The Dangers of the "Business Outcome" TV Buzz
As someone that works with networks and their marketing and sales teams, with measurement companies, ad tech companies and major advertisers directly, I understand the urgency to transform TV advertising to be like digital -more measurable, transparent and responsive to marketers' needs, in real-time. And while I'm thrilled to be on the front lines of that transformation, I also see an industry in danger of getting it wrong.
No doubt, the new wave of business outcome-based TV buying and selling is an amazing step forward for the industry.
For the first time since the inception of TV, brands can correlate ad airing exposures to digital actions, foot traffic, and more. The hunch that TV works is giving way to the proof that there is real lift and performance in a medium known usually for its reach.
While digital can deliver millions of "targeted" views to an ad, a brand pays for reach and influences to help that well-crafted video get 5 maybe 10 seconds of time! Yet in one day, TV delivers billions of 25-30 second views. That's not hyperbole -- that's a fact. Automakers, for instance, will appear on TV screens a billion times on an average Wednesday, with a high completion rate often at a better CPM.
That kind of new-digitally enriched TV data helps car brands (just as one example) match TV populations with digital segments for true cross-channel marketing. This kind of intelligence helps identify what shows people that purchased a car are watching (or not), and target accordingly. And those same car brands can now see how those TV ad exposures correlate to local dealer website visits, lot visits and sales. The proof of TV working to impact lift comes at a time when TV needs more than #goodluckTV wishes from Rich Greenfield. Linear is fragmenting like so many other institutions we grew up with; from radio to newspapers, there is no reason why appointment viewing TV would be different. This kind of business outcome-based attribution can handle that fragmented OTT viewing- if and only if, brands are able to see it all at scale. This poses a problem for networks that have taken to selling business-outcome based lift attribution as a campaign service.
It's great that networks are working to provide brands a lift metric that proves a single TV campaign works. But there are three problems there that threaten to undermine the validity:
Campaign-based "measurement" from a data provider or from a network group is inherently flawed because it is hard to isolate ad exposures on one set of networks and OTT apps unless a brand only advertises on owned and operated properties. It's like judging the landscape of the Grand Canyon through a pay-per-quarter binocular machine. You get a limited view, with limited context, for limited time and you've discounted the magnitude of the entire landscape. What you've got is a narrow, highly perishable snapshot in time. If it is not always-on, and viewing all ad exposures on TV+OTT, how can it be legitimate? 
Self-Graded: Networks and agencies shouldn't be grading their own homework. Agencies looking to stay competitive often throw half-baked attribution data into a sales deck to a brand as an up-charge. But of course, those paid to activate media have a vested interest in a positive outcome. Increased pressure also means these sellers often can't afford the kind of expense and due diligence it takes to select the right models, controls and benchmarking required for real attribution. They are up against politics and planning tools that promise "good enough" and competition that will cut corners without thinking twice. On the network side, we've seen this disaster happen before. Facebook, for instance, has so many missteps in selling its own measurement, it now finds itself being investigated by the DOJ. Networks can provide guidance and with proper co-operation from a brand, perhaps do a better job. But ultimately buyers need impression delivery and performance validation done in house where they are free to test and compare against competitors.
Garbage In: The rush to have a fast, low cost solution usually involves incorporating questionable data. Set-top box data that is notoriously slow and messy and doesn't know or care if the TV is on is a problem with no easy fix. Snippets of audio data gathered using microphones on Android apps tied to children's games or gambling apps aren't exactly reliable and representative sources. Data that is so heavily modeled that one person, in one market, could be the actual piece of evidence a brand uses to make a decision. Raw data that isn't cleaned. Location data that is aggregated without proper controls, making it fourth party by the time it is baked into models. And if that's all not so bad, imagine making those data sets talk to each other accurately so you can make a business decision from it. You know what they say about 'garbage in'.... 
If this TV attribution thing is going to work, it needs to work at scale. And there needs to be a real conversation about what actually does work and what is a Trojan Horse.
As a wave of system integrators and small agencies rush to the market making a variety of claims that all sound roughly the same, the market will need to sort out things such as data integrity and measurement rigor. And in doing so, brands will need to resist the charms of companies that license a lot of data but cut corners on the way to delivering actual insight. This new TV infrastructure should be built on what is best for brands and for consumers, not what is cheapest or most convenient. This can’t be done by pulling in a bunch of data from various sources, throwing that through a magic context calculator, declaring it done and selling against it. And TV networks, which finally have some evidence of how they are delivering value back to brands, are anxious to "have something to say" to fight off digital. But they are smart to avoid calling sell-side attribution the magic bullet, because right now, without scale in these offerings, the only gun that can shoot is aimed squarely at the networks's feet.
