Blog

Experts from Airbnb, Stitch Fix, and Bark Box on multi-touch attribution & incrementality

By John Koetsier November 1, 2019

Multi-touch attribution, incrementality, and lift are huge issues and opportunities for modern marketers. There’s a ton of confusion and disagreement on where they fit and how valuable they are. So what happens when you get world-class experts on marketing science together to talk about them?

You get the unexpected.

Especially about multi-touch attribution. It’s a hot attribution model right now, but is the hype justified? Does a marketer need to use MTA to understand all the customer journey touch points in order to perform accurate marketing attribution?

Perhaps not, a panel at Singular’s recent UNIFY conference said. In fact, almost certainly not … at least not yet.

“It’s a buzzword,” says Airbnb growth marketing lead Dmitry Stavitsky.

Multi-touch attribution: just a buzzword?

Stavitsky joined Bark Box‘s Todd Kane and Stitch Fix‘s Eric Kovalkoski on-stage at UNIFY. The conversation, moderated by Singular CRO Vince Cortese, started on cross-platform and cross-channel attribution. It quickly zeroed in on multi-touch attribution.

“It’s the holy grail of marketing,” says Kovalkoski. “Our data science team thinks — and I would agree with them — that it’s hard to trust. It’s still a ways out.”

One challenge, Airbnb’s Stavitsky said, is what you’re using MTA for. And most people aren’t using it for the right reasons, which seriously limits multi-touch attribution’s ability to provide value.

“Some teams look at MTA as a way to re-slice the pie,” he said. “It kind of makes one channel look better, or one team look better … but the pie doesn’t get bigger. The key is to use multi-touch attribution to make the pie bigger. Ideally, MTA shows you some level of truth as to how your world of marketer works. In other words, what your customer journeys actually look like. What customers actually do. Where they actually touch you, or your marketing.”

Using multi-channel attribution should highlight more marketing touchpoints, places where your campaigns have impacted a customer. It should provide more insights than a linear attribution model, and should help you adjust your marketing mix. And it should provide better analytics on future marketing campaigns.

All of this, however, is not easy. Hence the “buzzword” comment, despite the real value that multi-touch attribution promises.

But one thing that is definitely not a buzzword is incrementality.

Incrementality: not simple to do right

But even if multi-touch attribution is not working amazingly well for you — and outside of B2B, not too many marketers have truly figured it out — incrementality is well within the realm of possibility. And incrementality does provide good insights into the value of your marketing efforts in each of the marketing channels you test.

“Our north star is incrementality,” says Stitch Fix’s Kovalkoski. “Everything we do is based off of randomized control trials.”

Sometimes called uplift, incrementality is testing the specific impact of one variable in a marketing campaign. When used in reference to ad networks and marketing partners, incrementality is the degree to which new partners access new audiences … and can engage with them.

“Five years ago when you were talking to an ad network about incrementality, they probably didn’t understand what you really meant, and what you wanted from them,” says Airbnb’s Stavitsky. “[They’d say] ‘of course it’s 100% incremental … of course these are all brand-new users that you would not find anywhere else … work with us!'”

But incrementality, like multi-touch attribution, isn’t easy to get right.

Even if initial results are positive.

One of the scenes at UNIFY 2019

“We started to experiment with it five years ago,” Stavitsky said. “When you run your first series of lift tests it goes up and to the right, and [you feel] … now I understand how the world works. [But] when you continue running those tests … the more you know the less you understand.”

The results generally look good initially, but when you keep running lift tests, reality becomes more and more complex. Your marketing analytics become more and more confusing. There is hope, however. Eventually, Stavitsky said, you settle into a scientific method of marketing: doing Y, looking for X.

Experimenting, gathering evidence, and proceeding based on the results.

There’s art as well as science

One challenging thing: it’s a little different for everyone.

Different businesses with different products, revenue models, customer profiles, and paths to monetization will need to figure this out for themselves. A different customer journey means a different model. Different touch points change the math. So there’s science … but there’s also art.

One very real challenge that marketers deal with is channel exhaustion. And that’s something that marketers simple have to measure themselves, says Kovalkoski:

“One big channel that I can’t name on-stage … was trying to prove value to us,” he said. “But we did the analysis ourselves and found that the actual incremental CPA was actually 3X higher.”

Holdouts: the next big thing

Multi-touch attribution is hard. Incrementality is too, but it’s at least achievable for many marketers, and most ad networks have ways and means of enabling testing. Marketing mix modeling may not work for all.

The next major mountain for marketers to climb might be holdouts.

Holdouts might be more common traditionally in email marketing, where you can simple leave a percentage of your list off a series of communications in order to check if your newsletter or communications are actually successful in generating lift. But holdouts are also tremendously important in advertising campaigns.

And while marketers have been successful in fighting for incrementality and lift test, they shouldn’t rest on their laurels, says Stavitsky:

“The next thing we should be asking for is regular holdouts. I want a 5% lift test holdout always on … keep telling it to Facebook, keep telling it to Google, keep telling it to ad networks.”

Adding offline marketing to the digital mix

Interestingly, many of the most mobile-first or digital-first companies are now also hitting offline marketing channels … making multi-touch attribution and cross-channel measurement even more important.

(And opening the door to mix modeling and other marketing attribution models.)

“We’re launching physical gift cards this holiday period,” said Bark Box’s Todd Kane.

Bark Box is looking forward to learning how this impacts the customer experience. Bark Box customers are the company’s best sales people, Kane said, and the company is trying to understand what that looks like in physical retail spaces as opposed to digital marketing.

And that brings up a whole new list of challenges — at least for digital marketers — that will remind old-school professionals of their early careers.

“We were very proud because our online side of things is very much evidence based,” says Airbnb’s Stavitsky. “But at the same time, the offline side of things was very faith-based. Faith-based marketing … if you believe that it works, it works … but over time there were less and less believers.”

Ultimately … it comes down to clean data

Whether you’re doing multi-touch attribution or incrementality or lift testing or holdouts (or all of the above) it ultimately comes down to data, the panel agreed. If a marketer has the right data from his or her marketing channels, they’ll be able to use multiple marketing attribution models to understand the consumer journey.

Because data fuels marketing attribution. Data informs marketing insights on customer touch points. Marketing data enables cross-channel attribution. Data tells you which campaigns were successful. And whatever attribution solution you use, whatever attribution model you follow … data will inform your future marketing mix.

Good data.

Clean data.

Portable data.

And whatever strategies you take to understand your customer and your customers’ journeys will be improved by that good, clean, portable data.

“A couple years from now there will be another shiny thing that everyone will be obsessed about,” says Stavitsky. “I would do one thing: structure my data and build my systems in such a way that it’s easy to plug any attribution model in: last click, MTA, or MTA version two tomorrow …”

And that’s something that Singular can help with.

Get a first-hand look

Want to get a first-hand look at how companies like Airbnb, Stitch Fix, and Bark Box got so good at marketing? How they do marketing attribution and cross-channel attribution? What data-driven insights they generate from campaign data, social media, organic search, and other data-driven marketing?

Whether you use multi-touch attribution, last-click attribution, or other ways of attribution modeling, Singular can help you get data-driven insights from digital channels, online and offline sources, and all your marketing campaigns.

Get a customized tour of Singular’s technology today.

Request a demo

Or, watch all of the UNIFY 2019 sessions here, in our UNIFY @ Home masterclass

Stay up to date on the latest happenings in digital marketing

Simply send us your email and you’re in! We promise not to spam you.