Howard Shimmel: The Dollar Value of Stability

When Howard Shimmel was running research at Turner he and Dan Aversano created a very advanced optimizer which used forecasted audience data rather than historical audience data When Howard resigned from Turner it was not long before he used his knowhow to with Dan Aversano and others create datafuelX DFX a very advanced optimizer for television networks to use to create the best targeted outcomes for specific advertiser datadriven advanced audience linear campaigns Of course that optimizer forecasts future audiences Therefore stability is very important because instability obviously causes increases in the size of errors in forecasted dataHoward has done what I think is the smartest analysis I have ever seen of the effect of audience measurement stability on the economic risk networks take by guaranteeing audience sizes I will share with you the high points and the complete analysis is available hereThe analysis compares the Mean Absolute Percentage Error MAPE in forecasting audiences based on panel only versus panel plus big data At the time some months back when Howard was doing the analysis he was not able to get all the big data but only a subsample of it and he is now repeating the work using the full Nielsen big data sample plus panel Therefore the stability of panelBD BDbig data is even better than is shown by this analysisThe advanced audiences used were MRI fusion product users with a sexage audience demo target of adults 18In the following tables a minus sign is a good thing it means a reduction in instability The larger the size of the minus number the betterThis first table represents two network groups and the individual networks in each groupAs you can see the addition of even a part of the big data sample made doubledigit improvements in stability as compared to the panel aloneThe average improvement was 23 less instability for one network group and 72 less instability for the other network groupThe medical insurance target group was the smallest target and therefore it is not surprising that it showed the largest improvement for one of the network groupsHoward suspected that the lower the ratings for a given network the more improvement would be seen so he analyzed the networks after putting them into groups based on quintiles of average rating sizeSure enough the data support a fairly consistent pattern of increased improvement the lower the rating size of the network The lowestrated networks benefit the most from the addition of big data to the panelNow comes the really interesting part the financial implications For this I shall take a straight extract of Howardamp8217s own wordsFinancial Impact ModelingTo understand the financial implications of improved forecast accuracy we modeled a standard crossnetwork daypartrealistic audience plan using average panel audience budget and average panel target CPM benchmarksThe key metric we analyzed was Estimated Financial Exposure which represents the dollar value of average variability between forecasted and actual impressions In other words this shows how much publishers or advertisers might be working within amp8212 positively or negatively amp8212 due to forecast uncertainty on a given audience dealBudget 415868Target CPM 18037Average Absolute Difference from Target Delivery Panel 1938Estimated Financial Exposure Panel 80595amp160 amp160 amp160This means publishers can be operating within a amp17780K swing in forecastbased audience deliveryWith the improved accuracy seen using big data 1485 absolute difference the exposure is reducedEstimated Financial Exposure Big Data 61755Net Savings 18840 Savings 24These results reflect just one campaign amp8212 but the financial relevance scales significantly Across larger portfolio deals or quarterly planning cycles even modest improvements in forecast precision can yield substantial reductions in liability and exposure delivering real value both operationally and financiallyAnother way to think about the financial benefit is from the perspective of inventory utilization Letamp8217s assume that the publisher wanted to insure 100 delivery to target and scheduled ADUs to cover the forecasting error Using Nielsen Panel data that publisher would need to schedule 24 ADUs for every 100 paid units using Big Data that publisher would need to schedule 17 ADUs for every 100 paid units While this doesnamp8217t seem like a huge savings of inventory 7 increase in inventory yield across all DDL deals is materialamp160What Howard is showing is that the benefits to the sell side of increased stability even based on only a portion of the big data can in normative cases be in the ballpark of from amp87647 to amp876424 More inventory available to sell less risk of having to give free units For the big players whose annual ad revenues are in the region of 5B each these percentage improvements translate to hundreds of million of dollars per yearBigger Menus Make Viewers More FickleThere is evidence to suggest that there is more unpredictable disloyal viewing behavior now than there was in the 1960s when there were essentially three viewing choices The more choices we have the more fickle we become the more we are willing to go beyond the programs we know always entertain us This is most obvious when big data are compared to panel data in terms of amp8220do they spike up and down in the same placesamp8221 In this earlier article I demonstrate that there is instability in the viewing reality that cannot be removed by measurement improvements This fickleness factor makes it all the more important to include big data in the most scientific ways possible with the best available panel data since there is a degree of instability in the real world so it is best to minimize as much of the sampling error as we canCover Image source BeetTV Howard Shimmel video interviewPosted at MediaVillage through the Thought Leadership selfpublishing platformClick the social buttons to share this story with colleagues and friendsThe opinions expressed here are the author39s views and do not necessarily represent the views of MediaVillageorgMyersBizNet

Become a Thought Leader. Share your thought leadership article with industry professionals & have them archived in Thought Leadership

Howard Shimmel: The Dollar Value of Stability

Author

Publish date

Channel

Archived in

Bill Harvey

Hispanic Leadership Roundtable: Inclusion, Mentorship, and Building the Next Generation (Video) - By Juan Ayala

More Videos:

Become a Thought Leader. Share your thought leadership article with industry professionals & have them archived in Thought Leadership

Howard Shimmel: The Dollar Value of Stability

Enjoying This Commentary? There's More to Love

Bill Harvey

TAGS:

Hispanic Leadership Roundtable: Inclusion, Mentorship, and Building the Next Generation (Video) - By Juan Ayala

The payment was successful

Cookies