Marketing In Five Dimensions
Epsilon is taking the next step in marketing analytics in a constantly changing business environment.
Topics
Competing With Data & Analytics
Andy Frawley is the newly minted CEO of the self-described “global marketing company” Epsilon, a post he earned in December 2014 after 5 years with the company. But don’t let his short time at the helm fool you — he’s an industry veteran with 25 years of operating experience, including 20 years at the senior management level, and expertise spanning digital marketing, email marketing, CRM, database marketing and customer value management.
The company he leads has a unique mission. In his words, Epsilon’s goal is to “fuse data, analytics, creativity and content together to connect brands with people.” In an era of rapidly changing technologies, that can be a significant challenge. The sheer volume of data, the ever-expanding multiplicity of platforms and the perennial problem of finding data scientists are but a few. But in many cases, the challenges are perpetual, as industries and companies struggle to adapt to a rapidly changing data universe.
Andy shared his perspective on these challenges — and where his industry may be headed — in a conversation with MIT Sloan Management Review’s guest editor Sam Ransbotham.
What, in your own words, does Epsilon do?
What we do is bring a series of different marketing practices together, everything from strategy and analytics to creative design and execution, as well as enabling technology, so that every interaction the brand has with a consumer can be fueled by as much information and personalized content as possible. We also do a lot of work trying to measure which of strategies are creating the desired business outcomes.
Can you talk about the different approaches to understand data across all devices, and where you are now?
It’s an interesting time for that, in our view. The amount of data that the digital channels are manufacturing, whether it’s the social feed or behavioral information on the web, all the way through to things like telematics — is more data than most brands or companies have ever seen before. It dwarfs the size of the traditional first-party data warehouse projects.
The challenge is, you’ve got a mix of authenticated data, and anonymous data and it’s difficult to marry the two. We actually just completed a large acquisition in late 2014 that helps fuel our ability to link both together and to be able to use that information to understand people — not just when they’re transacting or buying something, but when they’re in the consideration cycle – and engage with them accordingly.
Big data needs to move from an analytical insight to something that drives actions. From our point of view, and based on the needs of our clients, big data needs to be more closely linked across the channels that are actually driving communications, whether that’s the web or mobile or call centers.
Or Fitbit or whatever the next wearable is.
Right. The telematics, the wearable stuff, and all the next-generation Internet of Things advancements are amazing. I was with a company the other day that has a marketplace for home services. Smart homes and intelligent thermostats are going to completely change their business, because the house is going to tell people when they need something serviced, not the other way around.
A lot of companies’ marketing execution framework was built around 10 campaigns a month or something like that. And the framework we talk about with clients now is that you need to have thousands of campaigns running all the time. That changes the framework.
Are you talking more about something like real-time retargeting campaigns?
It could be. It could be linking into other operational systems. For example, at this time of year we do a lot of work retargeting people in stores through e-receipt programs. We have an offering where we look at e-mail, which is still a very productive marketing channel for most of our clients. In fact, it’s the most productive channel on an ROI basis.
But the behavior of how people look at e-mails has changed radically. Now, about 60% of e-mails are opened on a smart phone, yet not many people actually convert, or make a purchase, on a smart phone. So we have the ability to recognize consumer behavior, see when they open e-mails, what device they open e-mails on, and then retarget them accordingly.
So if I look at my e-mails on a smart phone at six in the morning but don’t click through or buy, and then I’m on my desktop at nine in the morning where I’m more likely to convert, the brand should deliver the e-mail to the desktop at nine in the morning if it’s a retargeting e-mail that somebody has opened but hasn’t acted on — those sort of things are required in today’s real-time world.
And then obviously the marketing aspect connects quite well there.
We did a project for a deregulated Texas utility, which has done a lot of work with smart thermostats. Utilities are tough because they all just try to compete on price. There’s no real differentiation in the product or service, per se.
We basically told them that the right advertising campaign was ‘buy less of what we sell,’ which is sort of a radical idea for most advertising campaigns. And then we built up a whole series of mobile tools to let people manage their energy consumption in their house, and that was the actual brand idea. We built a whole series of campaigns around it, and it went over incredibly well with consumers.
Let me push back on that a little bit — “buy less of what we sell,” that’s tough on the ROI there.
Well, not when you increase the efficiency of acquisition and reduce churn. They’re using less energy, but they have more customers, and the customers they have stay longer.
How far along are you in that process of being able to really link all the activity for a customer to a purchase decision?
We’re pretty far along. We do it for the large brands. There are a series of techniques to do it, and obviously a fair amount of what has been the common practices like using cookie matching, which has a level of imperfection to it. There are other ways to drive that that also fill in the gaps where cookies don’t really work, particularly on a lot of the mobile devices.
So the company we just bought, Conversant, which is a big media company, has built a common ID that goes across devices, like mobile, tablet and desktop. They have something like 190 million people identified — a pretty decent swath of the population. We’ll be integrating that from an Epsilon standpoint with our first-party matching technology that tends to be terrestrial, based on name, address, phone number, e-mail address or something like that. That is in the works right now.
How much do you get into analyzing people versus providing the data to companies and a more raw format, to providing more of the answers and answering business level questions?
Deriving insights from data is effectively what we do. The legacy Epsilon business was more around data management, but today, we have a large group of quants who do deep analytics, PhD-level people who are data scientists. We also do a lot of data discovery work.
Again, we’re typically driving toward some actual communications. Our lens on all this data is that it’s not that useful unless you can use it to improve the customer experience and drive engagement between a consumer and a brand. And we have thousands of people who do that sort of work.
We’ve got lots of people looking for those same people that you’re looking for to do those jobs. How are you competing in this market, or acquiring or developing the talent?
It’s a great question, and you may have some perspectives that we would find helpful, too.
We have a group that historically has been a quant group — PhDs building predictive models. We still do a lot of that work, but as we’ve migrated into this big data world, those job skills change. And in some ways it’s good, because you don’t need a PhD to do everything.
As we’ve been looking at this data scientist role, we have found it very hard to hire those people. So our approach has been much more to develop them ourselves. We do a lot of college recruiting. Not exclusively, but we have a lot of good experience hiring young people right out of school. I had 150 college kids this year, and one of the tracks that we put them on, if they have the right sort of aptitude, is that data scientist track.
What do you particularly think is important when you’re looking for taking people from college, who may not have the actual skills? What is it you’re looking for when you’re looking for these students?
They’re not all computer scientists. Yes, we do hire computer science majors, but we also look for people who have some math or CS background, but haven’t chosen to major in it.
The other track is for people who are on more the pure marketing side, and they tend to start in junior account management roles.
So you’ve gotten these people. You found them hard to hire. You pulled them in from college and you’ve given them a bunch of data skills. What stops them from disappearing? Taking these newfound skills and going elsewhere?
In reality, nothing. But it’s a challenge. We actually have pretty good retention — the attrition of our college hires is actually lower than our general population. If we hire a kid out of college, they get a 6-month review, and they get much larger salary increases than the rest of the group. You know, we recognize that somebody with 2 years’ experience is worth way more than somebody with no experience, so we have to sort of tweak their salary. You can’t give them a 3% or 4% raise or you will lose them all.
We have the luxury of being big enough that we can move people around. They can move around to different clients if they find that maybe they aren’t as quantitative as they thought they were. Or they can move more into something that might be more creative, or more of an account role.
If they’re interested in what we do, we can give them a very well-rounded background. And we promote them quickly. A lot of them are very talented. I was talking to some recent grad who is 25. He’s worked for us for 4 years and he’s running a $20 million project right now. It’s a lot of responsibility.
What’s changed, to where people don’t need a PhD for this?
A lot more can be done through data visualization. And even some of the tools that bring predictive approaches can be used by a less technical person. When we get in the hardcore machine learning and stuff like that, those tend to be people with more advanced degrees.
The other dynamic that plays into this is that a lot of these analytics need to be in real time. Before, somebody would say, “I want to build a model to predict who is going to open a mortgage,” and we’d spend 6 weeks building a model and another 6 weeks deploying it. This stuff has to be happening all the time in real time now, because that’s the pace at which consumers are acting now. And that drives a very different cadence from a marketing standpoint.
The companies you’re working with — how far along are they at moving, say, from the non-analytical approaches to customer segmentation and understanding customers, to your more avant-garde analytical approaches?
It’s all over the board, as you’d expect. There are some industries where they’ve done quite a bit of it, and others with more laggards. Almost all of our clients use fairly sophisticated techniques to do segmentation analysis. Some sort of predictive approach to understanding which segments people are in, and whether they’re behavioral segments or attitudinal segments. It varies across category, and across company.
What’s more on the leading edge is to integrate all the broader data that is often used to trigger something. In terms of segmentation analysis, there aren’t many companies doing that.
I just wrote a book [Igniting Customer Connections] that has some of this in it. For example, we found in the research we did in this book is that emotional segmentation, really trying to predict who is emotionally engaged with the brands, is a high driver of business outcomes. That’s interesting and relatively intuitive in some ways. But when you step back and say, well, we actually now have data from the social feed, from various points, that could give us a perspective on who we could meet more emotionally and engage with as a brand, it becomes really powerful.
What is the hesitancy there? What keeps people as laggards, I guess is what I’m trying to get at?
One is just complexity. To do this requires different tools and techniques. There’s the use of the data, and there are also different levels of adoption of how much people are doing in things like social, which is a relationship there.
And then I think there’s some categories, financial services probably being the leader, that the privacy aspects of this have been uncomfortable to them. I think that’s actually changing now, and they’re leaning forward more. But just this area starts to bleed into the privacy discussion and what data can be used in what ways.
So you said you think that’s changing. What’s driving that change, or how is it changing?
I think it’s becoming an imperative. If you’re a big bank, you can’t just keep using direct mail. It’s too expensive. So they’re leaning back a little bit on some of this stuff. But given the way consumers in general, and particularly younger consumers, are behaving now, you have to be in this game if you’re going to have any sort of a real digital marketing approach.
If everybody feels like it’s becoming an imperative and some of the privacy-related issues may be fading a bit, what happens next? Where is this all going?
I think we’re at the point in the adoption curve where the technical capabilities exist. It may not be easy, but they’re maturing quickly. You have a lot of different choices. Now it’s more around the application of these big data technologies, and using them to not just observe things, but to act on them and create business outcomes. Again, in our lens that’s marketing, but obviously there are lots of other functional areas that use big data.
And I think identity is increasingly going to be a big part of it. The linking we talked about earlier, that’s going to be an interesting evolution of who is positioned to help companies with identity. We think we have an interesting posture because of the assets we have. But clearly, companies like Facebook are big into the identity business; Google, big in the identity business. Our view is that there’s going to be a market for a company like Epsilon, who can be the trusted third-party advisor and partner.
What makes you feel like Epsilon is better positioned in terms of trust than others?
We do it. We managed customer data for all 10 of the top 10 bank and for for 8 of the top 10 pharmaceutical companies. So because we do a lot of first-party data management, that’s kind of what our business was originally. It still is a big part of it, so we sort of have the chops to do that. It’s not easy; it’s not simple; and it’s not cheap in today’s world, but we have that capability and expertise already.