COMMERCIAL SALES IN CABLE TELEVISION - AN OVERVIEW AND A LINEAR MODEL

Transcription

1 COMMERCIAL SALES IN CABLE TELEVISION - AN OVERVIEW AND A LINEAR MODEL Jagu P. Aiyer 1 and Jayant Rajgopal 2 ABSTRACT This paper deals with the sales of commercial advertisements in the cable television industry. It describes in detail the dynamics of the entire process, and develops the notion of an audience loss function, the minimization of which is equally desirable to both networks as well as advertisers. A simple, yet robust linear model is described for optimizing this loss function, subject to the constraints that are typically inherent in the process. This is followed by a numerical illustration that uses real-world data, and a discussion on the use of the model. 1 Management Science Associates, Pittsburgh, PA Department of Industrial Engineering, University of Pittsburgh, Pittsburgh, PA 15261

2 INTRODUCTION In the United States, approximately 64 million households have access to cable television and there are approximately 80 channels which are carried by cable operators. New channels, both basic as well as premium, are sprouting by the day, and cable networks have shown a steady annual growth of about 3% in viewership (McConville, 1995). With cable providers having access to millions of households nationwide (and in some cases worldwide), they present a very effective medium for advertisers to reach their target audience. Cable advertising is a rapidly growing business; in 1994, advertising revenues totaled approximately $2.85 billion (Dempsey, 1995) and revenues have been growing at the rate of 14-17% per year (McConville, 1995). Perhaps the most significant feature of cable audiences is that they tend to be segmented very clearly with respect to demographics (Batra and Glazer, 1989). This makes them especially attractive to advertisers who prefer a pure, demographically segmented audience to which they can deliver specially tailored messages (Battino and DePalma, 1989). In fact, cable networks have had a significant influence on the ability of broadcast networks to retain viewers (Walburn and Yucelt, 1995). Given that advertisements constitute the prime source of revenue for most cable networks, the rapid growth in this area has made the selling of cable spots extremely competitive. In order to maximize revenues, it is imperative for cable networks to make optimal use of the time available for advertisements, while retaining the flexibility to dynamically meet its customer requirements. The objectives of this article are twofold. The first is to describe in detail the dynamics involved in the sales of commercial advertising spots, so that this area may be further pursued by other researchers. Given its importance, the degree of contact between the cable industry and academics as well as the level of academic research in this area have been disappointing. This point is noted by Glazer and Batra (1989) who clearly emphasize the need to stimulate formal interest by academics in the practical concerns of cable networks. The second objective of the

3 article is to present a simple, robust, mathematical model that represents the selling process and provides a means for cable companies to decide on a strategy to maximize advertising revenues. The first half of the paper is descriptive in nature and deals with the first objective. It provides a detailed overview of how commercial inventory is created, what its characteristics are, and how it is sold. This portion of the paper provides the background necessary for identifying research issues in this area, and provides the setting for the second half of the paper. The latter begins with a discussion of what makes a particular sale efficient from a network s perspective. This leads to the development of a loss function and a mathematical model, which are then illustrated with a real-world example. COMMERCIAL SALES - PROCESSES AND ISSUES The Nature of Commercial Inventory Commercial inventory is typically imbedded within programs aired by a network and the value of the inventory is a direct function of the program, its air-time and its audience. Inventory is created when programs are scheduled. The shortest programs usually last half an hour, while others are scheduled in increments of thirty minutes. Each program is made up of show segments and commercial breaks. Typically, the show component is approximately two thirds of the program length and the other one third is devoted to commercials; a viewer can thus expect about twenty minutes worth of commercials in a one hour program. Commercial inventory is usually expressed in terms of thirty second spots, this being the modal copy length for advertisements. Thus, the one hour program above may create an inventory of about 40 spots. The broadcast industry uses the term avail to denote this inventory. The total commercial time is further partitioned into breaks. Each program has a generic break structure. For instance, in the one hour show discussed above, the twenty minutes of commercial time may involve eight breaks of 150 seconds each. A break

4 contains contiguous lengths of time termed as chunks. Continuing with our example, each break may contain three chunks of 60, 60 and 30 seconds. Commercial inventory has specific characteristics associated with it. For example, there are different types of inventory: some common inventory types are national, local, and promotion. In addition to the above categorization, there are other ways of viewing the inventory. For instance, inventory is also viewed on the basis of criteria such as product category avails, copy length avails and A-spot avails. The former refers to spots that are devoted to products in some specific category, e.g., automobiles or pet-food. Copy length refers to the length of the commercial itself. A-spot avails refers to the first commercial that airs in a break; in our example we would have a total of eight units of A-spot inventory. These characteristics are relevant because sales contracts are often constrained by them. For example: (1) a commercial promoting a particular product category normally precludes any other commercial belonging to that category from being aired in the same break; (2) an advertiser may specify a ratio of different copy-lengths; (3) each chunk in some break may be devoted to a different type of inventory, etc. Drivers of Commercial Sales A cable network tends to assume a specific image - Comedy Central specializes in comedy shows, whereas CNN Headline News specializes in summarizing current events. Akin to their image, networks also assume a corporate demographic profile. For instance, the Lifetime Channel identifies most closely with adult women, while the Disney Channel identifies with children. Since a network acquires programs that are in accord with its corporate image for the entertainment of viewers belonging to its corporate demographic profile, sales of commercials tend to be strongly driven by these factors. Indeed this is what sets cable apart from broadcast networks; cable channels are becoming increasingly distinguished, and by the turn of the century, audience

5 fragmentation is likely to be the rule (Spain 1995, Wilke 1995). Appealing to specialized audiences is of great attraction to advertisers as they pursue target or niche markets (Fahey 1991). In addition to the network s characteristics, individual programs also influence commercial sales. First, programs too tend to have a dominant viewer demographic profile; e.g., the dominant viewer profile for the show Thirtysomething was working, single women. While with many programs this profile coincides with the corporate demographic of the network, it is not necessarily true in every instance. Occasionally, there may also be significant differences in viewer characteristics for the same program. These factors play a major role in the sale of commercial spots since advertisements are typically aimed at specific demographics. Secondly, a program by itself has various attributes, any of which could play a role in the sale of commercial time within it. Examples of these are break and inventory structures that are unique to the program, the time of day when it airs, whether it is a single program or has multiple episodes, whether it is live or taped, whether it is a first run program or a rerun, etc. Finally, we look at revenue related drivers of commercial sales. Cable advertisements are usually aimed at some specific demographic associated with a corporate buyer-cpm (cost per mille) which is the average price that an advertiser is willing to pay per thousand viewers of that demographic. This determines the advertising rates, and hence revenues for a program. A related notion is that of a network-cpm which is the latent cost per thousand impressions to the network for each unit of commercial inventory that is created within the program. To illustrate these concepts, consider a one hour program with an inventory of 40 spots for which the audience is estimated to be 50,000 (or 50 mille) households. If the program costs the network $4,000 then this implies that each spot on the program costs the network $100 to create. Since the estimated program audience is 50 mille, the network-cpm is therefore equal to $2. If a spot within this program is sold for $200 dollars then the buyer-cpm is $4. In general, there is a buyer-cpm

6 associated with each specific demographic. This is because while the rate for a particular spot within a program is fixed, the number of impressions may be different for different demographics. When networks acquire programs, their main goal is to pay an amount that will determine a network-cpm that is well below the buyer-cpm. The Sales Process The process of selling cable spots must deal with several different issues, all of which play a role in determining revenues. First, cable companies sell spots guaranteeing a certain exposure, or in industry parlance, a certain number of impressions. When the total delivery of impressions is insufficient, additional spots must be run free of cost to compensate for the deficiency. Second, programs exhibit considerable diversity in terms of the dominant demographic of their audience. Given that advertisers are interested in a principal demographic profile, they typically have preferences for specific programs and/or segments of the day in which to air their commercials. Third, the sale of spots is a dynamic process and therefore decisions made at some earlier time clearly affect the decisions that are made later. Fourth, decisions involving one customer directly influence the decisions that can be made for others. Finally, there are other collateral issues such as commission structures for sales force compensation, cost amortization and price discounting strategies; we will ignore these in this paper. The sales process starts with a management team establishing selling titles; these may be viewed as being akin to different product lines or models. As a simplified example, if the network is airing 24 one-hour programs during a broadcast day, it may sell spots embedded in these programs as 24 different selling entities. Alternatively, they could decide to sell spots for programs grouped together in some rational manner. For example, a network may market an entity called prime time that includes four one-hour programs airing between 6:00 and 10:00 P.M. For the purposes of this study we will assume that there is a one to one correspondence between the selling

7 entities and the program titles, although this is not a restrictive assumption. The sales management team also establishes a rate or a ratecard for every program in which spots are sold. Often, price discounting is used as a device to stimulate the sale of slow-moving inventory by permitting the sales force to discount spots within limits. In the extreme case, a totally stagnant program may even be replaced with a more marketable one. Companies also spend significant amounts on promoting programs to increase viewership (Levin 1995). Networks typically operate a broadcast calendar starting in October and ending in September of the following year. The upfront season is a period that lasts about a month during July, and the sales for the next calendar year usually begin during this time; the sale of children s programs starts as early as May. Much of the commercial inventory is sold during the upfront season, with the percentage varying between 60 and 100 for different networks. Revenues during the upfront season in 1995 topped out at $1.7 billion - approximately 55% ahead of the 1994 upfront season (Burgi 1995). The remaining inventory is sold during the broadcast year in the form of scatter sales. In general, the price of a spot will be slightly higher for a scatter sale when compared to the price for an upfront sale. Networks manage spot sales through their sales forces which include account executives (AE s), sales planners, and sales assistants. Buyers specify their requirements in terms of impressions required, a target demographic, flight (a specific time interval) and a budget. For example, a specification may be for 3000 mille impressions of men between 25 and 50 years of age at a budget of $18,000 to air in the last two weeks of December. The AE prepares a proposal which lists the programs during which the buyer s commercials will air, the number of spots in each program, the week(s) when the spots will run (the flight), the copylength, the rate per spot, the number of impressions estimated to be delivered by each spot, and the total dollars and total number of deliverable impressions guaranteed. The proposal is presented to the buyer (the industry

8 commonly uses the term deal to refer to the proposal) and is revised based on negotiations before it is written as a contract. Often, several deals are written for the same request so that the customer is given more than one option. The situation with multiple scenarios is examined elsewhere by the authors (Aiyer and Rajgopal 1996); in this paper we are concerned with drawing up a single deal. The delivery of impressions guaranteed in a deal are written on the basis of expectations. When the program airs, the actual impressions (as reported by Nielsen for example) may deviate from this prior estimate. If the actual number of impressions delivered is lower than the estimated delivery, the guarantee calls for the network to run additional spots at no cost in order to compensate for the deficit. These additional units are termed as Audience Deficiency Units (ADU). Deals are reviewed periodically for their delivery in order to assess ADU requirements that may arise. If a deal delivers impressions in excess of the contracted amount, the network does not get additionally compensated and the additional delivery is essentially a free gift to the buyer. Finally, a percentage of spots are canceled by advertisers. In such cases, the contract is reviewed at the cancellation point in order to discern how the guarantee and CPM will apply to the partial contract. PLANNING SALES CONTRACTS Determinants of an Efficient Contract Based on the groundwork laid in the previous section, we now examine what constitutes an efficient sales contract and how sales may be planned in an optimal fashion. Summarizing the contents of the previous section, a deal should deliver the required gross impressions of a specified demographic at the average buyer-cpm specified. The deal comprises a set of scheduled programs and one or more spots within them in which the customer s advertisements will be aired. The complexity of the planning problem arises from the fact that there are a finite number of spots with each having its own potential audience, but there are numerous customers who

9 compete for these same spots. Thus a contract written for one customer has a direct impact on what can be contracted to another customer. If future customer requirements are all known in advance (e.g., all sales are made simultaneously) the optimal strategy is to simultaneously arrive at deals for all customers by optimizing the entire system, i.e., by taking into account interrelationships between various customer requirements. In practice though, sales transpire sequentially in a dynamic fashion and deals need to be proposed as these requirements from buyers come in at different points in time. During the early stages, there may be several alternative ways to formulate a deal since most of the inventory is available and there is considerable flexibility in how a deal is proposed. As the inventory is depleted over time, the set of feasible options shrinks and there is a concurrent reduction in flexibility as well. At later stages, when much of the inventory may have been depleted it may not even be feasible to formulate deals satisfying the specifications of a buyer, since the remaining inventory of programs may not be able to deliver impressions of demographics demanded by the buyer. Therefore it is critical that an early deal that enjoys the availability of most spots, should consist of programs that are exactly appropriate for meeting the requirements of that buyer. Mistakes made in selecting programs during early sales activity may leave the remaining inventory difficult, or even impossible to sell. To illustrate this point, suppose that there are two open spots, the first in Program A which has estimated audiences of 70 milles and 20 milles for demographics d 1 and d 2 respectively (where d 1 represents men and d 2 women, say), and the second in Program B which has estimated audiences of 50 milles each for the same two demographics. Now, suppose a customer specifies a requirement of 50 mil impressions of demographic d 1 at a CPM of 3. This may be done by choosing one spot from either program in which to run the customer s commercial (at a cost of $150 for the spot), since the customer s requirements are met either way. However, from the

10 perspective of the network, despite the additional (free) impressions delivered the first option is probably superior since Program A has a much smaller impact on the secondary demographic d 2 than does Program B. To see this, suppose there is a second customer who requires 50 mille impressions of d 2. If the spot in Program B had been used up to satisfy the first customer, the network would be unable to satisfy the second one using the spot in Program A. Since demands come in sequentially over time, sales planning is thus a dynamic problem. One must ensure that programs chosen for a customer specified demographic have a minimal impact on other demographics. Since a network maximizes its revenue by minimizing the loss of valuable impressions, in a perfect selling environment the network does not lose any impressions from other demographics, and one would choose a program for which the audience comprises 100% of the demographic of interest and 0% of all others! Since this is infeasible, one must measure the impact of the sale on these secondary demographics. This provides the basis for an objective function where the network attempts to minimize this impact. Details of this objective are provided in the next subsection. A Mathematical Model for Planning Sales We first introduce some notation as follows: p m w A pm Q pw β m index for programs, p=1,2,...,p index for demographic profiles, m=1,2,...,m index for weeks (units of time) in the flight, w=1,2,...,w estimated audience (in milles) for program p and demographic m available inventory of spots in program p during week w a relative weight associated with demography m C p rate paid by the buyer for a spot in program p (usually the same for all spots in p) m r index of the reference demographic profile specified by the customer

11 D B demand for impressions (in milles) of demographic profile m r budget specified by the customer Let us define the variable X pw to be an integer representing the number of spots to be assigned to program p during week w. It will be non-zero if the program is chosen for that week and 0 otherwise. Then the problem may be formulated as an integer linear program in the X pw. The constraints of this program are as follows: The specified minimum number of impressions should be delivered: P W Apm X pw D r p= 1 w= 1 (1) The total value of the deal must be no greater than the budget specified by the customer: P W C p X pw B p= 1 w= 1 (2) The number of spots assigned in a program for any week cannot exceed availability: 0 X Q pw pw for all p,w (3) The objective is to minimize the negative impact of the assignment on other demographics. Specifically, revenue is maximized when the loss of potential commercial impressions from secondary demographics is minimized. However, not all demographics are equally valuable. In order to specify the objective one needs to assign weights that measure the relative importance of all the demographics. Several approaches are possible to estimate these weights, three of which are discussed below. The first approach is based on subjective assessment. The network simply assigns weights between 0 and 1 for each demographic to denote its relative importance. It is reasonable to expect that the corporate demographic will be assigned the highest weight assuming that the modal

12 cumulative demand is for that demographic. The only further constraint the network needs to observe is that the weights sum up to unity, i.e., M β m m= 1 = 1 (4) The second approach involves estimating the weights based on the cumulative demand for impressions from all customers during the past. Assuming that the pattern of demand among the different demographics for the current year will be similar to that in previous years, one can easily develop a sophisticated demand forecasting model; the details are not explored here. In general, data from recent years are used to forecast demand of impressions among different demographics for the current year. If F m represents the cumulative forecast of impressions for demographic m, then the weight β m corresponding to that demographic is: β m = F m M F j= 1 j (5) The third method is based upon the schedule of programs for the current year. Considerable research efforts go into acquiring programs and scheduling them, and the estimated delivery of impressions based on the schedule is used in preparing proposals. It is therefore reasonable to assume that the relative importance of a demographic is proportional to the total audience for the same. If E m represents the cumulative estimate of impressions in demographic m deliverable by the current program schedule, then the weight β m for that demographic is: β m = E m M E j= 1 j (6) Now suppose X pw > 0 for some p, i.e., we choose to run some spots in program p during week w. The reference demographic m r is specified by the customer and thus all other

13 demographics are secondary. For some demographic m other than m r, the number of impressions lost due to this assignment is equal to A pm. We may now define a weighted loss function L for the assignment X pw as M P L = β [ A ( X )] m= 1 m mr W m pm pw p= 1 w= 1 (7) where β m is chosen via (4), (5) or (6). Thus the objective of our program is to minimize the function given by (7) subject to constraints (1), (2) and (3). This is a straightforward integer linear program, a mathematical programming problem that has been addressed extensively. Also note that the above problem is easily extended to account for additional constraints such as specific types of spots, maximum and/or minimum numbers within a program, non-standard copy length, etc. Before ending this section, a few words about the audience estimates A pm are in order. Nielsen provides actual viewership data based upon a projection of samples. The data are provided on 42 atomic demographics. However, the demographics specified by typical cable customers tend to be much broader. For example, a contract might specify a demographic bracket of women ages 21-35, which actually includes three atomic demographics: women ages 21-25, and Merging atomic demographics into a wider range is common practice and our model readily supports this because we need only index those M demographics that are of business significance to the cable company. Some of these may overlap but each one is distinct and has its own weight computed via (4), (5) or (6). An Illustration To illustrate how the model would work, consider the following data which is a scaled down version of real-world data used by cable networks. Table 1 presents audience estimates A pm

14 in milles for P=8 different selling titles and M=9 different demographics (HH refers to households, P to people, M to men and W to women, while the numerical values refer to ages). The last row of the table presents values for the weights β m (which were obtained by a combination of the first and the third approach detailed in the previous section), and the last column presents the rates (in $) charged for a spot (C p ) in each of the titles. We consider a flight of W=4 weeks and Table 2 provides the inventories currently available in each of the four weeks for each of the eight titles; note that the earlier weeks have more of their inventory depleted than the later ones. Suppose the customer has specified an impressions requirement of D=30,000 milles for demographic m r =1 (households) and a maximum budget of $275,000. The resulting mathematical program has 32 variables, two structural constraints and the upper bounding constraints on all variables. The optimum value for the loss function was and the corresponding solution has [X 11 =4, X 41 =2, X 61 =1, X 12 =6, X 62 =1, X 13 =10, X 63 =9 and X 14 =2], i.e., schedule 4 spots in the early morning, 2 in the late afternoon, and 1 in prime time during week 1, 6 spots in the early morning and 1 in prime time in week 2, etc. This solution delivers exactly 30,000 milles at a budget of $273,400. DISCUSSION AND CONCLUSIONS The industry s major objection to the application of quantitative techniques is that the models abstracted for solution are not easy to implement and data are not readily available in the form required by the model. A key feature of the proposed model is that its information requirements, while comprehensive, are readily met. At present, rapid expansion in the cable industry is leading it to acquire state of the art systems that support its managers in all functional areas. For cable companies to be successful they need to use accurate ratings and delivery data and have tight control over commercial inventory (Fahey, 1991). The industry is moving from a set of isolated and fragmented information systems to an integrated systems environment. A few vendors specifically support the information needs of the cable television industry; Management

15 Science Associates (MSA) located in Pittsburgh is the leader among them. MSA s product Gabriel is a an integrated programming, sales proposal, and inventory accounting system. Such an integrated system allows for the estimation of deliverable impressions, maintenance of rate cards, and proposal preparation. The system also provides current information on the available inventory on a real-time basis. Thus, audience estimates, inventory data and information required for estimating demographic weights are all readily available when an integrated system is in use. A number of real-world factors are also easily handled. These include program cancellations or changes, revisions in the pricing scheme (for example, discounting to sell slowmoving inventory), changes in demographics and in estimates on their relative importance, and variations in inventory availability. Since the model that is constructed to prepare a deal for a customer uses the most current information available, these dynamic factors present no problem as long as the data base used to construct the model is up-to-date. The optimal assignments to two different customers with identical requirements in terms of demographics, flight and budget could be quite different because the two came in at different times with different input data to the model. We should also state that there is an abundance of literature that addresses issues of media mix, campaign planning, quantifying effective exposure, etc. However, these issues are addressed from the point of view of a media buyer. The loss function defined here is equally desirable to the seller as well as the buyer, so that this paper expands the perspective to include the interests of the seller of a specialized medium without sacrificing the requirements of the buyer. Finally, there are excellent and reasonably priced software systems for solving integer linear programs, and the form of the model yields useful information as well. The linear relaxation for our model (i.e., where the integer restrictions are ignored) is a linear program (LP) that is very easy to solve. It yields a quick lower bound on the true minimum value as well as other potentially Gabriel is a trade mark of Management Science Associates

16 useful information which can be extended to the harder integer problem. For instance in our example, the relaxation becomes infeasible if D increases to 31,360 milles or higher, given a budget of $275,000. Similarly, it is also infeasible if the budget is $263,000 or lower given a requirement of 30,000 milles. Although the limits when the actual (integer) problem become infeasible will of course be tighter, these limits give us a rough idea of what minimum budget is feasible for a given value of D, or what is the maximum number of impressions deliverable for a given budget B. The bound provided by the LP also provides information which can be extended to the harder integer problem. To illustrate this, the LP relaxation yields a value of for our illustration. The first feasible (integer) solution found by the branch and bound solver yielded a value of for the objective. This value is within 0.34 % of the lower bound and could easily be adopted. The solver also yields other intermediate solutions; we obtained solutions with values of , and before finding the optimum solution with a value of In practical terms, this presents us with alternatives that are very similar to each other in quality, so that one may also use extrinsic, qualitative factors to choose between one or the other. Acknowledgements The authors wish to acknowledge two anonymous referees for their constructive suggestions.

17 REFERENCES 1. Aiyer, J. and J. Rajgopal Sales Directed Commercial Inventory Management in Cable Television. Working Paper, Department of Industrial Engineering, University of Pittsburgh, Pittsburgh, PA. 2. Batra, R. and R. Glazer Cable TV Advertising: A Strategic Overview. In Cable TV Advertising, (R. Batra and R. Glazer eds.). New York, Quorum Books. 3. Battino and DePalma Batra, R. and R. Glazer Cable TV Advertising: Is the Promise Being Fulfilled? In Cable TV Advertising, (R. Batra and R. Glazer eds.). New York, Quorum Books. 4. Burgi, M Cable TV: The Industry is Wired for Growth. Mediaweek. 5(35): S24-S Dempsey, J Cable TV Ad Revenue Rises 16.9%. Variety. 358(7): Fahey, A Cable TV Critiqued. Advertising Age. 62(17): Levin, G Basic Cable Nets Give Themselves Promotion. Variety. 360(7): McConville, J. Cable Numbers Are Up. Broadcasting and Cable. 125(38): Spain, W Fishing for Sales Through Cable. Advertising Age. 66(13),: S Walburn, W.B., and U. Yucelt The Effect of Cable Penetration upon Network Audience Size in the United States. International Journal of Advertising. 14(1): Wilke, M Cabooses, Two or Three Engines. Advertising Age. 66(13): S4.

18 Selling Title Demography (m) (p) HH P18-34 P25-34 P35+ F18-49 F25-54 M18-34 M25-54 M35-64 Rate(C p ) Early Morning ,300 Late Morning ,600 Early Afternoon ,300 Late Afternoon ,700 Early Evening ,700 Prime Time ,400 News ,800 Late Night ,900 Weights (β m ) Table 1: Estimated audience A pm (milles)

19 Week (w) Selling Title (p) Early Morning Late Morning Early Afternoon Late Afternoon Early Evening Prime Time News Late Night Table 2: Current Commercial Inventories Q pw