- Gross Impressions (“GI,” the sum of individual AQH numbers, rolled up across stations and dayparts);
- CPM (Cost Per Thousand=price/GI*1000);
- GRP (Gross Rating Points=GI/population*100);
- CPP (Cost Per Point=price/GRP);
- Reach (can be derived in conjunction withFIG. 6 below);
- % Market Reach (Reach*100/population);
- Frequency (GI/Reach); [0035] CPMNR (CPM Net Reach=price/Reach*1000);
- AQH (for the daypart in question, does not depend on spot count);
- AQH Rating (AQH/population*100);
- Cume (for the daypart in question, does not depend on spot count);
- Cume Rating (Cume/population*100).

AudiencePro engine

45 can maintain this output data and more (including input data and intermediate data) in multi-dimensional arrays or matrices with different variables indexed as appropriate, such as by book, market, demographic, station, Arbitron daypart, or client daypart.

As explained further below, in order to produce theseoutputs AudiencePro engine45 executes several calculations including computing the Cume of an arbitrary daypart from the Cume of a limited set of dayparts, and computing a Cume for multiple stations from Cume values of individual stations. The outputs ofAudiencePro engine45 are accordingly dependent on various user inputs, such as the list of stations, demographic, flight dates and schedule for each desired station (including specific dayparts, spot counts, and any weighting adjustments).

FIG. 3 is a flow chart for a process of translating and rolling up AQH and Cume values to calculate the various outputs ofAudiencePro engine45 and particularly Reach, in accordance with one implementation of the present invention. InFIG. 3 and the figures that follow, AQH is represented by the letter “S”, and Cume is represented by the letter “C”. The radio station ratings analysis process begins whencomputer system10 receives information from the various sources, particularly S and C of all available stations for limited (Arbitron or Nielson) dayparts, a station list, and a proposed schedule including client (AudiencePro) dayparts (51). The radio station ratings application of the present invention can consolidate the data streams from multiple vendors such as Arbitron and Nielson to present a single interface to the user. In this manner the system can switch between sets of input data to provide output based on a selected vendor. The input S values for each client station are then translated from the limited dayparts to the client dayparts (52). If the client daypart is contained in an input daypart, then the S value from that input daypart will simply be used for the translation. If instead the client daypart is not contained in an input daypart, then the smallest set of input dayparts that contain the client daypart will be used for the translation. In this case, the S value for the client daypart is the weighted average of S values for the input dayparts in this set. The average is weighted by the portion of the client daypart contained in each input daypart used. After S values are translated, they can be extended (rolled up) across multiple (client) dayparts and multiple stations (54). To roll up the S values, gross impressions (GI) are added up across order lines which can span multiple dayparts and multiple stations. Then, the rolled up S for the schedule (which can span multiple dayparts and stations) is the total GI divided by the number of spots.

Once these intermediate AQH computations are complete, similar computations are performed for Cume. The input C values for each client station are translated from the limited dayparts to the client dayparts (56). A preferred computation for this C translation is discussed further below in conjunction withFIG. 4. After C values are translated, they can be rolled up across multiple (client) dayparts and multiple stations (58). Rolling up C values across dayparts is accomplished used the same algorithm for C value translation; the input daypart for the algorithm is given as the sum of dayparts in the proposed schedule. A preferred computation for rolling up C values across stations is discussed further below in conjunction withFIG. 5.Delivery information50 is then computed according the formulas given above (60), and the process is complete. The translation and rolling up of S and C values can introduce inaccuracies to varying degrees but the resulting outputs are still considered excellent approximations for marketing and research purposes.

In some embodiments the translation of C values from a limited set of dayparts to an arbitrary daypart (56) is accomplished by the process illustrated in the flow chart ofFIG. 4. This process begins whencomputer system10 identifies the smallest input daypart which still covers an entire (user-specified) daypart, referred to herein as the minimal parent. A list of C values is created and opened at first without any entries (64). The process loops through all of the input dayparts, identifying in each iteration the maximal (longest) input daypart that is still within the arbitrary daypart. The C value of the maximal input daypart is added to the list, and the maximal input daypart for this iteration is subtracted from the arbitrary daypart (66). The process then loops through selected input dayparts to find any intersections (overlaps) with the remaining arbitrary daypart. In the implementation for Arbitron input data, the selected input dayparts are M-F AM, Sat AM, Sun AM, M-F MD, Sat MD, Sun MD, M-F PM, Sat PM, Sun PM, M-F EV, Sat EV, Sun EV, and M-Su ON. If an intersection is found a C value is computed for it and added to the list. The intersection is removed from the arbitrary daypart, so any intersections are mutually exclusive (68). The C values for the intersections may be computed in various manners, preferably using a hyperbolic fit. The fit uses a hyperbola of the form y=Ax/(Bx+C) that passes through the points (x₁,y₁) and (x₂,y₂), where x₁is 1, y₁is the AQH of the input daypart which is being subtracted from the arbitrary daypart, x₂is the number of hours in the input daypart which is being subtracted from the arbitrary daypart, y.sub.2 is the Cume of the input daypart which is being subtracted from the arbitrary daypart, A is y₁*y₂, B is (x₂*y₁−x₁*y₂)/(x₂−x₁), and C is x₂*y₂−Bx₁.

A binomial method can be used to calculate C for the arbitrary daypart based on the compiled C values in the list (70). This binomial method is the same as that described below in conjunction withFIG. 5, but the C value of the minimal parent is substituted for the population, i.e., the Cume value C_arbitraryfor the arbitrary daypart is

C_{arbitrary} = 1 - Π_{j - 1, m} (1 - \frac{C_{j}}{C_{small}})] * C_{small}

where m is the number of Cume values in the list, C_smallis the Cume value of the smallest input daypart, and C_jis each Cume value j in the list. The arbitrary daypart can represent a sum of component dayparts in a proposed advertising schedule.

In a simplified example, consider input dayparts which include a four-hour daypart of Monday-Sunday 6-10 a.m., and 1-hour dayparts of Monday-Sunday 6-7 a.m., Monday-Sunday 7-8 a.m., Monday-Sunday 8-9 a.m., and Monday-Sunday 9-10 a.m. The client is considering a schedule which includes the 3-hour daypart of Monday-Sunday 6-9 a.m. In this case, the minimal parent would be the four-hour daypart, and the accumulated input dayparts would be Monday-Sunday 6-7 a.m., Monday-Sunday 7-8 a.m., and Monday-Sunday 8-9 a.m. The binomial calculation would then operate on the three C values for these three input dayparts, using the C value of the four-hour daypart as the population.

This binomial calculation is shown in further detail in the flow chart ofFIG. 5, but for rolling up C values across multiple stations rather than multiple dayparts. The C values for individual stations are based on given ratings parameters, e.g., a particular geographic market, a particular demographic, and a particular daypart. The process begins whencomputer system10 receives those C values and associated populations (P) for a given market over a specific demographic (72). The process then identifies a set of the multiple media stations based on user input (74). These stations have C values denoted C_i. The probability that a given person in the demographic is listening to the given station at a given time can be expressed as C_i/P. Accordingly, the probability that the person is not listening to the station at the given time is 1-(C_i/P). It also follows that the probability that the person is not listening to any of the stations in the market at the given time is

Π_{i - 1, n} (1 - \frac{C_{i}}{P})

where n is the number of stations in the set, and that the probability that the person is listening to at least one of the identified stations in the market is

1 - Π_{i - 1, n} (1 - \frac{C_{i}}{P}) .

The final Cume for the multiple media stations as a group can thus be computed as

\begin{matrix} [1 - Π_{i - 1, n} (1 - \frac{C_{i}}{P})] * P . & (76) \end{matrix}

As noted above, the various translated and rolled up S and C values can be used to generate a variety of outputs. One audience statistic that is very important to marketers/advertisers is reach. In some embodiments the present invention uses an inverse exponential model to compute the reach of an arbitrary radio advertising schedule, i.e., the estimated number of different people actually hearing an ad.FIG. 6 illustrates this model which can be constructed according to the formula

Reach=n*S*C/[(n*S)+C−S],

where n is the number of spots in a given schedule, and S and C are the aggregate (translated and rolled up) values for the schedule. When using this formula, for just one radio advertising spot the reach is S, while for a very large number of spots the reach is C.

Although the invention has been described with reference to specific embodiments, this description is not meant to be construed in a limiting sense. Various modifications of the disclosed embodiments, as well as alternative embodiments of the invention, will become apparent to persons skilled in the art upon reference to the description of the invention. For example, the invention is applicable to other media stations besides terrestrial radio, such as internet radio, cable or broadcast television, or satellite. It is therefore contemplated that such modifications can be made without departing from the spirit or scope of the present invention as defined in the appended claims.