Blog post

Why agile marketers need AI testing

Why agile marketers need AI testing
Written byNathaniel Rounds
Published8 Mar 2023

A conversation with David Edelman, Harvard Business School Senior Lecturer & former Fortune 50 CMO

Nathaniel Rounds, OfferFit Product Marketer: So David, you led the digital marketing practice of McKinsey & Company for a number of years.  What were some of the trends you saw in the work you did with your clients? 

David Edelman, Harvard Business School Senior Lecturer: I think the biggest trend – and frankly, it's been a trend for quite awhile – is the inexorable march towards personalization at scale. That leads to the hunt, especially by companies who have first party data, to leverage data to drive more personalization. That might be in the way they acquire customers, the way they engage customers, or where they cross-sell. Companies are trying to figure out how to crack that code – having the data, the content, the experience, and being able to deliver it. So I think the number one trend is personalization at scale.

A second trend is the recognition that in order to get there, you've got to do a lot of testing and learning. And so one of the big things we did was help companies set up what we call agile marketing capabilities. That's become a bit of an overused term, but it's real. Instead of just simply saying, “Here's our campaign and we're going to go out and do it,” agile marketing says “No, we're going to set a target for where we want to go. But the pathway to get there is going to be chosen based on what we learn along the way, and what we test, and the innovations we come up with.” 

So you have to constantly test and learn. Put together cross functional teams that can bring analytics, marketing strategy, operations, creative, sometimes legal together to come up with ideas, get them out to market, get those ideas back, and then constantly improve. That was a significant part of the work we were doing – helping companies pivot towards agile. And when I became Chief Marketing Officer at Aetna, one of my top priorities was to change the marketing operating model to be dramatically more along the lines of agile marketing. 

Nathaniel Rounds: So as a marketing leader, you were copying the “waterfall to agile” transformation that happened a few decades before in software development?

David Edelman: I’m not unique in borrowing the term “agile” from software – that has been talked about a lot in marketing circles. There's actually a book, The Six Disciplines of Agile Marketing. It's just not easy to do.

Nathaniel Rounds: What do you think makes it so hard? Why isn't everybody agile, if it's such a good idea? 

Davide Edelman: Because change is hard. 

I hate to use the word “paradigm,” but it’s just a different paradigm of thinking about marketing operations. So much of marketing has focused on the content and the media buy and not enough on the operational aspects of simply trying to do that faster. There's been a shift in the companies that are doing agile, where marketing operations is getting more and more important. At Aetna I actually set up a new position on my team at the VP level, reporting directly to me, leading marketing operations.

Organizations are set up hierarchically for the most part – they are functionally aligned, even within marketing itself. There's an analytics team, the strategy folks, the creative folks. The way things get done, to borrow the term from software design, is waterfall – things get handed off down the line. There's all kinds of iteration that needs to happen before you get something out the door. That’s slow, but that is the way a lot of organizations are set up. 

If you're going to move towards agile pods, with clusters of people from different functions working together, you're changing the unit of operations. Functional leaders don't actually manage the day-to-day work as much. They are managing the capability development of their people, sharing, learning, and thinking about how to better use tools. Functional leaders are also much more sensitive to what's going on strategically and where to take their teams. But they're not managing the day-to-day output that's coming from these cross-functional teams. That's a pivot. That's a hard mental pivot to do. 

Nathaniel Rounds: What were the main things you had to do to make that pivot at Aetna? 

David Edelman: When I came into Aetna, one of the most important things we had to do was to sell Medicare health insurance. Medicare is direct-to-consumer insurance for people over 65, and 60 percent of Medicare sales happen within 60 days in the fall. So it's very concentrated. I started at Aetna at the beginning of September, and I knew Medicare was coming up. I said “I want to see the Medicare plans.” They showed me the planned campaign, and I asked, “Well how do we know that's going to work? Things have changed from last year. We have new products, the competitors have changed all kinds of things. We need a different model; we're going to move into agile.” 

The president of Aetna’s Medicare business was an accountant by training, and she was very comfortable with the traditional way marketing had operated – she had seen the campaigns, she knew the funding. So she was hesitant when I told her how I was planning to change things, especially that we would be taking more risks and constantly testing. I put her in touch with a couple of clients I had worked with, who helped her understand what was involved in doing agile and the benefits of doing so. It helped her get more comfortable, and she let me try it. 

Two weeks in, I invited her to come and sit with us. The way we worked – this was before COVID – is that we physically got everybody together in a room. We had stand-ups every day at 11:00 where we looked at the data from the day before. We made decisions about where to move money around – different channels, different geographies. We also made ideas for new tests we would want to run. We had her see this in the room. She loved it. 

In fact, she said to me, “You have to share this with the rest of the leadership team.” And I said, “Well, why don't we make a video of it?” She loved that idea, gave me some money to make a video, and I showed that video to the broader leadership team. It really opened their eyes to how marketing was changing. And so we ended up bringing agile to three other parts of the business, and it just expanded their whole sense of the art of the possible. 

But yes, it took some selling to get them to buy into that. 

Nathaniel Rounds: That's really interesting. It sounds like you were an articulate marketing leader who evangelized this way of doing business. 

David Edelman: Well, it helped that I had helped other companies do this before. In fact, when I was hired, it was one of the things the CEO and I talked about. It was one of the things that intrigued him. He knew that Aetna was going to have to dramatically change the way it did marketing – getting more consumer centric, more digital, more analytically driven. So I had air cover from him, although he didn't really understand the details. And it did help when I connected the president of the Medicare business with other people who had been through it. 

But it is different for a lot of companies that have just not really thought this through.  Change is hard, but there's more and more talk about it. People are starting to recognize the need. The question still is, how in my company do I make this pivot to agile, if people have never done it before? 

Nathaniel Rounds: What connection do you see between agile marketing and AI? 

David Edelman: With AI we have the instrumentation to be able to test rapidly. The technology allows us to put more tests out, learn, get gradually more granular, and move towards something that's more and more personalized. 

AI works best the more you feed it. The more you get the AI data, the more you get variance. So you’ve got to test more and more things in order to make the AI more powerful. Well, how do you do that? Software is definitely key, but you also have to have an operating model that will be comfortable with doing what will feed the AI. You need to make sure all the organization issues – all the approvals, everything you need to get something out the door – are going to happen fast enough to feed the pace you want to go. 

Nathaniel Rounds: So you saw the link between AI and agile. How did you come to be interested in OfferFit?

David Edelman: While I was at Aetna, one of the particular programs we needed to develop was called Behavior Change Marketing. As an insurance company, we wanted to get people to do healthier actions. If you're over 45, get your colonoscopy. If you’re on meds for a chronic condition, stay on your meds and renew your prescription. If you're a diabetic and you haven't had your levels checked in three months, go get your levels checked. 

There were about 50 of these different basic behaviors. If we could get people to do them, they would be healthier. They would save money and we would save money. We'd also, if they were over 65, get points towards our Medicare Stars rating. So there were a whole lot of things conspiring to make this an important engine, to use marketing to help healthcare. 

So I would know, for example, that I want to get Nathaniel to do a particular action. I knew who, I knew what I wanted you to do, but I had no idea how to get you to do it. So we had to test, and it’s a lot of different things to test: copy lines, incentives, the channel we use, the timing, the sequence, and so on. We used the agile pod format to set this up, but it was a massive multivariable testing exercise. 

We didn't have OfferFit, but we had some really damn good analytics folks. We created something on our own that was a bit more kludgy, but had AI capabilities that would help us manage things like minimal test cell sizes, so that we can figure out the most cost-effective way to test multiple variables at a time. It was very successful, but it was kludge. It required a lot of data science capability that wasn't easy to hire. 

After the CVS merger, I left. And within a few weeks of when I left Aetna, one of the early investors in OfferFit, who was someone I had worked with before, reached out to me. He had heard what I had been doing at Aetna because I had talked about it in blogs and on LinkedIn and such. And he said, “You got to meet these guys.”

And so I met OfferFit's co-founders George and Victor, and saw what they were doing. Immediately the light bulb went off – “Wow, if I had this at Aetna, it would have helped us enormously.” I absolutely saw the value. I said, “I haven't seen anything else like this. I definitely get it. I recognize it's going to take operational changes to do, but in terms of the software and the algorithm capabilities, this makes total sense.” 

So I jumped on board as both an investor and an advisor to help evangelize what you guys are building. 

Experimentation unleashed


Ready to make the leap from A/B to AI?