Optimal Decision Making in TV Program Selection - Bill Harvey - MediaBizBloggers

By In Terms of ROI Archives
Cover image for  article: Optimal Decision Making in TV Program Selection - Bill Harvey - MediaBizBloggers

The commoditization of TV GRPs, which arises from bulk buying of a few oversimplified sex/age targets, undervalues advertising, television, agencies, and brands, creating an ideal atmosphere for private labels to continue their onslaught. Fortunately today this realization is rippling out over the TV buying selling community and along with TRA singlesource, there is also a renewed utilization of fusion tools which at least take baby steps in the right direction. This blog posting takes a step back to consider the high level theoretical underpinnings of the decision model for buying the right TV media.

In any decision making there is the concept of optimization: the idea that there is one best set of decisions/actions that comes as close as possible to perfection. Perfection is defined as the meeting of certain quantitative criteria. Where criteria exist that are not quantitative, the user must quantify those criteria in order to perform optimization. Thomas Bayes, who lived in the early 1700s, argued convincingly that it is better to include all variables in a decision rather than just those for which one has the best quantitative information, pointing out that even squishy estimates of a variable serve the decision maker better than the exclusion of the variable.

Of course, he would always take harder measurements as greatly preferable over squishy SWAGs (scientific wild ass guesses) derived from ratio estimation, wherever harder measurements were available.

Today we have new names for old SWAG heuristics: ratio estimation and operations at the household level, which produce mathematically equivalent results to ratio estimation, have been granted the high-falutin' term "fusion," and multiple regression of the coincidence of sales and marketing mix trend lines by DMA is called "econometric modeling." These inferential techniques for producing ballpark estimates have proven to be highly useful in the absence of anything more bulletproof. Enter singlesource.

Singlesource, by longitudinally tying hard measures of ad exposure, media exposure, purchase, demographics/psychographics together at the household level, fill in some of the blanks with harder data than we have ever had before as an industry. Mars Inc. (makers of M&Ms and a lot more) have disclosed their TV ROI increases (70 cents to $2) using singlesource. We have entered a new era of marketing almost overnight. The world will never be the same again.

As we contemplate the optimization of TV vehicle selection, these are the variables that must be properly weighted together by the optimizer:

1. The overall audience size of each alternative vehicle. Today the U.S. currency for this variable is Nielsen Media Research. NMR has earned this role by covering all sets in all types of TV households over a long period of years.

2. The density within that audience of people likely to respond to the message with more purchasing of the advertised brand. These would be known product purchaser households, since the industry's purchase measures are all at the household level and always have been. TRA is aimed at being used for this variable, using hard singlesource on massive representative samples.

3. The negotiated cost of the spot in that vehicle (possibly including cost of brand integration). This variable comes directly from the agency after extended head-butting with the sellers.

4. The effect of that program's environment upon that commercial, either upward or downward. Here TRA, like Apollo, is aimed at being used for this variable using multiple regression at the household, not DMA level, to tickle out the environment ROI effect for specific creative and specific program types and possibly even for individual programs where sample sizes permit, e.g. the Olympics.

5. The effect of the brand integration, true sponsorship, or other non-traditional ad unit that is available for that vehicle. The same vehicle can be considered with and without this brand integration enhancement as if the optimizer is looking at two alternative vehicles. TRA is also aimed at being used for this variable through use of multiple regression analysis to measure the net ROI add of brand integration. The TRA system is already being used to reveal analyses of this kind, sometimes finding strong effects and sometimes finding weak effects, depending upon the psychology of the execution of brand integration in each case.

6. The presence in the room with the TV set of the type of person in the household that the brand considers it essential to reach, either because that person is the purchasing agent who goes to the store, and/or the primary consumer of the product, and/or a primary influencer of the brand choice decision. For example, adult females are considered the people target for most food brands, mostly due to dominant purchasing agent status as well as being most motivated to make the right decision as to what to serve the family. Nielsen Media Research is the currency for this variable through the use of the pushbutton peoplemeter to estimate presence. True, the method is beset by noise from nonresponse as well as response error. Arbitron's passive PPM was used by then BBM's Pat Pellegrini and Ken Purdye in landmark research early this century to show that active peoplemeters understate additional viewers beyond the first viewer. Thomas Bayes would probably use PPM to make adjustments to NMR if Thomas were running a media agency today. J

7. Recency effects pointing at heavyup on specific days of the week. This would become important if it were determined in a particular case that the homes most likely to sales-respond tend to shop more on a specific day of the week, in which case the ads ought to be scheduled to hit those homes a day or two before the shopping trip.

8. We invite readers to indicate other variables not already included in the ones we have identified. Note however that we have purposely excluded commercial recall engagement, eyes-on and other interim measures of effectiveness since these would net out in ROI (variables 4 and 5 above) and their application as additional variables would be a form of double-counting that could move the optimized solution away from the maximum ROI.

So the optimization possible today with the hardest data available for each variable would use NMR and TRA together with media cost data. It has struck some people as ironic that TRA, while introducing scalable singlesource, is actually recommending the use of multiple sources, i.e. NMR as well as TRA. This is simply a matter of using the best data one has for each variable. Sure, it would be nice if you could get all the data – even the actual cost which will always be impossible – from one source. It would be more convenient and it would be neat, appealing to the OCD in each of us. But it is not essential. What is essential is making the best decisions, because to do otherwise costs money that can affect companies and the lives of the people running those companies.

Bill Harvey has spent over 35 years leading the way in the area of media research with special emphasis on the New Media. Bill can be contacted at bill@traglobal.com.

Read all Bill’s MediaBizBlogger commentaries at Bill Harvey - MediaBizBlogger.

Copyright ©2019 MediaVillage, Inc. All rights reserved. By using this site you agree to the Terms of Service and Privacy Policy.