Attribution Wars: In Defense Of The Last Click (Part I)

I went to a party last week. It was put on by a big online player, mainly to thank agencies who buy large gobs of display ads for their clients. And strangely, one of the hosts began apologizing to me (over the gurgling sounds of me thanking him for the free drink—how they let me in there, I’ll never know). “As an individual, between you and me, I’d never advocate a lot of this inventory. Hey, I come from the performance side. I worked [in a senior position for a search advertising platform].” But in his current role, he was doing a pretty good job; thanking the buyers of digital display media is a fun job, made easier by their rampant demand for it.

Search ads sell themselves with performance, to those who obsess about it. Venue: spreadsheet or analytics interface. Brain usage: left brain, mostly.

In a funny way, display also sells itself, to agencies and advertisers who—for other reasons—just plain want it. Venue: the social scene. Brain usage: right brain, mostly.

Are those two worlds about to collide? Maybe they are, and it could be the death of those thank-you parties. You guessed it. The display people are now armed with spreadsheets, analytics and attribution models. And while the search nerds go rock climbing to beef up the neglected studly parts of their personas, the guitar players, prom queens, and jocks at the agencies want into the nerds’ world. Armed with slick studies and recently-purchased designer specs, they’re banging on the door of performance marketing. It’s high school in reverse!

But are there really lenses in those glasses? And are the new attribution models favored by the display people making any sense?

One of the got-good-grades analytical types, Wister Walcott of Marin Software, doesn’t think so. At SMX East recently, he threw down the gauntlet. At 6-foot-8, Wister is a nerd you don’t want to mess with.

Leaving display aside for now, and focusing just on the paid search side of the equation: a lot of people now believe in the myth that for accurate bidding, you would need to look at clicks prior to the last click. Walcott, using data from brand etailers Marin Software works with, busts that myth.

Problem: 74% of purchases can only be attributed to a single click anyway. By definition, that’s the last click. Attempt to “attribute” that purchase to prior clicks or other factors? Good luck.

Beyond that, when multiple clicks can be tracked, you don’t find confirmation of the convenient “purchase path” fairy tale that people begin with one type of search (say, a generic term) and wind up with another (say, a specific branded term). Some buy cycle behavior may resemble this stereotypical model, but on the whole, the real picture of what influences people is much more complex than that.

So the next revelation: when you bucket keyword queries down into distinct types, only 7% of sales can attribute significant influence to some kind of click other than the last click.

Ergo: unless you’ve got some awfully compelling evidence to the contrary, paid search marketers, don’t be dislodged from your current practice. Which is, as you know, relying on performance data associated with the last click prior to purchase.

To be sure, there are probably several shortcomings to this case study. It makes no attempt to latch onto users over extended periods of time, doesn’t refer to a variety of potential influences such as email, direct navigation, display ads or even organic search.

But that scenario mirrors the real-world optimization environment you work in as a paid search marketer. The choice is clear: you can react as best as you can to (admittedly imperfect) data that is pretty close to 100% certain, or you can rebalance your keyword spend portfolio based on little more than hand-wringing and hearsay evidence.

So what might be “compelling evidence” of purchase behavior being influenced by various touchpoints? Don’t expect consensus here, but do stick to your instincts.

Former Canadian Prime Minister Jean Chretien perhaps said it best: “A proof is a proof. What kind of a proof? It’s a proof. A proof is a proof. And when you have a good proof, it’s because it’s proven.”

Incoherent babble, on one hand. But it aptly captures the determination to stick to your guns, for now, when you suspect you’re in the process of being bamboozled.

And how might we gather better evidence to better confirm our suspicions one way or the other? Two ways: by trusting complicated studies that aggregate other people’s results, or by gaining better tracking capability and looking at your own analytics data.

Most advertisers employ popular analytics solutions (and paid search platforms with reporting) that lack the capacity for multiple attribution. When current analytics tools make this new functionality available to the wider marketplace, you’ll get closer to being able to verify competing claims for yourself—as long as you understand how the technology works.

For maximum accuracy and comprehensiveness of user behavior tracking, these solutions are likely going to have to come from huge companies like Google that have access to more behavior data. Since the DoubleClick acquisition, and as their content networks and exchanges gain reach, their potential to develop a tracking standard has increased even more. That said, this will run up against privacy issues, especially in international markets.

I suspect it won’t be much longer that we’ll all be fumbling along using ecommerce attribution data from Google Analytics’ quirky means of tracking behavior from multiple sources, comparing it to our stats in AdWords, and wondering if there isn’t a better way to track things, and to customize our models.

Prediction: since some competing vendors have already raced ahead in this regard, then it won’t be much longer before Google plays catchup. Remember, buyers and sellers of media have often sought a mutually-agreed standard to audit performance: that’s what the “DoubleClick era” was all about back in the day. As with analytics generally, a shakeout is inevitable here. That shakeout will accelerate when Google releases products that address the new debates about multiple attribution.

I hope I’ve piqued your interest in this debate, but it’s just getting warmed up. Next time I’ll identify and assess the four major “camps” in the attribution wars.