AdExchanger: Who are you partnering with for programmatic TV inventory?
JIM CARUSO: We’ve been working with AudienceXpress, AdMore and Clypd the longest (on the supply side), but there are about a dozen TV sources we have access to. There’s real value to be had in targeting and optimization, and being able to say, “Instead of placing a TV buy and waiting however long, I can turn this campaign around in a day.” With some of the TV platforms, we’re now able to get day-old delivery data and say, “These are the markets and day parts which are working and start to optimize accordingly.”
What’s your biggest challenge buying television inventory, programmatically?
How granular the data is and how quickly you can get it. With programmatic (display), we typically get 30-40 data points about every impression we serve. Some of these platforms aren’t able to give us that data, so sometimes they have to roll it up and tell us they served 50% of the buy to the top 10 geo markets whereas we’re getting ZIP code-level data from our other buys. Both are improving, and a year or so ago it was challenging. The providers did not want to give any transparency into the data to the platforms, but I think that’s changing.
Why do local markets seem more amenable to programmatic buying?
A lot of that is because of the MVPDs and how that inventory is packaged and sold and who has the rights to sell it. The two minutes of local air time on these cable stations is more easily accessible, whereas a lot of the stuff at the national cable or broadcast level is packaged up and sold well in advance in the upfronts or scatter and there’s reluctance with price and availability.
If they decouple this inventory for programmatic sales from their standard Nielsen rating, they want to know, “Can I compare this and make money on this?” A lot of this is channel conflict. They are very strict with what advertisers they’ll let us work with if they’re already working with them directly.
What are the most common metrics or data points you’re applying?
In addition to Nielsen and Rentrak, a lot of clients are asking us to layer in existing or new purchase behaviors. One client wanted us to target people who bought certain kinds of their products before, but not others. This is where programmatic TV is going. You get the sense people feel there has to be a better way to do this.
If they invest in developing 15- or 30-second spots, they should be able to target them effectively not only when they’re in front of a TV, but on YouTube and elsewhere. If the best bet is serving an ad on TV in primetime with a direct buy, how do you make the rest of your buy smarter? Making sure the TV portion is informing the digital portion of this, or how digital video can help you better optimize your TV buys.
What does it mean when a video demand-side platform calls you a preferred partner, as Tremor Video did?
The preferred word is a little bit loaded since we have a lot of platforms we test and some tend to work better for certain campaigns and clients than other, and sometimes we develop deeper relationships with them to get access to certain inventory or to get access to certain features.
But evaluating DSPs is a big part of what we do, and we do that based on a number of criteria – their interface, access to reporting, can we get log-level data in real time? There are about three dozen different variables that weigh in to decisions about which DSPs we want to work with and which ones we want to use more.
How frequently do clients come in with a preferred DSP? And how often are you making additional recommendations?
Agencies are typically pretty flexible, but brands may have partners they’ve worked with in the past. Even if they have a DSP they prefer, we say, “Can we try at least one or two more partners?” We take budget and test other guys to expand results. Maybe 25% of the time they come to us, and want us to run campaigns through a specific platform.
How do you measure?
We use Nielsen and comScore with about a dozen other viewability vendors. We want to be able to give clients the right mix. A big problem in the industry is clients might want to use DoubleVerify, Integral Ad Science and one other one.
But if they’re using IAS to pre-bid, sometimes they’re using comScore to measure. So you start to have these permutations where you have to test five vendors, but it’s almost 25 combinations you have to go through. That’s where a lot of this gets tough. How do you match this all up? Sometimes we use a solution that does both pre-bid and post-bid reporting, and you get two radically different sets of results.
I’d love for this to be at the ad-server level. The argument [against embedding this at the ad server level] is “let’s not give them any more data.” But at the same time, if we could handle this at the ad-serving level and not have to wrap all these extra tags that give you all these different results, it’s a nightmare operationally to handle.
Google acquired Spider.io and they obviously have that as part of their offering and its not fully baked, but the whole process is so disjointed right now. I need something we can all agree on and something that can be baked into the ad server.
How are you talking to clients about viewability?
That’s where we do a lot of education for clients and help them understand there’s a lot of fake traffic out there, but it’s more about helping them navigate it while setting the expectation that you’re not necessarily going to eliminate it completely.
A lot of clients task us with very specific outcomes. It may be a bank that wants to drive sign-ups for checking accounts or to drive more sales of paper towels – very specific actions that a human needs to do. We’ve had to educate them about there are some impressions that are fake and some that aren’t viewed, but as long as we’re providing good results, there’s a balance. As we do more video, more brand awareness, brand lift campaigns, those metrics are where you can kind of get bogged down in this malaise of fake traffic.