In terms of tactics that occupied the time of US digital marketers in 2015 vs. what they expected would occupy their time in 2016, cross-channel measurement and attribution grew by over 20 percent, according to an August eMarketer report. But as marketers well know, adopting measurement practices and implementing the best attribution models is often a slow and pain-staking process, and one that involves multiple business units outside of marketing to get right.
In a recent interview with eMarketer’s Lauren Fisher, our president of technology Ari Buchalter spoke about MediaMath’s focus on attribution capabilities and why cross-device is a crucial part of the puzzle. Here is an excerpt of the interview below:
eMarketer: Where do most marketers stand on transitioning to more advanced attribution solutions?
Ari Buchalter: It is safe to say that every sophisticated marketer understands the challenges inherent with the old way of doing things vis-a-vis last-touch modeling, cookie-based solutions and so forth. We are at a point where everyone agrees on the problem and is trying a series of solutions, but is hitting the limitations of those solutions and trying to figure out how to deal with that.
Over the past five years, the solution has been to build custom, multitouch attribution models that are either delivered by a third-party player that specializes in this sort of thing or to build a model yourself. But many folks are now facing challenges with the fact that the people who build these models are often not connected back to the real-time execution systems or the systems that decide how to spend the money.
eMarketer: What do marketers do then?
Buchalter: We see a lot of frustrated marketers that have models they know are better than what they had before, but the insights end up trapped in PowerPoint or Excel, and the only way they can [take] action is to make very high-level, aggregate decisions like, “I’m going to spend less on this part of my plan and more on this part of my plan.” It’s a shame, because those models are actually built off log-level, granular data.
So what we try to do is focus on the integration point that allows you to take the granular output of those models and plug it right back into the machine learning and systems that can [make] decisions on that information in real time. You can take the outputs of your multitouch attribution and have those weights move your bidding up and down, in real time, based on the characteristics of that impression. We call that “closed-loop attribution,” and that’s something we think is critical to get out of this PowerPoint and Excel paralysis.
Read the rest of the interview here.