It wasn’t so long ago that advertisers were still reluctant to invest in online TV (OTV). But today OTV is undoubtedly the hottest channel within digital, sometimes taking as much as 80 percent of total digital budget.
This tremendous growth can be attributed to the emergence of Internet Gross Rating Point (iGRP), a methodology which has helped traditional marketers better understand OTV buying, with budgets naturally following. But now with programmatic OTV coming onto the scene, armed with better targeting options and behavioral datasets, it’s supposed to be even easier to reach a brand’s audiences. But is this really the case? Not while the industry insists on using traditional panel-based metrics to measure newer digital tactics.
iGRP is a Double Edged Sword
On the one hand, while the iGRP integrates traditional TV and online TV channels, on the other, it sets digital, especially programmatic, back by using a panel-based measurement.
Traditional TV has to rely on panels because not every home can be tracked, hence the best we could get was inferred demographics based on limited sample sizes.
Digital and programmatic is different in that we can essentially track every impression, so if we have all the data reverting back to limited panels simply does not make sense.
Take a programmatic OTV campaign as an example: We have a plethora of different third-party browsing behavior data that we piece together to construct the consumer portrait. We use all that data to select which impressions to buy on the exchanges. In the end, we only take a portion of those unique visitors (UV) and try to match them to limited panels. In essence, we’re buying impressions based on a behavioral dataset and measuring those impressions based on a panel dataset. This mismatch of a measurement methodology causes severe data discrepancies as well as supporting a “luck-based” buying approach in programmatic OTV.
Panels and Behavior Cookies Data Discrepancies
On the topic of data discrepancies, one can argue that either panels or cookies are more/less accurate. But I see it merely as comparing apples and oranges – it’s not that one or the other is superior, but rather we should standardize the buying mechanism and the measurement methodology. Panel data is based on sampling, but for real demographics data, we cannot discredit its authenticity, and when applied to scale, will inherently lead to some errors. Cookies data is collected at scale, but as demographics are largely inferred based on browsing behavior, it also will have certain percentage of error.
The issue with using both datasets is that we have no way of knowing how a certain UV will be classified under different methodologies. So we have circumstances where in a demand side platform (DSP) the UV is classified as female, aged 20-30, but the panel’s data is saying it is a male, aged 40+. Who’s right? And how do we control the result of such large discrepancies except by leaving it to pure luck?
Moving to First-Party DMP as the New Measurement Paradigm
Instead of mucking around with data discrepancies, advertisers should take measurement and verification into their own hands through the use of a first-party data management platform (DMP). The DMP collects cookie data from the brand’s digital assets (website, email programs, apps, etc.).
Third-party data such as search and e-commerce behavior should also be mapped with a brand’s first-party data to enrich the dataset. With this data, we can verify the UV impressions based on the following dimensions:
- visited the brand’s website in the last week
- searched for category keywords in the past two days
- bought category products within the last month
This would consolidate both the buying mechanism as well as measurement into a behavior-based framework. It would also allow us to use all the data that’s collected on digital, rather than panel-based samples. A pure behavior-based measurement methodology is definitely a drastic paradigm shift. But I believe that ultimately advertisers themselves should be the judges of whether their ads are considered on target, and this is done through an accumulation of their own data in a brand-owned DMP.
This article was written for ClickZ