Advanced TV Targeting: A Small Step Forward

On TV And Video” is a column exploring opportunities and challenges in advanced TV and video.

Today’s column is written by Chris Peterson, managing partner at R2C Group.

I’ve sat through at least a dozen TV ad tech presentations that promise pretty much the same thing: the complete transformation of customer targeting for TV advertising. While the pitch outlines and content are all about the same, the user interface of the PowerPoint and their respective software vary.

The stated problem to solve is that TV targeting today is incredibly vague, typically employing broad demographic criteria, such as “adult 25-54.” No argument there.

The stated solution is to apply “advanced data sets” to target customers like never before. OK, show me more.

Typically, the targeting includes taking a brand’s customer data and bumping it up against one of three TV viewership panels; there’s Nielsen’s own data set and viewership data from either cable set-top boxes or connected TVs. Marketers should find out what customers watch the most and use that to guide the media plan and go get more of them. It’s a basic lookalike model that has been employed in digital media for years.

By the way, am I the only one who wonders how cable set-top box data can provide accurate viewership data? Cable boxes are always on, so how do you know if the TV is being watched? I’ve been told that there are “sophisticated algorithms that look at changes in volume and channels in order to precisely determine if someone is watching TV.” I’ve asked several times to get some kind of back-up for this – anything – but no one seems to have it. Until I get something, I’m prone to think that this data set is skewed by people who fall asleep in front of the TV or have dogs that jump on couches.

Let’s assume that somewhere, someone has figured out that looking at volume and channel changes does, in fact, give you some indication of what a household is watching. There are still a number of problems that make “advanced TV targeting” a small step forward. These issues tend to come to light when you actually try to apply this kind of targeting to TV in real life.

First, you learn that there is cruel math related to these data sets that prevents you from learning much. Let’s say, for example, you have 500,000 customers and use a cable set-top box or connected TV data set to identify what customers watch. For simplicity’s sake, let’s say the viewership panel has 4 million households, which is roughly 4% of about 100 million TV households. This means you will have a read on 4% of your customers. In this example, you end up with 20,000 customers.

Using 20,000 customers should be fine for understanding what networks customers watch because there are only 180 national networks. But, there are 16,000 programs a week that can take a 30-second ad. This means when you get down to the daypart or program level, you will probably end up with two or five customers who are watching. Hardly enough for significance, right? Getting insight at the network level is nice, but hardly earth-shattering. It certainly isn’t going to help fine-tune a media plan that much.

The second problem related to advanced TV targeting using customer data is that the results reflect a composite view of the customer. That is, the viewing data averages all customers. The vast majority of brands tend to have distinct segments with different profiles. If you attempt to segment your customer database before applying viewership data, then the math gets far worse.

The third problem with this type of targeting is that it only reflects the past with no insight into the future. If we’re going to apply the word “advanced” to “TV targeting,” I would hope that there are ways to understand where a brand should go to grow revenue versus reflecting what’s already been done. For some brands, this will be OK because market penetration may be low, but at some point, it becomes an issue.

Don’t get me wrong: You can learn some things with advanced TV targeting, but the value just isn’t that high right now. To make advanced TV targeting more sophisticated using customer data, brands either need to get millions more customers or the data sets need to grow exponentially. Neither will happen any time soon, so we are a bit stuck.

In spite of a lack of targeting, TV advertising continues to grow, albeit slowly. Certainly, there are big changes in consumption afoot with drops in live viewing among younger audiences and correlated rises in over-the-top viewing. But the old black box on the living room wall still moves needles for a lot of businesses. The typical ad unit – the 30-second spot – seems to have enough power to overcome the shortfall in targeting. Imagine running digital advertising with targeting that is “adults 25-54” – you’d lose your shirt. The quality of the unit demands really great targeting.

The ability to target TV will only get better, so it will be interesting to see how the medium fares in general as it dukes it out with digital media for market share. If Google and Facebook, which account for 85% of all digital spend, ever hit a wall, then the momentum might shift.

Follow R2C Group (@r2cgroup) and AdExchanger (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!

1 Comment

  1. marshall wong

    The author is concerned about using advanced TV data sets because a dog can jump on the couch, step on the remote and change the channel. If that’s one of the top concerns around advanced TV data efficacy, sign me up.

    A few more technical counterpoints to the author’s framing of advanced TV targeting, which in the context of this response will focus on audience indexed TV buying

    Counter to what the author expressed, audience indexed TV buying in many cases does not rely on look-alike models. The programs, networks and dayparts that have the highest density of viewers a brand is trying to reach are identified using matched digital to TV data sets and assessed with cost and reach to form a media plan.

    The first problem the author sites centers around concerns around the scale of the data being indexed. Brands can and index all digital 1st, 2nd or 3rd party audiences in addition to offline audiences. Brands are not restricted to indexing current customers – in many cases new customer segments are the goal – so assembling strong audience data sets, with sufficient scale, that align with the brand’s objectives is pretty straight forward. In short, if a brand is having success with a specific digital audience in its digital marketing, that same audience can be used to index its TV buys.

    The second problem raised is around generating TV indexes against a composite of a brand’s customer segments. We covered above why scale is not a concern when assembling audience segments for indexing. In short brands can absolutely have different indexes run for different brands and creatives.

    Lastly, the author laments the ability for advanced TV to provide insight around where a brand should “go to grow revenue.” Unpacking that a bit, historical data is all we have to inform the future. The beauty of modern indexed based TV buying is it wildly compresses the timeline needed to inform future decisions. Fresh digital audiences can be indexed in a day and on air in under 2 weeks as opposed to legacy techniques which can take months and sometimes up to a year. And to top it off, audiences can be A/B tested using new TV attribution techniques which have finally come into their own.
    All of which raises the question: what is the cost and process overhead for diving into audience based TV buying and is the juice worth the squeeze? That is the discussion I believe we should be having.

    Marshall Wong
    Rhombus Technologies