For many in marketing, deploying multi-touch attribution is more of an aspiration than a current reality, but that doesn’t mean we should defer efforts to find a complete attribution solution. There are many considerations when constructing a workable model, including variations in the consumer journey based on products and categories, the impact of unmeasurable factors like word of mouth and the ceaselessly expanding choice of datasets — some valuable and some, depending on the advertiser and category, not so much.
I had the opportunity to sit down with Matthew Krepsik, Global Head of Analytics for Nielsen to talk more about facing the challenge of attribution.
Charlene Weisler: Nielsen has always been proud that they own their data. Can you talk about the new reality where Nielsen will need to go beyond their own data and partner with other data suppliers who, in turn, control their own datasets?
Matthew Krepsik: From our point of view, we have a lot of unique and valuable data, just like other suppliers with walled gardens have incredibly valuable data. What we see as the biggest opportunity is our ability to co-mingle those data sets. If we think about our constituents and our users, whether marketers or other business executives, what they really want is greater intelligence. We understand the complementary nature of all of these datasets and we can bring them together for specific use cases and help in enabling desired outcomes. This is where we generate the most value. As we think about the next generation of growth and innovation as a company, we think that the innovation stems from building across different partners. We have opened up access to our data with partners to make it easier to use and more permissible for marketers and brand owners across the value chain and in places where we don’t operate.
Weisler: How do you manage the de-duplication of data when you use other walled garden datasets?
Krepsik: There are a couple of dimensions of de-duplication around identity, around devices and around the cookie itself. So if you think about the ad model side, the ad tech really starts with the cookie. Your phone right now probably has more than 50 or 100 cookies on it. All of those cookies roll into a device. That device has to roll into other devices. So de-duplication is improving right now. Is it perfect? Not at all. But the first challenge is getting from cookies to devices and devices to people. I would say that we are getting more robust at the device level. The challenge is that they keep reinventing and updating the device. Every customer is getting a new phone. They are getting new laptops. They are getting new devices at home. Most households get new devices about every six months. So the challenge is that we have gone past the period of “build” and we have to constantly reinvest in the updating. The technology we have makes it easier to do this than ever before. The bigger challenge is de-duplicating your on-boarding data. That gives us a tremendous opportunity to get better at connecting to a digital consumer from an offline consumer. There is still a lot of work to do there.
Weisler: Looking at multi-touch attribution, where do you see it today, grading it from A to F?
Krepsik: I am going to answer this question along two dimensions. When I think about the overall need of a marketer or a CMO, I would give most attribution models a grade of D. If I think about the overall need for digital media planning, I think attribution models today are approaching a B+ / A- grade.
I say this because for a digital media manager, what you really want to know is whether this creative, or this site, or this device, or this placement working better than another one? Is this audience working better than another one? How do I make decisions and trade-offs across all of the possibilities out there? Today’s attribution models are really good at allowing digital owners to understand the cornucopia of media channels and how they can get more improvement out of them.
That being said, from an overall CMO standpoint, out of every revenue dollar that goes to the cash register, 25 cents is spent back on some form of marketing investment. Digital is only two cents of that quarter. So what they want to know is “should I be spending three cents on digital or should I be spending one cent on digital?” If you think about most attribution models today, they are not expressly measuring incrementality. They are not taking into account a lot of the “last mile” problems.
Weisler: Where do you see attribution going in the next couple of years?
Krepsik: Where I think the attribution industry has a chance to grow and where the marketing mix industry can take a step forward is bringing those two pieces together. Attribution models bring speed and granularity together and the marketing mix world offers scale and coverage. Where I see the industry going in the next two years is that both of those pieces come together; Leveraging the technology, real-time nature, and granularity in the attribution model with the scale, coverage, and sophistication of marketing mix modeling.
This article first appeared in www.MediaVillage.com