The data science behind multi-touch attribution with TV data

The data science behind multi-touch attribution with TV data

Tom Weiss, Thu 01 February 2018

Tune-in is the poster child for closed-loop attribution on TV because we can quickly identify TVs that have seen a promo for a show with those who have subsequently watched that show. With the massive data sets now available from set-top-box and smart TV data providers, tune-in is becoming an everyday part of campaign planning and measurement.

But as networks start to run cross-platform campaigns, buying on digital and addressable TV, as well as linear TV, it becomes harder attribute uplift to a specific market channel, let alone when we want to drill down to the separate lift in OTT and VOD viewing.

We've helped many clients set-up their multi-touch attribution process. In this article, we're going to outline the process that we used to do this for a client.

Starting with the TV data

For any closed-loop attribution activity for TV marketing, networks will have access both campaign exposure data, and conversion data, from multiple different sources. For tune-in, both the linear campaign data and the conversion data will typically come from the same source.

For this specific project, the client was running a campaign across linear TV, addressable TV, and digital. For the digital campaign, they received IP address logs from a tracking pixel. For addressable, they had reported impressions from the MPVD running the ads, and for linear, they were using Smart TV data.

Each of these datasets had only a partial view of the campaign and in some cases a partial view of the market. The smart TV data only covered a subset of the homes exposed to the commercials. The addressable data and digital data should both be complete datasets. Because the MVDP provided just aggregate statistics for the addressable ads, we could not directly match this data together, but we could match the Smart TV viewing along with the digital exposure using the IP address.

The raw data that we started with once the campaign had run was as follows:

Smart TVs that saw the promo 61,127
Smart TVs that tuned into the show 25,597
Digital uniques 3,619,017
Addressable uniques 505,673

Uplift and effectiveness?

Taking the 25,597 Smart TVs that had viewed the show, we ran an IP match against the digital data. Of these Smart TVs, we matched 58% to an online identifier that we could match against the digital campaign.

With this in hand, our data science consulting team broke down the campaign into four different segments. People exposed to linear TV only, just addressable TV, digital only, linear TV and addressable TV, and those who saw the ad on both linear TV and digital.

With these figures in hand we were able to model the conversion rate, uplift, and effectiveness of each of these channels:

Customer engagement

The tune-in rate is the overall percentage of national exposures that then went on to tune in to the show. The uplift is the number of additional households that tuned in they saw the commercial. The effectiveness index is a measure of probability. It describes how much more likely people who saw the commercial are to tune in compared with those who didn't see the commercial.

As we can see a significant uplift in viewing based on the campaign, it is simple to calculate the return-on-investment, but of course what the client now wants to know is how to make it more efficiently next time?

So what's the best approach?

When our data science consulting team plotted the reach vs. the effectiveness of the different media, they saw a distinct difference between the number of people reached with each media type against the effectiveness of the campaign.

Customer engagement

These results suggest the most effective approach is to target people on both linear and addressable, but because of the low availability of addressable inventory, this only gives us limited reach and is thus hard to use to scale a campaign. Both linear and digital on their own provide scale but at the cost of lower effectiveness. The sweet spot between reach and efficiency comes from a combination of both digital and linear.

These conclusions are of course only valid for this specific campaign, and as more addressable inventory becomes available, we would expect it to grow in importance. For this campaign at least, the combination of linear and digital seems to deliver the best results.

Need help? Get in touch...

Sign up below and one of our data consultants will get right back to you

Other articles about Advanced Advertising

Dativa is a global consulting firm providing data consulting and engineering services to companies that want to build and implement strategies to put data to work. We work with primary data generators, businesses harvesting their own internal data, data-centric service providers, data brokers, agencies, media buyers and media sellers.

145 Marina Boulevard
San Rafael

Registered in Delaware

Thames Tower
Station Road

Registered in England & Wales, number 10202531