We’ve worked enough with Nielsen's panel data to know that’s a fair-weather indicator of TV viewing. We’ve analyzed set-top-box data to see that it fills in a lot of the gaps that aren’t covered by panel data. Now we’ve got smart TV data from Inscape, and we've been comparing their data with set-top-box and panel data to see how they stack up together.
How did we compare it?
Nielsen publishes regular ratings, as do the set-top-box vendors, providing estimated national percentages of homes tuning in to shows. We calculated the same metrics using Inscape's data on a DMA by DMA basis and then normalizing it up to be nationally representative.
The chart below shows 15-minute ratings for two local affiliate networks in a small DMA. We chose these because they are areas where there are a minimal number of panelists available and few set-top-boxes from MVPDs who share their data. It's a classic case where the smart TV data should be able to show divergence from the norm.
What's interesting here is just how well the smart TV data tracks against both the panel data and the set-top-box data. There is more variability on both of the other datasets, which we'd expect based on the sample size, but overall we were impressed at consistent the datasets were.
We also found it interesting that the smart TV data didn't need anywhere near the level of modeling that we'd need to perform if we were dealing with raw set-top-box or panel data. By just applying simple arithmetic we can generate numbers that not only correlated well with the other data sets but provided very similar overall results.
So what gaps does smart TV data fill?
One of the most significant benefits of Inscape's data is that it's independent of the TV ecosystem. Nielsen is paid for by the networks. MVDPs pay the channels for their programming. Except for the influence of Netflix - whose app is more important than any other feature on the smart TV - the TV manufacturers are entirely independent of the networks and the MVPDs. MVPD’s won’t release granular information on viewing that will affect their carriage deal negotiations, and Nielsen is unlikely to adopt a methodology which would lead to a decline in the big four networks.
Only the Smart TV manufacturers are independent arbiters of what people are watching on their screens, and agencies, in particular, can trust that.
Smart TVs know what is on display.
The big dirty secret about set-top-box data is that nobody knows whether there’s a TV connected to the end of the set-top-box and whether it’s switched on. If a family sits down and watches CBS all evening, or switches the TV off after the evening news, the set-top-box signal looks the same.
For sure, we don’t know whether anyone is actually in the room, or whether they’ve fallen asleep with the TV on, but we do know that it’s tuned in and there’s sound blaring out into the room.
Inscape's smart TV data gives us a breakdown by the input of the TV, so we not only know whether the TV was switched on, but we know if it had a set-top-box or a broadcast antenna attached, and we exactly know what was on the screen at that time.
We can match on IP address
Panel data’s unique selling point is also its Achilles heel: it tells you precisely who’s pressed a button on a people meter to say who’s watching the show at that time. At one fell swoop, this solves the problem of whether the set-top-box has a TV connected to it and whether there is anyone in the room. At the same time, it limits the number of people you can measure to those who are willing to press a button, and it restricts what you can do with them.
Most of the internet - which now accounts for more advertising than TV - is based on building up profiles of behaviors from different data sources and fusing them together. Panel data is horrible to combine because of the tiny number of panelists. Set-top-box data is often limited in how we can join it because of both the Cable Act and also the reticence of MVPD’s to share data on their subscribers.
Smart TV data has no such limitation, and with an active opt-in from consumers, granular viewing data is available tied to IP address which we can quickly match to online cookie pools.
The IP address takes us beyond the simple segmentation that we can do with Nielsen market break segments to direct fusion from brands first-party data and custom audiences from online DSPs.
The future of TV data
Ultimately, the best dataset depends on the application for which you are using it.
We’ve seen people combining smart TV data with first-party data to provide attribution and conversion metrics against tightly defined segments. We’ve seen people using smart TV data on its own for multi-platform tune-in. We’ve seen others stick with panel data.
Those who want the best granularity, are combining all of your datasets into a single pool. Set-top-box data is the best way to understand how people are engaging with the TV when they lean forward to grab that remote. Smart TV data is the best dataset for the overall viewing experience and matching data online. Panel data is necessary if you need to precisely know who is in the same room as the TV.
Which dataset you use depends entirely on your use case. Smart TV data matches well against both set-top-box and panel data, and its granularity and speed opens up a new set of use cases that we can’t address with set-top-box data or panel data alone.