By Ben Maughan, VP of Product Management, Advanced Media and Advertising, TiVo 

Unless you’ve been vacationing the past twelve months, you’re well aware that the linear television advertising ecosystem is finally changing. There are new ad inventory units, new measurement methods and new point-to-point delivery capabilities. A huge driver of this change has been the explosion of readily available, actionable, linear-TV viewership data offerings that are comparable to the type of data only available from online video viewing.

Only a few years ago, there wasn’t the amount of household-level and even device-level viewership data for linear TV that there is today. Two major sources of this type of data are Automatic Content Recognition (ACR) viewership data and clickstream, set-top box (STB) viewership data. The former is primarily sourced through IP-connected devices, such as smart TVs and mobile screens, and the latter is primarily sourced through two-way set-top boxes leveraging the ALOHA network back channel (my video engineering colleagues just cringed at how much I simplified that).

Now, with all the hype about these two data sources, it’s about time for some real talk that outlines the pros and cons of each.

Let’s start with ACR-produced viewership data.

Pros:

Available Scale.  There are more homes in the US that have clickstream capable set-top boxes than there are with ACR-embedded smart TVs. But, when it comes to available, licensable viewership data, the single largest available dataset from a single provider is ACR-produced. However, don’t be fooled when some ACR-companies claim they have tens of millions of devices – they may be stretching the truth. Their software tends to be dormant on tens of millions of devices, and only a fraction are actually reporting actionable viewership data. Make sure you have “real meat in your hamburgers, no filler.” But, all in all, if you want scale, ACR-produced viewership data may be your best bet.

Fast.  ACR-produced viewership data requires an active internet connection to recognize the content via watermarks or fingerprints. This near real-time recognition means that the resulting viewership data can be processed, cleansed and redistributed very quickly. If you’re looking for near real-time cross-platform retargeting on data that isn’t perfect, but mostly right, you should consider ACR-produced viewership data.

Cons:

False Positives. Most ACR-produced viewership data relies on audio- or image-based fingerprints to recognize content. I’ll spare you the unpleasant details, but many ACR-based viewership data companies use several dozen digital satellite receivers to index (aka fingerprint) live, national TV, and then usually use a 3^rd-party company for the top 10 to 30, sometimes up to 50, DMAs. So, for ACR-produced viewership data you need to a) have original content fingerprints in your index and b) not have multiple places that content can be viewed at once. It’s worth mentioning that second bit because ACR software can struggle to figure out which airing was which; this doesn’t work well in non-linear worlds with simultaneous airings, time-shifted viewing, program reruns, or multiple airings of the same ad. And as we know, this is how the world is currently viewing video content. That being said, ACR-produced viewership data shouldn’t be the only source in your licensed viewership dataset when you’re trying to do real measurement, campaign planning, attribution, or true targeting.

Now, let’s talk about clickstream-produced (STB) viewership data.

Pros:

Accurate.  Return-path viewership data is by far the most accurate source of viewership data. . A set-top box is either tuned to “channel 2” and watching “NBC,” or it’s not. You can’t circumvent the tuner in this case. A set-top box either played back that exact recording or it didn’t. I’m either watching Judge Judy on the CW at 2:35 PM, or I’m watching Judge Judy on CBS at 2:35PM – the set-top box always knows which of the two it is that I am watching. If you want more accurate viewership data, then return-path produced viewership data must be in your licensed viewership dataset.

Authoritative Data.  Typically,  return-path viewership data doesn’t rely on IP-addresses to match viewership data to a household.  Clickstream-produced viewership data aggregators leverage trusted third-parties such as Experian or Acxiom to handle personally identifiable information (PII) to create blind matches between anonymous identifiers which are then associated with viewership data. The cool thing about this is that Experian or Acxiom use authoritative video service provider household information to accurately, but anonymously match a household’s viewership data with non-personally identifiable information such as gender, age, presence of children, etc. in a privacy-compliant manner.  This combined data is incredibly useful to marketers and keeps consumers’ information safe as well.  If you want data that is authoritative and drives real attribution, then clickstream-produced viewership data must be in your licensed viewership dataset.

Cons:

Not as Fast.  Clickstream-produced viewership data does have its faults. It hasn’t reached its full potential with recency (the delay between when a viewership event happens and when that data is ready to be consumed by third parties). Because the ALHOA network was built in the 1970s, the back channel to the video head-end is temperamental, to say the least. Sending massive amounts of data in real-time back over the coaxial cable during prime time viewing or prime time video on-demand viewing could stop any MVPD’s VP of video engineering’s heart.  Therefore, video engineers in the 1990s set up most set-top boxes to send clickstream viewership data every night around 2AM over the span of several dozen minutes in order not to collapse the network.  Fast forward to today, where most set-top boxes are hybrid set-top boxes (meaning they have an ethernet port), clickstream should theoretically flow back in real time over the unmanaged network, not the managed network.  If you can handle data that is accurate, but delayed twenty-four hours, then clickstream-produced viewership data must be in your licensed viewership dataset. 

Truncation. One thing clickstream-produced viewership data nay-sayers always point to is the situation where a viewer turns off the television but leaves the set-top box powered on. Those critics often say, will say, “how do you know if they are really watching TV?.” Having looked at thousands of viewing events myself, it’s quite easy to understand if someone is present because of one simple fact: humans constantly change channels. The moment we stop changing channels is typically when we can gather that the TV has been turned off.  We’ve run extensive analysis at TiVo, compared that analysis to industry benchmarks and have found that if you truncate viewing events that don’t have activity for longer than two hours, your overall viewership ratings can easily be correlated to third-party ACR or panel-based ratings. In short, this is a red herring that shouldn’t prevent you from ensuring clickstream-produced viewership data is part of your licensed viewership dataset.

The Verdict: 

In reality, no viewership data set is perfect. If someone tells you that their viewership data set is perfect, let’s hope it’s the result of a lack of experience rather than bad intentions. Pick your viewership data sources and types based on your use cases. Having both types won’t hurt.  All data sets are complimentary, and knowing more about your audience will just make your goals that much easier to reach.