Attribution has always been challenging and it keeps getting tougher. Here we have two executives from different segments with different perspectives. Just what is the thinking on attribution nowadays?
Attribution has always been a fun conversation,” says Healy. “Marketing mix modeling has been a tool that we have used historically, we’ve been able to rely on it a lot. As it pertains to media planning, understanding where value is really coming from in the business, just the shifts in where media spend is going is very direct in terms of things like ecommerce, social media, walled gardens–there’s all of these areas that have made attribution even more difficult than it already was. And these are areas that are growing the most, but also the areas that we have the least amount of rigor. We’ve had so long to optimize how we understand marketing mix modeling on linear TV, for example. And now trying to figure it out on different sites, different social media influencer platforms, and how those things attribute to sales week on week has been extraordinarily difficult.”
What about the changes happening in the market such as GA4 coming from Google, and a cookie-less future in front of us. The future looks just as chaotic as the present and the past.
“It’s a really interesting problem, and there’s no perfect idea or clear path forward,” observes Zhu. “There are several trends I’m observing, such as the GA4 migration. Should I continue using GA4 and take all the efforts for migration, try to link all the historical data with those events, go and try to learn with the new tool—or should I just start brand new and migrate to so many other tools in the market. Start thinking about the long-term data strategy. What is the right tool we should enlist to solve the problem we really need to solve. What kind of data do we need to collect? The second trend is to assess the cookie-less future. It’s definitely a trend that people realize that third party data is not really reliable. How can we purchase reliable good second party data? And people started thinking about what kind of first party data we should invest in. Get the right technology. Another trend is definitely machine learning. The last trend is a focus on customer loyalty. They start thinking in a cookie-less future, the attribution is not clear. What we can do is we can build up loyalty, invest in CRM, really have a 360 view of our customer.”
Healy agrees with Zhu’s assessment of the trends. “These are things internally that, within PepsiCo, we’re working on a lot, in terms of first party data, direct to consumer information, how we connect with consumers. It’s such a huge puzzle to solve. And now it’s like, we’re just going to add three times as many puzzle pieces to the puzzle you already had, now figure it out,” he says.
There’s multi-touch attribution. Marketing mix modeling. Walled gardens. There is no one silver bullet solving everything. How can we actually come up with any value for anyone along this value chain.
Healy notes, “There have historically been companies that did not want to share their data. And then over time, what we were able to do is show the value that could come of a partnership of shared information. I totally understand where a walled garden perspective comes from. Everybody’s going to look for their angle. And what you have to do is to show that through the correct partnerships, you can trust us, let’s build those sorts of honest relationships that are built on this idea of mutual success. I know there’s some solutions out there. There are tools like that, which maybe you can then extrapolate into some marketing mix performance, but it just adds another layer of plus minus to a system that already has that built into it. I think one of the criticisms of a marketing mix model is it’s not exact. Every attribution model has a plus minus. And in every kind of stack that you build on that, creates an even bigger stretch. We just have to come together a little bit more in terms of creating systems of mutual success.”
Zhu points out, “I think it has come down to the regulations or if there is a centralized best practice in terms of the data sharing part. And, also, there is the ethical part–who owns the data? There’s no common goal. That’s why it’s hard to have a win-win situation. Regarding the technology stack, it’s hard to say this is the blueprint. A different team is needed to think about how the tech stack really fits. What’s your size? What’s your budget? What’s a scalable solution? What’s your business and data strategy? What are we empowered to do in five years or ten years? You have to really align with that. Another angle is to integrate fast to the puzzle. To Doug’s point we already have a puzzle. And now we cannot just throw away everything and get you the new puzzle. For your current puzzle, what are your strengths, your competitive advantage, what are the things you want to add on? Maybe the most important thing is to be transparent.”
Healy says, “Bring in a lot of different groups. Maybe that was part of the solution with marketing mix modeling too. Find that unified goal and how do you bring people along with you? We’re asking a lot of questions about tech, and we’re asking a lot of questions about how, but maybe it’s a what? If we can bring different minds together to your point of figuring out that stack, it’ll happen. But don’t do it in silos. Let’s bring the same objective to everybody. Get everybody in the same room and start to build these best practices and these strategies rather than go over here for this question and over there for this question, to have a centralized person that’s away from all that.”
The actionable point could indeed be internal. Setting up the right team. It’s a very people-based solution. The community does not seem short on solutions. It’s that we don’t necessarily need more different compartmentalized solutions.
See the video for the full discussion on attribution from Seth Adler, with Doug Healy and Sunny Zhu.
