to block1026 to increment to the next target product of interest.
• FIG. 11 is a flowchart representative of example machine readable instructions that may be executed to calculate competitive regular price variables atblock720 ofFIG. 7B. Atblock1102, the examplepre-model data generator208 calculates, for all store-week-UPCs, mean-center values for the value of the log of the regular price (LREGPR(k)). As discussed above, at least a second course of action that a competitor may take to affect volume sales is to change the regular price of a UPC item. To quantify this effect, the examplepre-model data generator208 calculates an inter-brand price for each competitive product (block1104) as set forth inEquation 7 below.
• $\begin{array}{cc}\mathrm{Inter}\text{-}\mathrm{Brand}\mathrm{Price}=\sum _{k=1}^{k=\max }{\beta }_{\mathrm{Inter}\left(i,k\right)}\times \mathrm{LREGPR}\left(k\right)& \mathrm{Equation}7\end{array}$
• InEquation 7 above, β_Inter(i,k)represents the Inter-Brand weight for the product of interest i as compared to the competitive product k.
• To quantify the price variable effects within the same brand as the target product of interest, the examplepre-model data generator208 calculates an intra-brand price for each alternate product that is also manufactured/produced by the manufacturer (block1106) as set forth inEquation 8 below. For example, similar to a competitor changing a sale price that causes a corresponding drop in sales of the target product, so too can a drop in a sale price of one or more of the manufacturer's other products cause their volume sales of the target product of interest to drop.
• $\begin{array}{cc}\mathrm{Intra}\text{-}\mathrm{Brand}\mathrm{Price}=\sum _{k=1}^{k=\max }{\beta }_{\mathrm{Intra}\left(i,k\right)}\times \mathrm{LREFPR}\left(k\right)& \mathrm{Equation}8\end{array}$
• InEquation 8 above, β_Intra(i,k)represents the Intra-Brand weight for the product of interest i as compared to another of the manufacturer's products k.
• FIG. 12 is a flowchart representative of example machine readable instructions that may be executed to calculate competitive distribution variables atblock722 ofFIG. 7B. As described above, at least a third activity that a competitor can take that may adversely affect sales of a target product of interest is to introduce additional UPCs on a store shelf. To quantify this effect, theflowchart722 ofFIG. 12 begins atblock1202, in which the inter-brand distribution is calculated (block1202) as set forth inEquation 9 below.
• $\begin{array}{cc}\mathrm{Inter}\text{-}\mathrm{Brand}\mathrm{Distribution}=\sum _{k=1}^{k=\max }{\beta }_{\mathrm{Inter}\left(i,k\right)}\times \mathrm{SELL}\left(k\right)& \mathrm{Equation}9\end{array}$
• InEquation 9 above, SELL(k) represents an indication that the particular product sold within a pre-determined number of weeks on a store shelf (e.g., a shelf duration). Specifically, some items may sit on a retail shelf for a relatively long period of time. As the shelf duration increases, the likelihood that the product will ultimately be sold drops. To smooth out potential problems related to slow moving sales, the SELL(k) variable is set to a value of one if the average number of UPCs for that product is greater than zero for the pre-determined number of weeks.
• To accommodate to future modeling application, the log of the Inter-Brand Distribution is calculated (block1204), as shown in Equation 10 below.
• 
  Log(Inter-Brand Distribution)=log(1+(Inter-Brand Distribution))  Equation 10
• Similarly, to quantify the effects of volume sales of the target product when the same manufacturer introduces new and/or alternate UPCs on the store shelf, the examplepre-model data generator208 calculates the Intra-Brand Distribution (block1206) as set forth inEquation 11 below.
• $\begin{array}{cc}\mathrm{Intra}\text{-}\mathrm{Brand}\mathrm{Distribution}=\sum _{k=1}^{k=\max }{\beta }_{\mathrm{Intra}\left(i,k\right)}\times \mathrm{MEAN\_NUMUPC}\left(k\right)& \mathrm{Equation}11\end{array}$
• InEquation 11 above, β_Intra(i,k)represents the Intra-Brand weight for the product of interest i as compared to another of the manufacturer's products k. MEAN_NUMUPC(k) represents the mean number of sold UPCs by each store-time period, such as, for example, a quadweek.
• To allow for use in one or more future modeling applications, the log of the Intra-Brand Distribution is calculated (block1208), as shown in Equation 12 below.
• 
  Log(Intra-Brand Distribution)=log(1+(Intra-Brand Distribution))  Equation 12
• To accommodate one or more interaction effects of independent variables with other variables, the examplepre-model data generator208 includes the addition of interaction variables based on user preferences related to model customization (block1210). For example, interaction variables may include, but are not limited to store categories that exhibit one or more predictive tendencies, such as some stores having a relatively high sales volume with diapers, and other stores having a relatively high sales volume with beer. Additionally, interaction variables related to known percentages of shoppers that are Hispanic, African American, Polish American, and/or geographic-based scores may be added.
• To accommodate one or more random coefficients typically employed in some regression models, random coefficient models, and/or variance components models, the examplepre-model data generator208 adds one or more random coefficient subject variables (block1212). Depending on the type of modeling approach selected by the user, certain assumptions and/or heuristics are applied to the coefficients that change based on the causal effects. For example, some of the random coefficient subject variables are unique based on whether the causal type is for a feature, a display, a combination of a feature and a display, or whether there is no promotional activity involved. Without limitation, other random coefficient subject variables are unique based on a feature advertisement size and/or location, such as the display being located at the end of an aisle, front of an aisle, and/or in a store lobby.
• Briefly returning toFIG. 5, the example flowchart also includes machine readable instructions that may be executed to perform trading area calculations (block508). Turning toFIG. 13, an example flowchart ofblock508 is shown in greater detail and begins atblock1302, at which the example trading areacompetitive activity assessor206 retrieves all commodities volume (ACV) data from each merchant (e.g., retailer, wholesaler, etc.). At least one source of ACV data includes the Spectra® services provided by ACNielsen® that, in part, build, manage, and maintain data sources related to store trading areas, demographics, social characteristics, and/or economic characteristics. The example trading areacompetitive activity assessor206 connects to the Spectra® services and/or any other trading area data source, which stores such trading area data in the example trading areasales data store218. Without limitation, the trading areacompetitive activity assessor206 may retrieve subsets of trading area data from the trading areasales data store218 that are relevant to the geographic region(s) of interest. Contents of the trading area ACV data include, but are not limited to, store name, store identification value(s), store characteristics (e.g., TDLinx® characteristics), geographic information, and/or geodemographic information.
• A competitive trading area table is generated by the example trading areacompetitive activity assessor206 for each target store of interest (block1304). The table includes a compiled list of competitive stores that surround the target store and which reside in the same relevant geographic area, such as the same state, the same county, and/or the same census block(s). However, the table discards other stores having the same organizational name as the target store (e.g., discard duplicate Jewel® stores that may have some census block overlap). Each census block may have varying degrees of overlap between the target store and a competitive store, thus the example trading areacompetitive activity assessor206 calculates a percentage overlap of the ACV data to identify store-pairs that exceed a threshold value (block1306) as set forth inEquation 13 below.
• 
  % ACV Overlap=(Trading Area Overlap)/(Target Store ACV)  Equation 13
• A user may select a threshold value of 0% in the event that analysis of all overlapping conditions is desired, while other users may select a higher threshold value when analysis of more significant census block overlapping conditions is desired.
• Independent trading area variables are calculated by theexample activity assessor206 to be used in one or more models, as described above. A target-store-week-UPC record is selected from the example transformed data repository222 (block1308), which was previously generated by the examplepre-model data generator208 described above. Based on the selected record, a corresponding set of competitive stores related to the target store are selected (extracted) from the table (block1310) that was generated inblock1304. For each identified competitive store related to the target store, theexample activity assessor206 calculates a percent overlap weighted mean of the inter-brand and intra-brand volumes (block1312). In effect, a resulting table includes summed values for the inter-brand and intra-brand incremental volumes for all competitive stores related to each target-store-UPC.
• To further facilitate calculation of factor-contributive sub-volumes (due-to driver scoring calculations), the example trading areacompetitive activity assessor206 calculates inter-brand store switching and intra-brand store switching values via, for example, Equations 14 and 15 below (block1314).
• $\begin{array}{cc}\mathrm{Inter}\text{-}\mathrm{Brand}\text{-}\mathrm{Store}\text{-}\mathrm{Sw}=\log \left(1+\frac{\mathrm{Total}\mathrm{Inter}\mathrm{Brand}\mathrm{Incremental}\mathrm{Vol}}{\mathrm{Base}\mathrm{Equivalent}\mathrm{Vol}}\right)& \mathrm{Equation}14\\ \mathrm{Intra}\text{-}\mathrm{Brand}\text{-}\mathrm{Store}\text{-}\mathrm{Sw}=\log \left(1+\frac{\mathrm{Total}\mathrm{Intra}\mathrm{Brand}\mathrm{Incremental}\mathrm{Vol}}{\mathrm{Base}\mathrm{Equivalent}\mathrm{Vol}}\right)& \mathrm{Equation}15\end{array}$
• In Equations 14 and 15 above, the BaseEquivalentVol is the target item base equivalent volume.
• The independent variables of Equations 14 and 15 are multiplied by model coefficients to generate intermediate store-specific variables for use in calculating factor-contributive-sub-volumes (e.g., due-to net store-switching, base store-switching, and incremental store-switching for both inter-brand and intra-brand circumstances). The example trading areacompetitive activity assessor206 retrieves an inter-brand store-switching coefficient (β_InterSS), and an intra-brand store-switching coefficient (β_IntraSS) from the model to calculate the intermediate variables BetaInterBrandStore and BetaIntraBrandStore as shown below in Equations 16 and 17.
• 
  BetaInterBrandStore=(β_InterSS)×Inter-Brand-Store-Sw  Equation 16
• 
  BetaIntraBrandStore=(β_IntraSS)×Intra-Brand-Store-Sw  Equation 17
• The example trading areacompetitive activity assessor206 next calculates the factor-contributive-sub-volumes (block1316), which represent sub-volume quantities that are attributed to (due to) one or more factors. In particular, theactivity assessor206 facilitates determining sub-volumes related to factors of competitive stores and/or same-stores (e.g., cannibalistic effects) in the trading area. In the illustrated example ofFIG. 13, the activity assessor is configured to employ Equation 18 below to calculate the net sub-volume due to intra-brand store switching. Additionally, example Equation 19 builds upon results from Equation 18 to calculate the base sub-volume due-to intra-brand store switching, and example Equation 20 further builds upon Equation 19 to calculate the incremental sub-volume due to intra-brand store switching.
• $\begin{array}{cc}\mathrm{Net}\mathrm{SubVol}\mathrm{IntraBrand}\text{Store Switching}=\text{}\hspace{1em}[\exp (\mathrm{BetaIntra}\mathrm{Brand}\mathrm{Store})-1]\times \mathrm{MAXIMUM}\lfloor \mathrm{UNITS},\mathrm{base}\mathrm{units}\rfloor & \mathrm{Equation}18\\ \mathrm{Base}\mathrm{SubVol}\mathrm{IntraBrand}\mathrm{Store}\mathrm{Switching}=\hspace{1em}\mathrm{Net}\mathrm{SubVol}\mathrm{IntraBrand}\mathrm{Store}\mathrm{Switching}\times \frac{\mathrm{base}\mathrm{units}}{\mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]}& \mathrm{Equation}19\\ \mathrm{Incremental}\mathrm{SubVol}\mathrm{IntraBrand}\mathrm{Store}\mathrm{Switching}=\left(\mathrm{Net}\mathrm{SubVol}\mathrm{IntraBrand}\mathrm{Store}\mathrm{Switching}\right)-\left(\mathrm{Base}\mathrm{SubVol}\mathrm{IntraBrand}\mathrm{Store}\mathrm{Switching}\right)& \mathrm{Equation}20\end{array}$
• In Equation 18 above, the factor-contributive sub-volume related to intra-brand net sub-volume is determined by subtracting one from the exponential value of the intermediate variable from Equation 17. The result is then multiplied by the maximum value of the corresponding UNITS volume and the corresponding base units volume for the store-week-UPC.
• While example Equations 18-20 above may be used to calculate net, base, and incremental factor-contributive sub-volumes related to intra-brand effects, the example trading areacompetitive activity assessor206 facilitates determining sub-volumes related to factors of inter-brand store switching (i.e., competitive stores). In the illustrated example ofFIG. 13, theactivity assessor206 is configured to employ Equation 21 below to calculate the net sub-volume due to inter-brand store switching. Much like the progress of example Equations 18-20 above, Equations 22 and 23 below build-upon the calculations of example Equation 21 to calculate the base sub-volume due-to inter-brand store switching and to calculate the incremental sub-volume due to inter-brand store switching, respectively.
• $\begin{array}{cc}\mathrm{Net}\mathrm{SubVol}\mathrm{InterBrand}\mathrm{Store}\mathrm{Switching}=\hspace{1em}\left[\exp \left(\mathrm{Beta}\mathrm{Inter}\mathrm{Brand}\mathrm{Store}\right)-1\right]\times \mathrm{MAXIMUM}\lfloor \mathrm{UNITS},\mathrm{base}\mathrm{units}\rfloor & \mathrm{Equation}21\\ \mathrm{Base}\mathrm{SubVol}\mathrm{InterBrand}\mathrm{Store}\mathrm{Switching}=\mathrm{Net}\mathrm{SubVol}\mathrm{InterBrand}\mathrm{Store}\mathrm{Switching}\times \frac{\mathrm{base}\mathrm{units}}{\mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]}& \mathrm{Equation}22\\ \mathrm{Incremental}\mathrm{SubVol}\mathrm{InterBrand}\mathrm{Store}\mathrm{Switching}=\left(\mathrm{Net}\mathrm{SubVol}\mathrm{InterBrand}\mathrm{Store}\mathrm{Switching}\right)-\left(\mathrm{Base}\mathrm{SubVol}\mathrm{InterBrand}\mathrm{Store}\mathrm{Switching}\right)& \mathrm{Equation}23\end{array}$
• Similar to example Equation 18, Equation 21 above determines the factor-contributive sub-volume related to inter-brand net sub-volume by subtracting one from the exponential value of the intermediate variable from Equation 16. The result is then multiplied by the maximum value of the corresponding UNITS volume and the corresponding base units volume for the store-week-UPC.
• Briefly returning toFIG. 4, the illustrated example application of prepared data and model coefficients to scoring algorithms (block406) is shown in greater detail inFIGS. 14A,14B,15A, and15B. In particular,FIGS. 14A and 14B illustrate additional detail related to base driver scoring calculations, while the example ofFIGS. 15A and 15B illustrates additional detail related to incremental trade driver scoring calculations.
• Turning toFIGS. 14A and 14B, initially, the basedriver data generator228 identifies the store, week, and product to be analyzed (block1402). For example, the store, week, and/or product may be selected by a user to specify that the user elects to analyze the sales volume of a particular product (represented by one or more unique UPCs) during a particular week and/or at a particular store (i.e., store-week-UPC sales volume), and the basedriver data generator228 can retrieve the user's selection. In some example implementations, two or more products may be analyzed simultaneously.
• The basedriver data generator228 retrieves the store-week-UPC POS sales volume data (block1404) from thesales data repository220 for each product selected atblock1402. Alternatively, the basedriver data generator228 may retrieve week-UPC sales volume data for a focus on products independent of one or more store-based effects. As also described above, in some example implementations the sales volume may correspond to the sales of a product sold in one packaging size (e.g., two-liter soda), in which case the sales volume data is retrieved based on only one unique UPC. In other instances, the sales volume may correspond to the sales of a product sold in two or more packaging sizes (e.g., two-liter bottles, 12-ounce cans, 6-ounce cans, 24-ounce bottles), in which case the sales volume data is retrieved based on two or more unique UPCs. In either case, the sales volume analyses described herein can be performed using base equivalent units by representing the sales volumes using common (e.g., normalized), standard volumetric units (or piece-count units). For example, if a soda product is sold in different packaging sizes, the sales volume analyses can convert the POS sales volume data for each packaging type into ounces such that a sale of a two-liter bottle would represent a sale of sixty-eight ounces of soda based on a base equivalent measure, while a sale of a six-pack of twelve-ounce cans would represent a sale of seventy-two ounces of soda based on the same base equivalent measure.
• The basedriver data generator228 then retrieves model coefficients (fixed-effect coefficients and random-effect coefficients) and net interaction terms from the model coefficients repository224 (block1406) and sets any non-significant fixed effect coefficients and non-significant random effect coefficients to zero (block1408). The basedriver data generator228 then calculates net store-UPC coefficients (block1410) based on the remaining, significant fixed and random effect coefficients and net interaction terms. In the illustrated example, net store-UPC coefficients are analysis-specific coefficients that represent different sales volumes effects for products in a particular store. While the model coefficients can be used to represent higher granularities of different effects on sales volumes, each net store-UPC coefficient can represent two or more aggregated model coefficients having similar sales volumes effects. In the illustrative example, the net store-UPC coefficients include a logarithmic regular price coefficient (β_RP), a lift coefficient (β_L), an inter-brand trade coefficient (β_interT), an intra-brand trade coefficient (β_intraT), an inter-brand price coefficient (β_interP), an intra-brand price coefficient (β_intraP), an inter-brand distribution coefficient (β_interD), and an intra-brand distribution coefficient (β_intraD).
• The basedriver data generator228 can determine the net store-UPC coefficients atblock1410 by aggregating the corresponding fixed and random coefficients and also applying net interaction terms to quantify the effects that the interactions between different coefficients have on the resulting net store-UPC coefficients. In example implementations in which two or more products are selected atblock1402, the operation ofblock1406 can retrieve model coefficients for each product, and the operation ofblock1410 can be performed for each product to determine respective net store-UPC coefficients for each product based on its respective coefficients.
• The basedriver data generator228 then discards any net store-UPC coefficient value that does not comply with a constraints table (block1412). The constraints table can be used to store rules to be satisfied or met by calculated net store-UPC coefficients to determine whether the net store-UPC coefficients are statistically correct. In the illustrated example, a constraints table is used to store positive indicators and negative indicators to indicate whether it is plausible for a coefficient to represent an increase in sales. For example, if a constraint is indicative of positive growth (which is typically the case when a promotion is run) and a net store-UPC coefficient calculated atblock1410 represents a decrease in sales, then the calculated net store-UPC coefficient is discarded atblock1412 as a statistical outlier. Inconsistent or statistical outlier net store-UPC coefficients may be attributed to missing sales data, incorrectly recorded or collected data or any other error or misstep in the data handling process. Discarding inconsistent net store-UPC coefficients atblock1412 reduces the chances of incorrectly influencing volume sales analyses.
• The basedriver data generator228 retrieves the independent logarithmic value for the non-promotion, regular price (LREGPR) of each UPC selected at block1402 (block1414). In the illustrated example, the basedriver data generator228 retrieves the independent variable LREGPR for each UPC from the transformeddata repository222. The basedriver data generator228 also retrieves independent variable decremental effect factors from the transformed data repository222 (block1416). Independent variable decremental volume factors (sometimes referred to herein as decremental effect factors) are factors having a decremental effect on the sales volume(s) calculated atblock1410 for the product(s) selected atblock1402. In the illustrated example, the independent variable decremental effect factors include, but are not limited to pantry loading (cannibalization) terms LIFT1=>LIFTn, inter-brand weighted category trade factors (INTER_TRADE), intra-brand weighted category factors (INTRA_TRADE), inter-brand weighted regular price factors (INTER_PRICE), intra-brand weighted regular price factors (INTRA_PRICE), inter-brand weighted distribution factors (LINTER_DIST), and intra-brand weighted distribution factors (LINTRA_DIST).
• The basedriver data generator228 determines mean-center values for independent pre-selected independent variables (1418). In the illustrated example, the pre-selected independent variables include LREGPR, INTER_PRICE, INTRA_PRICE, LINTER_DIST, and LINTRA_DIST. The basedriver data generator228 also retrieves the reference base unit volume (MEAN_BASE) for each store-week-UPC identified at block1402 (block1420 ofFIG. 14B). The reference base unit volume for each product is the mean sales volume for that product over a predetermined duration (e.g., a plurality of weeks) and is representative of typical sales of that product over that duration.
• The basedriver data generator228 and/or the scoring processor determines the products of net store-UPC coefficients and respective independent variables to generate decremental effect terms (block1422). The net store-UPC coefficients are selected from those determined above atblock1410 and the independent variables are those retrieved at block1416 (some of which were mean-centered at block1418). Thus, the product values (e.g., store-UPC coefficient terms, decremental effect terms), which may be represented as decremental effect terms are determined by the basedriver data generator228 are, for example, (β_RP×LREGPR), (β_interT×INTER_TRADE), (β_intraT×INTRA_TRADE), (β_interP×INTER_PRICE), (β_intraP×INTRA_PRICE), (β_interD×LINTER_DIST), (β_intraD×LINTRA_DIST), and (β_LIFT1×LIFT1) through (β_LIFTn×LIFTn). In addition, a pantry loading store-UPC coefficient term (β_pantry×PANTRY), which may be represented as a pantry loading effect term, can be determined by summing the LIFT store-UPC coefficient terms ((β_LIFT1×LIFT1 . . . β_LIFTn×LIFTn).
• The basedriver data generator228 determines factor-contributive sub-volumes (block1424). Factor-contributive sub-volumes represent sub-volume quantities attributed to (or due to) respective factors that, when aggregated, account for an entire dip (decrease) or spike (increase) in sales volumes. For example, factors A, B, and C correspond to respective ones of factor-A sub-volume, factor-B sub-volume, and factor-C sub-volume that, when summed, quantify an entire observed volume (i.e., total observed volume=sum(factor-A sub-volume, factor-B sub-volume, factor-C sub-volume). In the illustrated example, the basedriver data generator228 can be configured to use Equation 24 below to determine a regular price factor-contributive sub-volume, an inter-brand weighted regular price factor-contributive sub-volume, an intra-brand weighted regular price factor-contributive sub-volume, an inter-brand distribution factor-contributive sub-volume, and an intra-brand distribution factor-contributive sub-volume.
• $\begin{array}{cc}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\hspace{1em}\left[1-\frac{1}{\exp \left(\mathrm{store}\text{-}\mathrm{UPC}\mathrm{coeff}\mathrm{term}\right)}\right]\times \mathrm{MEAN\_BASE}& \mathrm{Equation}24\end{array}$
• In Equation 24 above, a factor-contributive sub-volume is determined by subtracting one from the inverse value of the exponential value of a corresponding store-UPC coefficient term determined above atblock1422. The result is then multiplied by the reference base unit volume (MEAN_BASE) for a corresponding store-week-UPC.
• The basedriver data generator228 determines pantry loading effect sub-volumes (block1426) as set forth in Equation 25 below. For example, the basedriver data generator228 can be configured to use Equation 25 below to determine a pantry loading net factor-contributive sub-volume.
• $\begin{array}{cc}\mathrm{pantry}\mathrm{loading}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\hspace{1em}\left[1-\frac{1}{\exp \left({\beta }_{\mathrm{pantry}}\times \mathrm{PANTRY}\right)}\right]\times \mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]& \mathrm{Equation}25\end{array}$
• In Equation 25 above, the pantry loading net factor-contributive sub-volume is determined by subtracting one from the inverse of the exponential value of the pantry loading store-UPC coefficient term (β_pantry×PANTRY). The resulting value is then multiplied by the maximum value of the corresponding UNITS volume and the corresponding base units volume for the store-week-UPC.
• In addition, the basedriver data generator228 can be configured to determine a pantry loading base factor-contributive sub-volume using Equation 26 below and a pantry loading incremental factor-contributive sub-volume using Equation 27 below.
• $\begin{array}{cc}\mathrm{pantry}\mathrm{loading}\mathrm{base}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\mathrm{pantry}\mathrm{loading}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\times \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\left[\frac{\mathrm{base}\mathrm{units}}{\mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]}\right]& \mathrm{Equation}26\\ \mathrm{pantry}\mathrm{loading}\mathrm{incremental}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\left(\mathrm{pantry}\mathrm{loading}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\right)-=\left(\mathrm{pantry}\mathrm{loading}\mathrm{base}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\right)& \mathrm{Equation}27\end{array}$
• In Equation 26 above, the pantry loading base factor-contributive sub-volume is determined by multiplying the pantry loading net factor-contributive sub-volume by the quotient of corresponding base units volume divided by the maximum value of the corresponding UNITS volume and the corresponding base units volume for the store-week-UPC. In Equation 27 above, the pantry loading incremental factor-contributive sub-volume is determined by subtracting the pantry loading net factor-contributive sub-volume from the pantry loading base factor-contributive sub-volume.
• The basedriver data generator228 then determines competitive trade effects sub-volumes (block1428). For example, the basedriver data generator228 can use Equations 28, 29, and 30 below, which are similar to Equations 25-27 above.
• $\begin{array}{cc}\mathrm{inter}\text{-}\mathrm{trade}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\left[1-\frac{1}{\exp \left({\beta }_{\mathrm{interT}}\times \mathrm{INTER\_TRADE}\right)}\right]\times \mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]& \mathrm{Equation}28\\ \mathrm{inter}\text{-}\mathrm{trade}\mathrm{base}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\mathrm{inter}\text{-}\mathrm{trade}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\times \left[\frac{\mathrm{base}\mathrm{units}}{\mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]}\right]& \mathrm{Equation}29\\ \mathrm{inter}\text{-}\mathrm{trade}\mathrm{incremental}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\left(\mathrm{inter}\text{-}\mathrm{trade}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\right)-\left(\mathrm{inter}\text{-}\mathrm{trade}\mathrm{base}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\right)& \mathrm{Equation}30\end{array}$
• The basedriver data generator228 then determines same brand trade cannibalization effects sub-volumes (block1430). Cannibalization occurs when a company offers a new product or service that takes sales or market share from one or more of its already existing products or services. The basedriver data generator228 can be configured to determine the same brand trade cannibalization effectssub-volumes using Equations 31, 32, and 33 below, which are similar to Equations 25-27 above.
• $\begin{array}{cc}\mathrm{intra}\text{-}\mathrm{trade}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\left[1-\frac{1}{\exp \left({\beta }_{\mathrm{intraT}}\times \mathrm{INTRA\_TRADE}\right)}\right]\times \mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]& \mathrm{Equation}31\\ \mathrm{intra}\text{-}\mathrm{trade}\mathrm{base}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\mathrm{intra}\text{-}\mathrm{trade}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\times \left[\frac{\mathrm{base}\mathrm{units}}{\mathrm{MAXIMUM}\left[\mathrm{UNITS},\mathrm{base}\mathrm{units}\right]}\right]& \mathrm{Equation}32\\ \mathrm{intra}\text{-}\mathrm{trade}\mathrm{incremental}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}=\left(\mathrm{intra}\text{-}\mathrm{trade}\mathrm{net}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\right)-\left(\mathrm{intra}\text{-}\mathrm{trade}\mathrm{base}\mathrm{factor}\text{-}\mathrm{contributive}\mathrm{sub}\text{-}\mathrm{volume}\right)& \mathrm{Equation}33\end{array}$
• If the store identified atblock1402 is a causal sample store, the basedriver data generator228 determines the aggregate raw trade driver volume for the identified product(s) being analyzed (block1432). A causal sample store is a store for which causal effect data and sales volumes are collected, while a volumetric (census) store is one for which only sales volumes are collected. The causal effect data collected at causal sample stores can be used to infer reasons for sales volumes at volumetric stores. The basedriver data generator228 can be configured to determine the aggregate raw trade driver volume using Equation 34 below.
• 
  raw trade driver volume=sum(pantry loading incremental factor-contributive sub-volume, inter-trade incremental factor-contributive sub-volume, intra-trade incremental factor-contributive sub-volume)  Equation 34
• As shown in Equation 34 above, the raw trade driver volume is the sum of the pantry loading incremental factor-contributive sub-volume determined in Equation 27 above, the inter-trade incremental factor-contributive sub-volume determined in Equation 30 above, and the intra-trade incremental factor-contributive sub-volume determined in Equation 31 above.
• The basedriver data generator228 determines the total category base equivalent units volume by store-week (block1434) for a particular product. Base equivalent volumes are calculated to determine total product sold regardless of the products' packaging. In this manner, for example, regardless of whether brand A soda was sold in a two-liter bottle package or twelve-ounce cans, the base equivalent volume will reflect the total ounces of the product sold regardless of packaging. The basedriver data generator228 then determines the average category base equivalent units volume for all weeks combined by store (block1436) for the particular product. For example, the basedriver data generator228 can use the total category base equivalent units volume determined atblock1434 to determine the mean of total category base equivalent units volume to calculate the average category base equivalents units volume atblock1436.
• The basedriver data generator228 then determines a category index of a particular product (block1438) by, for example, dividing the total category base equivalent units volume determined atblock1434 by the average category base equivalent units volume determined atblock1436. The basedriver data generator228 then determines a market driver volume (block1440) using, for example, seasonality factors, trend factors, etc.
• The basedriver data generator228 then determines the base residual volume of the product (block1442) using, for example, Equation 35 below.
• 
  base residual volume=base units−sum(regular price factor-contributive sub-volume, inter-brand weighted regular price factor-contributive sub-volume, intra-brand weighted regular price factor-contributive sub-volume, inter-brand distribution factor-contributive sub-volume, intra-brand distribution factor-contributive sub-volume, pantry loading base factor-contributive sub-volume, inter-brand trade factor-contributive sub-volume, intra-brand trade factor-contributive sub-volume, other driver volume(s))  Equation 35
• As shown in Equation 35 above, the base residual volume of a product is determined by subtracting the base units from the sum of the regular price factor-contributive sub-volume, the inter-brand weighted regular price factor-contributive sub-volume, the intra-brand weighted regular price factor-contributive sub-volume, the inter-brand distribution factor-contributive sub-volume, the intra-brand distribution factor-contributive sub-volume, the pantry loading base factor-contributive sub-volume, the inter-brand trade factor-contributive sub-volume, the intra-brand trade factor-contributive sub-volume, and any other driver volume(s).
• The basedriver data generator228 then determines a distribution volume of the product (block1444). For example, the base driver data generator can determine the distribution volume of the product by determining the mean of the base residual volume of the product determined atblock1442. The example process ofFIGS. 14A and 14B is then ended.
• Turning now toFIGS. 15A and 15B, initially, the incrementaldriver data generator230 identifies the store, week, and product to be analyzed (block1502). For example, the store, week, and product may be selected by a user to specify that the user elects to analyze the sales volume of a particular product (represented by one or more unique UPCs) during a particular week at a particular store (i.e., store-week-UPC sales volume), and the incrementaldriver data generator230 can retrieve the user's selection. In some example implementations, two or more products may be analyzed simultaneously.
• The incrementaldriver data generator230 retrieves the store-week-UPC POS sales volume data (block1504) from thesales data repository220 for each product selected atblock1502. As discussed above in connection withblock1404 ofFIG. 14A, the sales volume analyses described herein can be performed using base equivalent units by representing the sales volumes using common, standard volumetric units (or piece-count units) regardless of whether the retrieved sales volume corresponds to the sales of a product sold in one packaging size to the sales of a product sold in two or more different packaging sizes.
• The incrementaldriver data generator230 then retrieves model coefficients (fixed-effect coefficients and random-effect coefficients) and net interaction terms from the model coefficients repository224 (block1506) and sets any non-significant fixed effect coefficient and non-significant random effect coefficient to zero (block1508). The incrementaldriver data generator230 then calculates net store-UPC coefficients (block1510) based on the remaining, significant fixed and random effect coefficients and net interaction terms. In the illustrative example, the net store-UPC coefficients include a logarithmic promoted price index coefficient (β_LPPI), a temporary price reduction (TPR) coefficient (β_TPR), a feature coefficient (β_feat), a display coefficient (β_disp), and a feature-display coefficient (β_featdisp).
• The incrementaldriver data generator230 can determine the net store-UPC coefficients atblock1510 by aggregating corresponding fixed and random coefficients and also applying net interaction terms to quantify the effects that the interactions between different coefficients have on the resulting net store-UPC coefficients. In example implementations in which two or more products are selected atblock1502, the operation ofblock1506 can retrieve model coefficients for each product, and the operation ofblock1510 can be performed for each product to determine respective net store-UPC coefficients for each product based on it respective coefficients.
• The incrementaldriver data generator230 then discards any net store-UPC coefficient value that does not comply with a constraints table (block1512). The constraints table used atblock1512 is configured similarly to the constraints table described above in connection withblock1412 ofFIG. 14A. Discarding inconsistent net store-UPC coefficients atblock1512 reduces the chances of incorrectly influencing volume sales analyses.
• The incrementaldriver data generator230 determines N reference fact values (block1514). Each reference fact value corresponds to a different store-week-UPC identified atblock1502. Thus, N is equal to the number of store-week-UPC's used for the sales volume analysis. The reference fact values can be calculated once per year or at any other desired time. In the illustrated example, the reference fact values include a reference base equivalent (REFBE) units and a plurality of reference promoted price indices (PPI's). REFBE units represent the average base equivalent units sales volume for a particular UPC (or group of UPC's representing the same product sold in different packaging sizes) in a particular store. Base equivalent units are discussed above in connection withblock1404 ofFIG. 14A.
• The plurality of PPI's are neither store-specific nor UPC-specific, but instead represent ratios of actual prices-to-regular prices. The reference PPI's can be used to determine when promotions are present for different products. For example, during an analysis if the ratio of an actual price-to-regular price of a product is less than or equal to the reference PPI, it confirms that there is a promotion for that product. Otherwise, a promotion is not present when the actual price-to-regular price ratio is not less than or equal to the reference PPI. Each reference PPI corresponds to a corresponding type of promotion including a feature only promotion, a display only promotion, a hybrid feature-display promotion, or a price cut only promotion. A feature promotion is a circulated flyer (e.g., coupons or advertisements in newspapers, magazines, etc.) associated with the product. A display promotion is an in-store display (e.g., in-store posters, shelf talkers, coupon dispensers, etc.) promoting the product. A price cut promotion occurs when a product's price is reduced but is not advertised.
• The incrementaldriver data generator230 determines an incremental equivalent volume (block1516) and raw driver volumes for causal promotion types (block1518). In the illustrated example, the incrementaldriver data generator230 determines the incremental equivalent volume and the raw driver volumes for causal promotion types based on the net store-UPC coefficients (i.e., logarithmic promoted price index coefficient (β_LPPI), temporary price reduction (TPR) coefficient (β_TPR), feature coefficient (β_feat), display coefficient (β_disp), and feature-display coefficient (β_featdisp)) determined above atblock1510 and certain facts data for the product UPC(s) under analysis retrieved from the transformed data repository222 (FIG. 2). The facts data include equivalent volume, base equivalent volume (BEV), actual price, regular price, causal promotion indicators (e.g., variable CAUSAL having example values 0-7), and temporary price reduction (TPR) flags (e.g., 1 if price cut, 0 otherwise). The facts data also include the reference facts determined above atblock1514. While the illustrated examples described herein use variable CAUSAL to indicate causal promotional activity, any other variable nomenclature and/or associated value(s) may be implemented to represent the same.
• To determine the incremental equivalent volume atblock1516, the incrementaldriver data generator230 subtracts the base equivalent volume from the equivalent volume (i.e., incremental equivalent volume=equivalent volume−base equivalent volume).
• The raw driver volumes by causal promotion types determined atblock1518 include a price cut only raw driver volume, a feature only raw driver volume, a display only raw driver volume, and a feature-display raw driver volume. Atblock1518, the incrementaldriver data generator230 determines that a price cut only promotion exists for a product when its causal promotion indicator (e.g., CAUSAL) is equal to zero (i.e., CAUSAL=0) and a temporary price reduction flag is equal to one (i.e., TPR=1 to identify the occurrence of a temporary price reduction). If the price cut only promotion exists, then the incrementaldriver data generator230 determines a price cut only raw quantity and price driver volume using Equations 36 and 37 below.
• 
  price cut only raw quantity driver volume={exp(β_TPR)×[(reference PPI for price cut only)̂β_LPPI]−1}×REFBE  Equation 36
• 
  price cut only raw price driver volume={exp(β_TPR)×[(actual price/regular price)̂β_LPPI]−1}×REFBE−price cut only raw quantity driver volume  Equation 37
• Also atblock1518, the incrementaldriver data generator230 determines that a feature only promotion exists for a product when its causal promotion indicator (e.g., CAUSAL) is equal to a value of one, two, or three (i.e., CAUSAL=1, 2, or 3). If a feature only promotion exists, then the incrementaldriver data generator230 determines a feature only raw quantity and price driver volume using Equations 38 and 39 below.
• 
  feature only raw quantity driver volume={exp(β_feat)×[(reference PPI for feature only)̂β_LPPI]−1}×REFBE  Equation 38
• 
  feature only raw price driver volume={exp(β_feat)×[(actual price/regular price)̂β_LPPI]−1}×REFBE−feature only raw quantity driver volume  Equation 39
• Also atblock1518, the incrementaldriver data generator230 determines that a display only promotion exists for a product when its causal promotion indicator (e.g., CAUSAL) is equal to a value of four (i.e., CAUSAL=4). If a display only promotion exists, then the incrementaldriver data generator230 determines a display only raw quantity and price driver volume using Equations 40 and 41 below.
• 
  display only raw quantity driver volume={exp(β_disp)×[(reference PPI for display only)̂β_LPPI]−1}×REFBE  Equation 40
• 
  display only raw price driver volume={exp(β_disp)×[(actual price/regular price)̂β_LPPI]−1}×REFBE−display only raw quantity driver volume  Equation 41
• Also atblock1518, the incrementaldriver data generator230 determines that a hybrid feature-display promotion exists for a product when its causal promotion indicator (e.g., CAUSAL) is equal to a value of five, six, or seven (i.e., CAUSAL=5, 6, or 7). If a hybrid feature-display promotion exists, then the incrementaldriver data generator230 determines a hybrid feature-display raw quantity and price driver volume using Equations 42 and 43 below.
• 
  feature-display raw quantity driver volume={exp(β_featdisp)×[(reference PPI for feature-display)̂β_LPPI]−1}×REFBE  Equation 42
• 
  feature-display raw price driver volume={exp(β_featdisp)×[(actualprice/regularprice)̂β_LPPI]−1}×REFBE−feature-display raw quantity driver volume  Equation 43
• The incrementaldriver data generator230 then determines the raw net price and quantity volume (block1520) based on the raw driver volumes for each causal promotion type determined above atblock1518. In the illustrative example, the incrementaldriver data generator230 determines the raw net price and quantity volume using Equation 44 below.
• 
  raw net price and quantity volume=sum(price cut only raw quantity driver volume, price cut only raw price driver volume, feature only raw quantity driver volume, feature only raw price driver volume, display only raw quantity driver volume, display only raw price driver volume, feature-display raw quantity driver volume, feature-display raw price driver volume)  Equation 44
• The incrementaldriver data generator230 then determines the raw base effectiveness volume (block1522) based on the raw net price and quantity volume determined atblock1520 using, for example, Equation 46 below.
• 
  raw base effectiveness volume=(BE/REFBE−1)×raw net price and quantity volume  Equation 45
• The incrementaldriver data generator230 then determines the adjusted incremental volume (block1524) using, for example, Equation 46 below.
• 
  adjusted incremental volume=(incremental units)−(pantry loading incremental factor-contributive sub-volume)−(inter-brand incremental factor-contributive sub-volume)−(intra-brand incremental factor-contributive sub-volume)  Equation 46
• The pantry loading incremental factor-contributive sub-volume, the inter-brand incremental factor-contributive sub-volume, and the intra-brand incremental factor-contributive sub-volume used in Equation 46 above are determined as described above in connection withblock1426,1428, and1430 ofFIG. 14B.
• The incrementaldriver data generator230 then determines a raw trade driver volume (block1526) using, for example, Equation 47 below based on the raw net price and quantity volume determined atblock1520 and the raw base effectiveness volume determined atblock1522.
• 
  raw trade driver volume=raw net price and quantity volume+raw base effectiveness volume  Equation 47
• Also for each store with a promotion condition, the incrementaldriver data generator230 determines a lift factor adjustment (block1528) for each UPC using, for example, dividing the adjusted incremental volume for that UPC (determined above at block1524) by the total raw trade driver volume of that UPC (i.e., lift factor adjustment=adjusted incremental volume/total raw trade driver volume).
• The incrementaldriver data generator230 determines final trade driver volumes for each causal promotion type (block1530). In the illustrative example, the incrementaldriver data generator230 can be configured to use Equations 48 and 49 below to determine the final trade driver volumes for each causal promotion type.
• 
  {promotion type}quantity driver volume=(lift factor adjustment)×({promotion type}quantity raw driver volume)  Equation 48
• 
  {promotion type}price driver volume=(lift factor adjustment)×({promotion type}price raw driver volume)  Equation 49
• In Equations 48 and 49 above, promotion type is one of price cut only, feature only, display only, or feature-display. The lift factor adjustment is determined above atblock1528, the {promotion type} quantity raw driver volumes for the different promotion types are determined above atblock1518, and the {promotion type} price raw driver volumes for the different promotion types are also determined above atblock1518.
• The incrementaldriver data generator230 determines a base effectiveness volume (block1532) by, for example, multiplying the lift factor adjustment (determined at block1528) by the raw base effectiveness volume (determined at block1522) (i.e., (base effectiveness volume)=(liftfactor adjustment)×(raw base effectiveness volume)). The incrementaldriver data generator230 then determines the effectiveness volume for each causal promotion type (block1534) using, for example, Equation 50 below.
• 
  {promotion type}effectiveness volume=(incremental equivalent volume)−(base effectiveness volume+{promotion type}quantity driver volume+{promotion type}price driver volume)  Equation 50
• In Equation 50 above, the incremental equivalent volume is determined above atblock1516, the base effectiveness volume is determined above atblock1532, and the {promotion type} quantity driver volume and {promotion type} price driver volume are determined above atblock1530. The example process ofFIGS. 15A and 15B is then ended.
• FIG. 16 is a block diagram of an example processor system that may be used to implement some or all of the example methods and apparatus described herein. As shown inFIG. 16, theprocessor system1610 includes aprocessor1612 that is coupled to aninterconnection bus1614. Theprocessor1612 includes a register set or registerspace1616, which is depicted inFIG. 16 as being entirely on-chip, but which could alternatively be located entirely or partially off-chip and directly coupled to theprocessor1612 via dedicated electrical connections and/or via theinterconnection bus1614. Theprocessor1612 may be any suitable processor, processing unit or microprocessor. Although not shown inFIG. 16, thesystem1610 may be a multi-processor system and, thus, may include one or more additional processors that are identical or similar to theprocessor1612 and that are communicatively coupled to theinterconnection bus1614.
• Theprocessor1612 ofFIG. 16 is coupled to achipset1618, which includes amemory controller1620 and an input/output (I/O)controller1622. As is well known, a chipset typically provides I/O and memory management functions as well as a plurality of general purpose and/or special purpose registers, timers, etc. that are accessible or used by one or more processors coupled to thechipset1618. Thememory controller1620 performs functions that enable the processor1612 (or processors if there are multiple processors) to access asystem memory1624 and amass storage memory1625.
• Thesystem memory1624 may include any desired type of volatile and/or non-volatile memory such as, for example, static random access memory (SRAM), dynamic random access memory (DRAM), flash memory, read-only memory (ROM), etc. Themass storage memory1625 may include any desired type of mass storage device including hard disk drives, optical drives, tape storage devices, etc.
• The I/O controller1622 performs functions that enable theprocessor1612 to communicate with peripheral input/output (I/O)devices1626 and1628 and anetwork interface1630 via an I/O bus1632. The I/O devices1626 and1628 may be any desired type of I/O device such as, for example, a keyboard, a video display or monitor, a mouse, etc. Thenetwork interface1630 may be, for example, an Ethernet device, an asynchronous transfer mode (ATM) device, an 802.11 device, a DSL modem, a cable modem, a cellular modem, etc. that enables theprocessor system1610 to communicate with another processor system.
• While thememory controller1620 and the I/O controller1622 are depicted inFIG. 16 as separate functional blocks within thechipset1618, the functions performed by these blocks may be integrated within a single semiconductor circuit or may be implemented using two or more separate integrated circuits.
• Although the above description refers to the flowcharts as being representative of methods, those methods may be implemented entirely or in part by executing machine readable instructions. Therefore, the flowcharts are representative of methods and machine readable instructions. Additionally, one or more methods may be implemented entirely or in part by executing example pseudo code instructions, such as the example pseudo code instructions ofFIGS. 17A-17E,18A,18B,19A,19B,20A,20B, and21A through21C. For example,FIGS. 17A-17E generally illustrate example pseudo code to prepare data for modeling in connection with the example system ofFIG. 1. Additionally, example block708 ofFIG. 7A andFIG. 8 may be represented by examplepseudo code section1702 ofFIG. 17A, example block710 ofFIG. 7A may be represented by the examplepseudo code section1704 ofFIG. 17A, and example block712 ofFIG. 7A andFIG. 9 may be represented by the examplepseudo code section1706 ofFIG. 17A. Further, example block718 ofFIG. 7A andFIG. 10 may be represented by examplepseudo code section1708 ofFIG. 17B, and example block720 ofFIG. 7B andFIG. 11 may be represented by examplepseudo code section1710 ofFIG. 17B. Moreover, example block722 ofFIG. 7B andFIG. 12 may be represented by examplepseudo code section1712 ofFIG. 17C.
• FIGS. 18A and 18B generally illustrate example pseudo code to prepare data for store switching calculations in connection with the example system ofFIG. 1. For example, example blocks1314 and1316 ofFIG. 13 may be represented by examplepseudo code section1802 ofFIG. 18B.FIGS. 19A and 19B generally illustrate example pseudo code to specify model parameters in connection with the example system ofFIG. 1.FIGS. 20A and 20B generally illustrate example pseudo code to perform base driver scoring calculations in connection with the example system ofFIG. 1. For instance, example blocks1416 and1418 ofFIG. 14A may be represented by examplepseudo code section2002 ofFIG. 20A, example blocks1420 and1422 ofFIG. 14B may be represented by examplepseudo code section2004 ofFIG. 20A, andexample blocks1424 and1426 ofFIG. 14B may be represented by examplepseudo code section2006 ofFIG. 20A. Additionally, example blocks1428,1430,1432,1434, and1436 may be represented by examplepseudo code section2008 ofFIG. 20B.FIGS. 21A-21C generally illustrate example pseudo code to perform incremental trade driver scoring calculations in connection with the example system ofFIG. 1.
• Although certain methods, apparatus, and articles of manufacture have been described herein, the scope of coverage of this patent is not limited thereto. To the contrary, this patent covers all methods, apparatus, and articles of manufacture fairly falling within the scope of the appended claims either literally or under the doctrine of equivalents.

Claims (25)

1. A computer implemented method to identify a competitive store effect in response to a trade promotion at a target store, comprising:

identifying a target product sold at the target store associated with the trade promotion;

identifying a plurality of competitive products having a degree of similarity to the target product;

retrieving all commodities volume (ACV) data from a plurality of stores in a geography of interest that sell the plurality of competitive products;

calculating a percentage overlap of the ACV data to generate a competitive trading area table of store pairs within a threshold value of the overlap;

calculating a brand store switching value by dividing a brand incremental volume by a base equivalent volume for the store pairs, the brand store switching value indicative of the trade promotion store effect; and

generating a report including the brand store switching value.

2. A method as defined inclaim 1, further comprising applying an intra brand store switching coefficient to calculate a factor contributive sub volume of the target product at the target store based on the trade promotion at the target store.

3. A method as defined inclaim 2, wherein the applied intra brand store switching coefficient is generated via a hierarchical linear modeling technique.

4. A method as defined inclaim 2, wherein the applied intra brand store switching coefficient is generated via at least one of a linear regression or a multiple linear regression technique.

5. A method as defined inclaim 1, further comprising applying an inter brand store switching coefficient to calculate a factor contributive sub volume of the target product at a competitive store in the plurality of stores in the geography of interest, the inter brand store switching coefficient based on the trade promotion at the target store.

6. A method as defined inclaim 5, wherein the applied inter brand store switching coefficient is generated via a hierarchical linear modeling technique.

7. A method as defined inclaim 5, wherein the applied inter brand store switching coefficient is generated via at least one of a linear regression or a multiple linear regression technique.

8. A method as defined inclaim 1, wherein calculating the brand store switching value further comprises calculating a base volume of sales for the target store during a time period.

9. A method as defined inclaim 8, further comprising calculating base factor contributive sub volumes during the time period to identify at least one of a sales increase or a sales decrease.

10. A method as defined inclaim 9, wherein the base factor contributive sub volumes are based on at least one of a regular price factor or a distribution factor.

11. A method as defined inclaim 9, further comprising calculating a base residual volume indicative of the difference between the base equivalent volume and the base factor contributive sub volumes adjusted by at least one of seasonality factors or trend factors.

12. A method as defined inclaim 11, further comprising determining an incremental volume at the target store caused by the trade promotion by calculating the difference between actual sales and the base residual volume.

13. A method as defined inclaim 12, wherein the incremental volume is based on a coefficient associated with a combination of the target store and a universal product code (UPC) associated with the target product.

14. A method as defined inclaim 13, wherein the coefficient comprises at least one of a promoted price coefficient associated with the UPC, a temporary price reduction coefficient associated with the UPC, a feature coefficient, or a display coefficient.

15. An apparatus to identify a competitive store effect in response to a trade promotion at a target store, comprising:

a product reference interface to identify a target product sold at a target store, the target product associated with the trade promotion;

a target product comparator to identify a plurality of competitive products having a degree of similarity to the target product;

a promotion analysis engine to retrieve all commodities volume (ACV) data from a plurality of stores in a geography of interest;

a trading area competitive activity assessor to calculate a percentage overlap of the ACV data and to generate a competitive trading area table of store pairs within a threshold percentage value, the store pairs containing the identified plurality of competitive products; and

a pre-model data generator to calculate a brand store switching value by dividing a brand incremental volume by a base equivalent volume for the store pairs, the brand store switching value indicative of the trade promotion store effect.

16. An apparatus as defined inclaim 15, further comprising a model coefficient estimator to apply an intra brand store switching coefficient to calculate a factor contributive sub volume of the target product at the target store based on the trade promotion at the target store.

17. An apparatus as defined inclaim 16, further comprising a hierarchical linear model to generate the intra brand store switching coefficient.

18. An apparatus as defined inclaim 16, further comprising at least one of a linear regression model or a multiple linear regression model to generate the intra brand store switching coefficient.

19. An apparatus as defined inclaim 15, further comprising a model coefficient estimator to apply an inter brand store switching coefficient to calculate a factor contributive sub volume of the target product at a competitive store in the plurality of stores in the geography of interest, the inter brand store switching coefficient based on the trade promotion at the target store.

20. (canceled)

21. (canceled)

22. An apparatus as defined inclaim 15, further comprising a trading area sales data store to identify a plurality of census blocks associated with the store pairs of the competitive trading area table.

23. (canceled)

24. An article of manufacture storing machine accessible instructions that, when executed, cause a machine to:

identify a target product sold at the target store associated with the trade promotion;

identify a plurality of competitive products having a degree of similarity to the target product;

retrieve all commodities volume (ACV) data from a plurality of stores in a geography of interest that sell the plurality of competitive products;

calculate a percentage overlap of the ACV data to generate a competitive trading area table of store pairs within a threshold value of the overlap;

calculate a brand store switching value by dividing a brand incremental volume by a base equivalent volume for the store pairs, the brand store switching value indicative of the trade promotion store effect; and

generate a report including the brand store switching value.

25-33. (canceled)

Images