AI has changed how Kellanova thinks about its customers

At the Groceryshop conference in Las Vegas, Modern Retail sat down with Louise Cotterill, global senior director of insights and intelligence for Kellanova, who sits on the advanced analytics team. Today, much of the team’s work centers around how to use AI, with a focus on what kind of advantages it can give the company versus its current tools or processes.

For example, Cotterill’s team has clean room technology in the U.K. to analyze purchase behavior, attitudinal research and demographic data, and then create personalized advertising across Meta platforms and Pinterest. Data clean rooms are safe spaces where insights gleaned from the walled gardens are commingled with first-party data from advertisers for measurement and attribution, as Digiday explains.

As detailed in Ad Age, Kellanova’s use of clean rooms led to a shift in Special K’s marketing to focus on why it is worth more than other, less expensive brands by highlighting its ingredients and brand heritage. That led to a 36% sales lift in premium segments and 9% growth with price-sensitive shoppers, per the outlet.

Cotterill discussed what AI means for how brands analyze information about their customers and how consumer-facing companies can do this effectively. This conversation has been edited for clarity and length.

How did the surge in popularity of ChatGPT and other AI tools change how you do your job?
“Our brand teams are fighting fires every single day, and so, the opportunity to look into the future is a luxury that’s afforded to us in the global team.

The ChatGPT explosion or implosion happened, and suddenly, what you saw was that rapid response time transformed. [The brand teams] were immediately saying, ‘Yes, I want it.’ And then, they’re proactively coming to you. You’re not going to them. At that point, we had more projects than we could possibly actually resource.

Then, instead of it being us going and selling to the regions and brands, it became a prioritization exercise of, ‘Who can we help first?’ and ‘How can we best support them and their organization?’”

What kind of projects immediately came to the forefront?
“What we wanted to be able to do is, instead of using historical data to say what did happen, start to use data to say what’s going to happen. That was the biggest shift for us, in terms of starting to get into more prediction models and starting to add prediction to say, ‘We think these are the people most likely to buy, and these are the audiences we should go after, as a result.’

[We then] watch that and then measure that in — as close as we can get for CPG — a closed-loop way, so you can really attribute the quality of that model and the quality of the creative media placement to the work you’ve done.”

What are some of the biggest revelations you’ve made about your business through AI?
“When you build audiences based on their behaviors, they tend to not be grouped by demographics. You have people across different age bands. They may be 18, they may be 65. You have people of different economic thresholds. You have different family circumstances and things like that. But what unites them is the way they’re responding and interacting with your brand in a predictable way.

And so, I think that’s been one of the really interesting pieces — that these fixed definitions we’ve had of who the consumer is can actually become much more dynamic and much more fluid as you’re building out these models. …

What’s been really interesting — as we look at, certainly, the role value is playing, and in markets where the cost-of-living crisis has hit — is the role that grandparents are playing in food choices. … They’re stepping in a lot more with childcare, but they’re also playing a bigger role in navigating what foods and snacks kids are having. And so, what perhaps had once been a nice, neat definition of family is actually a very broad definition now. … That was an insight that came about from mining our own first-party data and then also using different models in the clean room.”

What have you learned about the technology?
“AI won’t make bad strategy good or just make it more expensive. Fundamentally, if you haven’t got the foundations right, AI isn’t going to be the silver bullet. … There’s a lot of talk about the hype, but really, you still have to have the foundations as key.

The other piece is that AI is dressed up as a technological transformation, but actually, it’s a leadership revolution. … You won’t embed AI in your organization if there isn’t an openness to doing things differently and if there’s not an openness to acting on data that tells you something different from how you’ve always done it.

If you’re using data as decoration to support your gut feeling, AI is not the right solution, and it won’t take you anywhere. But if you’re really open to following where the data takes you, that’s when you’ll see the biggest success as you go through.”

How do you use retailers’ first-party data through their retail media networks as part of your data strategy?
“I don’t think there is a silver bullet media solution available. You’ve got to pick the right tool for the right job. Retail media plays a really important part in our job. It gives us much more connection to the consumer, but it’s only through the purview of one retailer, and most consumers are shopping two retailers a week — probably up to four or five a month.

And so, you need a more holistic picture, and that’s when you use other tools for those situations. And so, it’s going to be an important part of our repertoire, but it’s going to be alongside other tools that we’re using, like our own clean rooms and things like that.”

For brands just starting out with using AI or analyzing their business in this sort of way, what are some of the biggest things you’ve learned that could apply to them?
“The No. 1 thing is: Really have a clear definition of what you want to get out of it. Whenever we start with technology, it never goes anywhere — it’s, ‘I’ve got this cool tool,’ ‘I’ve got this amazing data set.’ That just doesn’t work. You have to have something that’s going to embed into your business processes. really making sure that you’ve got a clear definition of the problem, that you can find a tool to solve it. … Otherwise, it just becomes a very expensive PowerPoint at the end.

The other thing is to really keep in mind the commercial model, as well: How much better would this need to be in order for it to pay back? It won’t pay back in the first program that you run, probably, but you have to know that it will pay back at some point. … You have to have patience. … If you don’t have that in a smaller organization, then it’s probably not the right tool.

If you don’t have the resources in-house, managed services are often a better way to go to get you started, and then you can understand what works and go from there.”

What keeps you up at night with AI?
“Making sure that it’s used properly is, I think, the No. 1 thing — that it’s not misinterpreted or bias comes in, and that we have proper governance around it.

We have an AI committee that makes sure everything we do is checked and validated before we do anything, but I’m not sure every organization has that, and it’s absolutely got to sit at the heart of what we do.

So, yeah, misuse of AI and it potentially then getting a bad rap just because it’s been misused [is a concern]. It’s got really great potential, but like any tool, it’s what it does in the hands of the right person [that matters].”