Predictive Marketing Analytics Will Change How You Approach Online Marketing: Interview with Dominique Levin

 
Interviewer: Today I’m very pleased to welcome Dominique Levin, V.P. of Marketing of AgilOne. Dominique runs marketing of AgilOne and she joined recently from Tatango, a leader in customer success management software; and was previously CMO at Fundly which helps non-profits with social media fund raising; and she was CEO and CMO at LogLogic, a big data analytics company which was sold to TIBCO. So who is AgilOne? One of the things I like about AgilOne is their mission statement. The mission statement is to bring revolutionary predictive marketing to every day marketers. AgilOne wants to restore the personal relationships companies had with their customers in the days of the corner butcher store. So, that’s a great mission statement Dominique. You know we were talking earlier about data driven marketers, and do they really need to become data scientists? Can you talk a little bit about that for us?

Levin: Absolutely, thank you for having me first of all.

Interviewer: My pleasure.

Levin: So, yeah, we believe that there is a true revolution going on in marketing and that marketers need to be data driven, actually have an opportunity to reconnect with their customers and deliver happiness to each and every customer by using information that they have on each and every customer. And this is made possible of course, by the incredible revolution and evolution and Cloud computing and machine learning and compute power being available, the big data revolution that we all hear about. And so it really allows customers to tailor their content, their offers, their products to each and every customer. Now of course, we believe this is a big change for many marketers, especially in the consumer marketing arena where marketing has been very creative and brand driven.

Interviewer: Right.

Levin: We don’t say throw that out, but we believe that there is some sort of perfect marriage between creative for human intelligence and machine learning that can make things much better. And we’ve seen some of our customers be very successful with that. One of my favorite examples and a company to check out is a company called Moosejaw. They’re an outdoor retailer.

Interviewer: Moosejaw?

Levin: Moosejaw yes, check it out. They’re out of Detroit, Michigan of all places, kind of an oddball place for an outdoor retailer. But, there were known for their incredible creative marketing. Their tag line is love the madness. And if you look at some of their YouTube videos they’re definitely mad, they’re very funny, very creative, but they have successfully used that to kind of inspire their creative and to validate what works and what doesn’t work, and to really grow their business incredibly. They’ve had incredible success in their category with this kind of blend of creative and data driven marketing.

Interviewer: Well talk a little bit about how they measure what works and what doesn’t work, that’s what we all want to know as marketers.

Levin: Right, well ultimately it’s about of course driving top line revenue, driving profitability. But they really look at this on a customer by customer basis and ultimately a measure of success is how many customers can you get to come back right, and develop a long-term profitable relationship with your brand.

Interviewer: Right.

Levin: In retail specifically, this is notoriously bad. I mean, a retailer who can get 30% of buyers to come back for a second time is doing well. There are people who have only 7-8% of their customers ever making a second purchase. So, for retailers like Moosejaw the challenge was how do we get these one-time buyers to convert in two-time buyers and come back multiple times.

Interviewer: Dominique are those statistics primarily for online retailers?

Levin: They are relevant both for online and offline. Of course, the offline world it’s much more difficult to measure. I mean, there is a resurgence of loyalty programs to try and measure of course purchases made and identify consumers and link their behavior and also link online and offline behaviors; I think that’s finally starting to happen of digital coming into the stores and of information about customers and purchases made in stores inspiring online campaigns and vice versa.

Interviewer: Right, okay.

Levin: But, yeah this is a pretty generic trend, and ultimately, the more repeat purchase and loyalty business you can drive and that’s a notoriously difficult metric to move. Another one of our customers is Mavi Jeans, they sell a lot of jeans to celebrities and such and they had a pretty average, like one order per customer per year type of average. But, using data to do these very personalized and targeted campaigns they were able to move that to over two orders per year. And so, you can see how that might have a dramatic impact on both revenues and profitability, as it is much more profitable to sell more to existing customers than it is to continuously have to spend money to acquire new customers.

Interviewer: So, can you talk about how specific we can get as marketers in marketing to the individual. Is it truly one-on-one?

Levin: So, yes there is one-on-one marketing possible thanks to big data, and the ability to track individual user behavior and then target. However, I would also caution marketers to get too hung up on one-on-one marketing. Just like people, yes we like to be treated as unique individuals, but we are also very much part of communities and like to identify ourselves with communities. And so, marketing of course is no different where individual one-to-one marketing is possible, but typically we recommend not to have that replace segmentation of customers and identifying customer personas or customer communities that you might have. And again, data science can help there too. We’ve helped many customers’ undercover personals that they didn’t know that they had. And then after you’ve uncovered this persona you can maybe throw in some personalized content. So, the way this might come together is you might discover that you have a group of soccer moms as part of your customer base. These are people that buy active wear and specifically a lot of active wear for their children. You might develop content specifically for that audience, either to acquire that audience to engage that audience on an ongoing basis. But then, when you send an individual email, let’s say you send an email about how to pick the right sport for your kid or whatever it is, you might want to include there a very personalized recommendation for that specific person based on products that they’ve bought in the past.

Interviewer: Right.

Levin: So, you’re trying to both segment address as a community, and then address as an individual, and those two often work together hand-in-hand.

Interviewer: Okay, so it sounds like you can get a richer experience for the customer by acknowledging that they’re both and individual and part of various communities.

Levin: Right and I do believe that customer expectations have changed dramatically too. I think we really expect brands to know who we are. So, we encourage marketers to really do collect customers 360 profile that they can use to better address the needs and wants of these customers. But, I think we’ve all these experiences, not to pick on my cell phone company, but where I call my cell phone company and I complain maybe about some item on my statement and I find a very terse and unfriendly phone representative on the other end of the line. Whereas, I have been a loyal customer of this phone company for 10 years and I have spent many thousands of dollars. And the fact that the person on the other end of the line does not recognize that and does not treat me with the respect that I believe as a VIP customer I deserve actually has caused me to switch phone companies in the past.

Interviewer: Um hm.

Levin: So I think we really expect these days brands to know these things about us and to really recognize us for who we are. So, I think that might be a little bit different when you haven’t done business with a brand before and it gets a little more sensitive I think when you track anonymous people around the web and so forth, but when it comes to companies that I have chosen to do business with I think we really see that shift in consumer expectations on personalized treatment.

Interviewer: Great. I want to go back to something you said a few minutes ago about helping identify new personas. Now we do a lot of persona development work and I’m really curious as to how data can help uncover new personas. Please tell us something about that.

Levin: Yes absolutely. So yeah it turns out that machines and algorithms are actually a lot better at pattern recognition than people ever could be. And so, what you really can do these days is you can throw that customer’s 360 profile data we talked about, which consists of all demographic data you might have about your customers, maybe even enhanced by things like U.S. Census data, and then combine that with any and all behavioral data that you might have collected, so these are things like email opens and clicks; websites; browse patterns; past purchases etcetera, and thing about if you throw that all in the top of this magic box and out on the other end come personals. And what algorithms can do is they can take into account many more factors than we as human beings ever could. So, we find for example that one customer found that they have different types of VIP customers that behaved very differently. They had one group of customers that were big spenders, yes, but the pattern of their spending was one where they would come very frequently and make small purchases; and they would buy certain types of products and they would buy certain types of products together; and they were mainly females. I guess they like to come and shop very frequently for fun. They had another group of VIPs that were spending a lot all at once an on certain types of products; and they, not to stereotype, but in this particular case were mainly men. I guess liked bulk up and get shopping over with all at once.

Interviewer: That’s right.

Levin: But anyway, these were interesting kind of personas. So, we see that again, that I can take into account. There’s a lot of things here, it could be demographics, again, behavior, types of products, frequency, things that people buy together, but ultimately yeah, in an outcome kind of what are the five, six groups of customers that math has been able to identify are most relevant for you. Are most distinguished or distinguishable from each other, and once you see that then the juices kind of start flowing like wow, I did not know I had this group of customers that only ever buy sweaters from me, but only when they’re on discount in the spring. Okay, when I know that I can now target and develop campaigns specifically to either attract or target this group of people. But, there’s no point in trying to bombard them with other messages that I might have and try and sell them my bicycles or my canoes or whatever else I’m selling. It might only cause people to unsubscribe from my mailing list, in which case they will no longer get my spring sweater sale announcements and so that would be bad. So again, really recognizing these customers, which goes back to something I said earlier, it gets the creative juices flowing and once you see these personas, like wow okay, now I know I have those people I know what I need to do with my creative to engage these groups.

Interviewer: So, question for you about the definition of predictive marketing. The example you gave of advertising or promoting sweaters in the spring to certain individuals, would you label that as predictive marketing in that based on historical information we have a pretty good sense of how they’re going to respond to that versus a campaign for bicycles?

Levin: Yeah, absolutely. So predictive, of course is it a direct reference to predictive analytics, and there are three mathematical models that are most useful to marketers we find in campaigns are clustering algorithms. So, those are the algorithms that I described that could – you let them loose and they find patterns and personas. So, these can be behavioral clustering, or brand based, or product based clustering algorithms. There is propensity models, so most specifically propensity to buy, so the likelihood of a customer to buy at any given point and time is very useful because based on that information you can, for example, identify customers that are at risk of leaving you. And if they’re at risk anyway, I mean, much easier again to turn them at that point than to wait for them to start shopping in a different store or start to engage with a different brand, or become a customer of a different company.

Interviewer: Right.

Levin: And so these propensity to buy models allow you to identify things like at risk customers then do preventative campaigns maybe to reengage those customers. And you do that with personalized content or offers. And the third model with clustering, likelihood to buy is the kind of lifetime value analysis. And so, that can be done both retroactively, so be very accurate about the actual profitability of a customer taking into accounts things like returns that people might have done and at our customer service cost. But, we can also start doing that predictive lifetime value. So, from the moment that I buy that first Gucci handbag, Gucci can actually have a pretty good sense of how many more of these handbags I will buy in the future even if I don’t know myself yet how much of my paycheck I’m going to spend with Gucci. But yeah, they can do that based on the standing of course of who I am, so some demographics and then comparing my behavior right. So, I haven’t just made that purchase, I’ve probably also browsed the website already a number of times, maybe engaged with some of their email campaigns, and now they can compare my behavior to other people like me and based on that predict my future lifetime value, which can be very valuable, especially when you’re trying to do things like allocating marketing spending across different channels, or maybe even specific key words. So, now you can start to decide which of my channels are producing, not just the most customers, but the most profitable customers in the long run.

Interviewer: What would be really interesting, and I don’t know who’s doing this now, you can tell me if you’re aware of companies doing this, if Gucci then sells that data to other companies, whether they be retailers or other fashion brands, because they now know something about that individual and their likelihood to purchase something else. Do you see that happening?

Levin: Yes and no. I mean there is – I’m not aware of people doing that. We definitely don’t do that at this point, because we very much, again, respect the privacy of the individuals. I think there are some privacy issues here. I think you’re again we’re walking this fine line of ultimately it’s about delivering value to the individual, and first and foremost it is for the brands to deliver a better experience to individual customers. And I think most customers would be okay with that. Again, they’ve chosen to the business with their brand and there is almost this expectation that technology is being used to serve their needs better.

Interviewer: Right.

Levin: I think once you start selling that information obviously it gets a little bit more tricky, but yeah, there’s certainly, I know people talking and thinking about things like that. I think for now we’ve take a position that we leave it to the brands to serve their customers directly. But, yeah I think the consumer perception will likely evolve and some of what’s okay with their data, even the legislation is evolving in this area in different countries. So, I think depending on that we’ll see more or less of that happening.

Interviewer: Alright interesting. Well good, so I have one last question for you. Step back for a moment and help us think about what the next one or two years look like. What does the future look like in predictive marketing?

Levin: Yeah, so I think what’s most exciting to us, you read kind of our mission statement, is that really this predictive marketing will be the norm. I mean there is – we did a survey last year that said that over 60% of – and this was in retail e-commerce specifically, over 60% of marketers are still doing one size fits all marketing, meaning that across their channels they’re having one message to all customers. And I think even though some want to use data science methodologies have been around for a long time they just have not been accessible to every day marketers.

Interviewer: Um hm.

Levin: You know you had to have big in-house data science teams. You needed to have large, million dollar data warehouse infrastructure to be able to perform some of this analysis, and not just perform it, but actually implement it in more or less real time on a customer by customer basis, so to make it actionable. And so, this is really what’s changing thanks to Cloud technology and thanks to the innovation of companies like AgilOne where we really put all that in the Cloud. Think of us as a data scientist in the Cloud. And really where predictive targeting becomes not just a better way to target customers, but frankly the easier way right, I mean think of predictive as an automatic way again to segment your customers. You don’t have to manually slice and dice, but you don’t have to guess what are the 10 behaviors that might drive likelihood to buy. There is just a likelihood to buy rank order of customers highest to lowest likelihood to buy and you can use that in your campaigns. And so really I think this is going very – this is all happening very fast and so first and foremost I predict that the next two-three years or so nobody will really practice on size fits all marketing anymore. And increasingly we’ll finally see happen the true integration of online and offline experiences and more kind of real time customer event driven marketing. These are not even necessarily campaigns anymore right, they’re just personalized treatments that happen in real time across channels thanks to the evolution of some of this data science and making data science truly easy and accessible to all.

Interviewer: Yeah, that’s really a good point about it’s not really about campaigns so much as an extended long-term relationship you have with that brand. So, I can definitely see the value in that for consumers.

Levin: Yeah, and it’s no longer a push, aside from it not being one size fits all, it’s no longer happening at the schedule of the marketer right.

Interviewer: Right, right.

Levin: It’s consumer actions that trigger a reaction and thus an on-going dialog.

Interviewer: Fantastic, well Dominique thank you so much. This has been very insightful. I appreciate your time.

Levin: Thank you very much.