April 25, 2016

Adweek "Stop Waiting for Perfect Data to Solve Your Attribution Woes" - Brian Stempeck


Brian Stempeck

When something is overwhelming and complex, people gravitate to simple solutions. For example, free trade and immigration are messy issues at the forefront of this year's presidential race. Yet 'build a wall' has emerged the fix-all campaign promise from a leading candidate.

Knee-jerk answers to nuanced problems are easy to mock, but a similar thing happens in marketing all the time. 'Which part of the media campaign worked best? Well, the empirical multi-touch attribution model is pretty complicated, so what's the last ad someone saw or clicked?'

Google, among others, has built an empire based on being the last click before a sale. It's easy to justify pouring money into the marketing channel that has the clearest direct link to revenue, even if a TV ad drove that search query in the first place.

We, in the ad tech industry, have done a pretty bad job at coming up with a good alternative to last click or last ad seen. I think our failure stems from a belief that the answer is always, "We need more data."

In the ideal world, a marketer could measure every single advertising touch point on every single consumer, and then determine what actually drove success. And if you look at the recent innovations in ad tech, such as cross-device and beacons, many of the new offerings try and solve that challenge.

But if we're all honest with each other, there are still giant gaping holes in our collective data set. Did anyone track the subway ads for Seamless I saw this morning? What about all the ads I saw during the college basketball tournament, delivered on a national broadcast, with no household or cookie-level tracking?

Many of the most effective ads still can't be tracked. Some will never be trackable. One of my favorite cocktail party anecdotes—the cocktail parties I go to are clearly not that fun—is that for all the perceived growth in e-commerce, people still spend an awful lot of money the old-fashioned way: Target and Wal-Mart only get about 3 percent of their sales online. E-commerce is about 7 percent of overall U.S. retail sales.

Because the marketing attribution path is missing some pretty big pieces, it hasn't become the killer app. In too many cases, I see marketers and agencies throw up their hands and say, 'Well, since we're still waiting on half of the inputs we need for a good attribution model, let's just talk about click-through rate.'

The average marketer does not have a decade to wait for the pieces to be stitched together. So what can we do to stop letting perfection be the enemy of the good? Here are three things I see advanced marketers and their agencies doing today:

When I worked for a large consumer goods company, we had amazing data about sales of every single product: By SKU, by shelf, by store, by zip code. Yet most media agencies and their tech partners get little visibility into this depth of data. If moving the needle on offline sales is the objective, the media buyer needs insight into what's working and what's not. It doesn't have to be at 1:1 consumer level, but if sales are plummeting in Topeka but spiking in Tucson, the media buyer can take action based on that knowledge. We've been working with our clients to pull this information, even in indexed form, straight into our platform so we can optimize media based on it.

One of my favorite companies in this space is called Cardlytics—the pitch is simple. You pass Cardlytics a data dump of all the cookies or devices that saw ads for a particular campaign, let's say Home Depot, and through its partnerships with credit card companies, Cardlytics can measure how many of those customers bought something in-store. That 100-lb bag of bark mulch probably got sold in-store, so why are we all optimizing to the e-commerce transaction?

Some of the sharpest brands and agencies we work with are ruthless in how they gauge the value of new media channels. They put their money where their mouth is. For example, how would convenience store sales of Doritos fare in suburban Denver when Snapchat is the only media channel being used? If the sales lift is isolated, and real, the brand will take it national. But it's a high bar to clear, and some marketers might pause at making one market a potential sacrificial lamb for advertising ROI.

Marketing measurement has come an enormous way, but we are still missing the forest for the trees. Next time you catch yourself or a co-worker resorting to CTR or last click when discussing attribution, see if there's an offline metric or a new way to test media channels instead. Remember: You're better than the pandering politician.