The quality of advertisement creative is often judged by one of two methods. One approach is the ability of ads to win awards at one of the many ad pageants held every year. The process involves a panel of eminent judges, who view all the creative work produced by ad agencies and pass judgment as to which are the winners. The Ad Awards approach, though highly coveted by the creative folks, is a process that is very subjective and can often be controversial. More importantly this process completely ignores the selling aspect of the ad and merely applauds the narrative.
Marketers, who actually pay for the creation of these ads, how ever use another method. They depend on market research techniques to judge ad copy. Marketers conduct surveys periodically on parameters such as awareness, comprehension, likeability and intention to purchase. Ad copies that score high on these parameters are termed as winning creative. The Market Research approach is certainly more objective but may be error prone if respondents differ between what they say and what they do. This method hinges on an assumption (often true but not always) that ads that are liked are indeed the ads that sell. While this assumption if quite often true, history is resplendent with examples of highly creative ads that did not work in the market place.
This paper proposes to use an analytical solution to the measurement of ad effectiveness, by isolating copy effects from those of media weights. This methodology does not use survey opinions of target audience or the beliefs of eminent judges at awards night. It is easy to implement and does not require too much additional costs. It uses just one parameter to make the call, and that parameter is the ability of the advertisement to sell the product. Many marketers will readily approve with this reasoning, and will even convey that the raison d entre of an advertisement is to sell. David Ogilvy the famous ad man once marked 'If it doesn't sell, it isn't creative".
For this paper we retain this fundamental idea that good copy should and must sell products. However the approach does not restrict the user to salesmanship alone and that the methodology is easily expanded to cover other brand track parameters of awareness, comprehension, like ability and intention to purchase.
Vector Auto Regression Model provides us the base model:
Sales = Base Sales + Distribution - Price + TV GRP + Print + Outdoors -Competition GRP + Season
The TV enters the model as ad-stock to include diminishing returns and carry over effects. The ad-stock was calculated using the sum of all GRPs across the 24 creative renditions. The Beta parameters obtained for this model was significant at 90% confidence and was also validated "out of sample".
Once the impact weights are provided, they are fit back into the original regression model to test if the weighted Ad-stock provides a better fit an the non weighted ad stock. Measures that suggest better fit are the Akaike Information Criteria and Mean Absolute Percentage Error.
The weighted Ad-stocks provided a better fit than the non weighted ad stocks thus confirming our hypothesis that the weights will improve the model. The contribution for each creative is then normalized for a 100 GRPs to isolate the effect of creative from that of media repetition. Finally, the 24 creative's are ranked in the order of their ability to sell. Our ranking provided a near perfect match with that of the copy scores provided by a leading research agency.
For the analysis we used the following data obtained from a leading FMCG company. The Vintage of the data was 3 years, at a monthly level.
We then mount the regression model into a non linear programming formulation (NLP). Here we ask the NLP to provide us impact weights for each creative. The weights must not be negative and must sum up to unity. The NLP algorithm chooses weights that maximizes the likelihood function and significant at 5% levels.