Transforming Panel Measurements into Impactful Universal Insights
The U.S. is comprised of over 116 million households, the vast majority of which have access to some kind of TV content. These households receive this content from a variety of different sources (satellite, cable and streaming) through any number of providers (DISH, Xfinity, Verizon Fios, etc.) typically concentrated in specific geographic areas. Further complications include many households having several TVs with access to multiple types of content on the same TV device.
Considering all these scenarios, it should come as little surprise the TV industry has resorted to a variety of panels for measuring responses to TV advertisements. Panels can be powerful, but they need to be made representative and scaled to the U.S. population before they can provide insights on truly impactful metrics, including a company’s return on ad spend or cost per incremental conversion.
Using Data Science to Balance Households
Scaling panels is challenging, but far from insurmountable for the scientists at iSpot. With more than 80 years of combined experience in data and analytics, the iSpot data scientists transformed the panel of 12+ million TV households into one that represents the entire U.S. population. Though there are many areas to tackle, the initial place to start was accurately accounting for household viewership. This meant properly weighting each household based on their demographics and the amount of TV they consume relative to the national average.
To do this accurately, iSpot designates TVs with heavy viewership as “primary” TVs with all other TVs being considered “secondary”. In general, if a panel is demographically skewed compared to the U.S. population and has nonrepresentative viewership patterns (majority secondary TVs), these biases are pulled into the extrapolation numbers with potentially disastrous results. It may lead to vast over reporting (or under-reporting) of impressions or any other metric generated through the panel. Because of this, it is vital to rely on the expertise of skilled scientists, as iSpot did, to account for and correct these skews.
Representative Panel for Media Measurement
In order to solve for the demographic biases inherent to the iSpot TV panel, we first partnered with another data provider, Epsilon, to obtain the demographic traits for each American household. Next, we combined that data with the TV households in our panel and kept only the highest-quality matches.
Upon completion, iSpot used a well-known survey technique called “Iterative Proportional Fitting” or “Raking” to debias the panel. Raking effectively lowers the weight of overrepresented households and increases the weight of underrepresented households. In this fashion, each household receives the appropriate weight, and the panel is balanced relative to the U.S. census.
Impact of a Representative Panel
iSpot built this initial panel using age & gender (Figure 1), education (Figure 2) and household income (Figure 3) demographics, though iSpot intends to add many more in the future. It is important to note that across all these figures, the Census-Balanced Panel produced by the raking algorithm is significantly more representative of the Census (ACS) values for U.S. demographics than the Raw Device Panel. The result is the updated panel unlocking accurate and reliable national ad measurement tracking for all iSpot clients.
Figure 1: Age & Gender Shares by Source
ACS 2017 5-year Estimates v. Census-Balanced Panel v. Raw Device Panel
Figure 2: Education Shares by Source
ACS 2017 5-year Estimates v. Census-Balanced Panel v. Raw Device Panel
Figure 3: Household Income Shares by Source
ACS 2017 5-year Estimates v. Census-Balanced Panel v. Raw Device Panel
Viewership Weights
iSpot’s mission to have the most accurate extrapolated metrics means going beyond just demographic weights. The second innovation that our scientists developed was to correct for viewership biases (aka time spent consuming TV) in our panel.
Understanding how the viewership times on these TVs compare to national averages allows us to place weights on primary, secondary and tertiary TVs within a household. The knowledge of whether the 12+ million TV’s in our panel bias toward “primary” TVs or lesser used TVs lets iSpot to continue to adjust its panel to ensure maximum accuracy. These viewership weights in combination with our demographic weights result in a significant increase in extrapolation accuracy.
National Level Conversions
The large size of iSpot’s panel makes us capable of reporting much more than simply national TV impressions. It also contains the scale necessary to unlock extrapolated conversions on a national level.
Currently, to adjust for the demographic skews coming from both the raw panel and client responders, we take the percent of U.S. households comprised by our panel and compare to the percent of a client’s conversions we can match into our panel. This generates an overall adjustment factor we can apply during the extrapolation process. If the match between the raw panel and client responders is higher than the percent of the U.S. that the raw panel represents, the factor will adjust the values downward. A downward adjustment indicates our panel and the client’s responders demographically skew in similar ways. Similarly, an upward adjustment would indicate the iSpot panel and a client’s responders skew in contrasting directions.
Answering Key Business Questions
The result of it all means iSpot is able to report on a client’s extrapolated conversions. With balanced and adjusted conversions, our clients can view the full effects of television buys at specific occurrence or campaign levels by network or by daypart. Clients can view their historical performance and compare multiple conversion types against one another. In short, it’s an unprecedented view of activity, letting you answer the question “How many actions were driven by my ad?”
TV Ad Measurement for Disruptive Brands
As iSpot moves forward, we will continue to update these products and methodologies to reflect our knowledge of data and the evolution of the TV industry. We are proud to be the leader in TV measurement and attribution, and are grateful to the brands who trust us to provide them with industry-leading analytics.