Modern TV attribution methodologies bring three worlds together. The first one, called top down, specializes in strategic measurement across all marketing channels. The second one specializes in measuring tactical allocations in addressable channels, such as digital. And the third one specializes in measuring tactical allocations in offline TV advertising. Getting access to the right type of data is key, but applying the right methodology to leverage its power is what makes the information actionable.
Impressions and rating points are good indicators of how many people saw your TV ad, but how does it influence individual consumer behavior? Consumers take action online in response to seeing your ad offline. But which types of TV ads trigger the greatest response?
To ensure you’re getting the most accurate answers to your questions, make sure your TV attribution methodology does the following:
Applies an algorithmic approach: The methodology should be able to ingest very granular TV and digital response data to build a highly accurate model that predicts the impact of TV on digital responses. Understanding TV’s impact by network, daypart, pod position, designated market area (DMA), timestamp, among others is critical to the TV buyer. Similarly, the same methodology should be able to ingest granular digital stimulation data. This will allow marketers not only to see the impact of TV on digital responses, but also see TV’s impact side-by-side with individual digital channels to gain a holistic view of performance.
Doesn’t make assumptions: Even though traditional TV is not addressable, some methodologies attempt to guess which of your digital users have also seen your TV ads. By artificially adding TV touchpoints to the sequence of digital touchpoints for some of your users, it negatively impacts the accuracy of your attributed metrics across all users. As a result, not only will the credit attributed to your TV advertising be inaccurate, but it will also skew the credit attributed to your digital channels.
Predicts the tail effect: The “tail effect” is the window of time between when a TV ad is viewed and when the viewer takes an action online in response to seeing that ad. It would be inaccurate to apply the same tail effect to all of your TV ads in order to make the assumption that all digital responses within the tail effect window should be attributed to the corresponding TV ad. There are too many variables that have an impact on tail effect. For example, daypart can play a big role. Early-morning TV ads may have a longer tail effect than prime time because viewers have more time to take action online in the evening. The algorithm should be able to figure out this changing tail effect at the spot level.
Accurate TV attribution takes the guesswork out of determining which combination of tactics drives the greatest digital response in order to more effectively allocate TV budget moving forward.
Some interesting points, Anto. I would note that GRPs, at best, are crude measures of how many times people might have seen your ad message. As a guess, these overstate the case by about 45% for most campaigns in terms of frequency but a lot less if total cummulative reach is your yardstick.
As for evaluating sales results after ad exposure, direct causal calculations are always a problem for they do not necessarily reflect the cummulative effects of a particular brand's positionning strategy as the campaign begins, matures and wears out over time. So, one has to factor in the weight of impressions prior to the most frequent one as well as where in the campaign you happen to be: beginning, mature phase or wearout phase. Obviously response rates, as indicated by sales, will vary considerably. At the outset, a new positionning strategy may woo a fair amount of brand switchers but later in the campaign, this may diminish while most of the sales results reflect buys made by dedicated users of the brand.
Great points, Anto. IMHO I would differentiate between direct and indirect impact of TV advertising. The direct impact - people immediately engaging with the brand after seeing the TV ad - can be measured quite well today, be it looking at call logs or at website uplift. This direct impact also has quite an effect on your digital paid search & affiliate campaigns as many viewers turn to Google to search for the product advertised instead of going to the URL shown in the ad directly. You could actually filter out these positive TV-driven effects on your digital advertising by taking the exact ad occurence into account when doing the analysis.
The indirect impact is harder to "capture" as it is long-term and - as Ed pointed out - quite hard to determince the causal relationship between running TV ads and response rates. A particular sale in a store as one response metric can be measured (although it is far from easy) by matching viewing data from set-top boxes with credit card purchase data in an anoymized way.
I want to mention two other examples that indicate the diffculty: A better brand image might lead to people more likely to click on a display banner six months later because they know and trust the brand. In an extreme case, think of Porsche - building their brand image starts at a child's age so that everyone wants one when they are older but only a few will be able to afford it 30,40,50 years later.
That said, as TV advertising often is the biggest part of the marketing budget, I absolutely agree with you that the right (i.e. proper) TV attribution and how we can measure it today should be part of any marketing attribution analysis.
Great article and comments. In my view there is significant value in further dissecting the direct impact of a TV commercial on digital. Besides knowing what happened after your spot aired on TV: who visited your website, who searched for your products, etc., it is almost as important to know what didn't happen: who saw the commercial but didn't interact on digital. This valuable insight enables marketers to further optimize their cross screen marketing strategy
Over the last couple of years we have seen that the combination of TV and digital is a potent strategy to increase brand recall and overall engagement. With TV moving towards IP based systems marketers will get access to more granular (and real time) TV data which will enable them to further remove the guesswork from their media tactics. Exciting times!
@Ed. You are right that GRPs are not always accurate. There are other ways to get the TV impression data too. Vendors like Rentrak, iSpot.tv, and certain SmartTV manufacturers provide TV impression data. A good TV attribution model should be able to understand the effects of TV ads using such impression data, granular dimensions and spot level media cost information.
To your second point, a good TV Attribution model will be able to understand the baseline effects and tease out the casual impact of each TV spot regardless of the constant TV advertising from the advertiser. It also need to consider the varying delays and maturity levels. Granular level data, multi-dimensional model, and establishment of a baseline are quite important to accomplish this.
@Andreas. You are right that the short-term effect(direct) and long-term effect(indirect) are different. TV Attribution models alone can only measure short-term effect. But using TV Attribution models along with Top Down Attribution models solves that problem. Top Down Attribution provides the brand equity, long term effect or indirect effect of TV and TV Attribution provides the incremental, short-term or direct effect of TV Advertising.
@stefan. Great points. Thank you. With the level of granularity of the TV data available today makes it much easier to differentiate the effect of each spot from the baseline.
Anto - thanks for the article. In your response to Andreas, you said the following: "using TV Attribution models along with Top Down Attribution models solves that problem. Top Down Attribution provides the brand equity, long term effect or indirect effect of TV"
Could you point me to more information for such Top Down Attribution models?
Thanks
Jim, Thanks for the comment. The following link can get you the answers to your question.
http://www.visualiq.com/resources/marketing-attribution-newsletter-articles/top-down-bottom-attribution-better-together