Glenn: Hi, everyone. I’m Glenn Gow, Founder & Advisor of Crimson Marketing. Welcome to Moneyball for Marketing where we talk about the incredible changes happening in marketing organizations around big data and marketing technology. We feature marketing technology insights from the top marketers in the world. The reference to Moneyball is from the story of how the Oakland A’s baseball team were able to win and win and win because they figured out how to use data and technology to their advantage. If you’d like to learn about how to use big data and marketing technology and marketing to help you win visit us at CrimsonMarketing.com or email us at info@CrimsonMarketing.com. And now on to our podcast.
Today I am very pleased to welcome Jim Herbold, the Chief Revenue Officer at Infer. Jim is focused on company building, growth, and sales success for all the companies go to Market Initiatives*. He’s specifically responsible for defining sales and marketing strategies, building the team that executes for growth, and embedding customer success at the core of Infer’s positioning. So, how does Infer describe themselves? Infer is taking the predictive power of Google to help companies win more customers. So, Jim it’s a real pleasure to have you here today.
Jim: Glenn, nice to meet you as well.
Glenn: Jim, we spoke about this whole category “Predictive Analytics” and I know that a lot of people are interested in learning more about what’s happening in this world. I’d love to hear your perspective on how would you define predictive analytics as it relates to the marketer? And give us some examples, perhaps, of how it plays out.
Jim: Yeah, absolutely. Predictive analytics, at the highest level, it’s just doing and analyzing trends and patterns in data sets to make inferences about future outcomes, right? Well, actually, a machine learning thing and we’re going to be building algorithms on top of sales and marketing lead flows and we’re going to allow a marketer to stack those flows, the leads, by propensity to convert. Okay, so when you really get down to it, again, at a high plain we’ve seen a lot of predictive happening in the B to C world, companies like Google, companies like Amazon, eBay, Pandora, Uber. There’s countless examples that any marketer or any person would relate to as a consumer on the Internet.
What’s happening is the power of Google is now being applied and focused on sales and marketing, right? So, there’s definitely some really interesting use cases that do evolve for companies. We can help you filter and prioritize lead scenarios. We can help companies identify outbound targets. We can help marketers optimize campaigns. So, one great example, a company that we work with in the e-signature space, they had us score their entire nurture database. It was actually flow of free trials that have signed up and they get a very large volume of them. We used machine learning to score the accounts that were sitting in nurture and we actually helped them identify a cluster of leads that they wanted to send into sales for qualification for sales processing and we actually helped them find the largest deal that they closed last year. So, these are the things that the machines can do for us now and this is the type of technology that’s available to sales and marketing shops all over the B to B landscape.
Glenn: So, Jim, I’m most interested in a little more detail on that example you shared with us. So, how is it that you helped find the right kind of customers for any company to focus on? How does that work?
Jim: Sure. So, what we do is we work with the company and get access to the lead and opportunity database, usually in the CRM system. We will go in and train a predictive model on the historical data, right, from lead through conversion through the funnel to an outcome and that could be a closed one or a closed lost outcome. And what Infer does is it hones in on firmographic and demographic signals or data-points that the customer has acquired.
We also will go and append additional firmographic and demographic signals to the dataset and we’ll also go out and append upwards of potentially—we’ll look at thousands of signals that can be associated with the individual or with the company. So, really quickly on the side the founders of Infer, we’ve actually got Google and Bing, that sort of experience in our founders and it’s bringing in all of these signals together to build a machine learning their algorithm pointed at historical outcome and we can use those algorithms to then predict propensity or the likelihood of converting in future lead flows.
Glenn: So, is it essentially looking at past behavior to find patterns?
Glenn: Or algorithms that tell us what future behavior is likely to look like?
Jim: That is correct. Yup.
Glenn: So, it seems conceptually easy to understand. My guess is it’s pretty hard to do well for any company. They’re going to need access to a lot of data and they’re going to need to process that data quickly especially if we’re trying to do this in real time. So, how does it work when a lead comes in as it relates to connecting the marketing activity to sales?
Jim: Sure. Well, the interesting thing with a predictive solution is that we can actually provide a score on an account or an individual coming in at the moment that it has come in. So, as soon as they’re signing up on a “Contact Me” form or as soon as they’re responding to a particular campaign or coming in from AdWords. We’re able to assign a predictive score to that record immediately and marketers will use that type of score to basically prioritize what’s going to happen next. Is it going to be sent in directly to a sales rep? Is it going to be sent into a group of lead qualifiers? Does the score indicate that this is cross-back that you might want to send into nurture. So it allows that marketer to quickly automate the routing of a particular incoming inbound lead to get it in the right hands or apply the right type of regime as quickly as possible.
So there is a time to process and benefit of using a score and certainly if you look at scores over time, right? The trend changes of scores over time from particular campaigns or from particular sources of leads, that’s terrific insight that a marketer can gather to see what’s happening with effectiveness of campaigns over time. So, not only do you get the benefit of the processing potentially more quickly, putting things into the sales side of the house, you can definitely just gauge quality over time of flows coming into your business and that helps marketers ask the right questions about what they need to double down on, what they might need to do less of, et cetera.
Glenn: So, let’s talk a little bit about that how far can predictive go, Jim, in helping me as a marketer understand the likely results I might get from a campaign? Does that capability exist yet?
Jim: Well, something for example we do ourselves, when we’re testing, for example, new AdWords combinations we are looking at the relative score of the flows of leads as we do change and test in a system like AdWords, so we do that today. There are other companies that we work with that do that today and I think that’s certainly one of the use cases. I wouldn’t call it the dominant use case but I would certainly call that one of the use cases that does exist with predictive.
Glenn: All right, another use case or example that I’ve heard about is when you sit down to design and integrate a marketing campaign by using predictive I can be smarter about what my campaign looks like so I can improve the results of that. Tell us a little bit about how that works.
Jim: Well, I think the smarter part, it certainly takes some of the hunches out scoring. What marketing automation systems for many years have had scoring tools that they make available to marketers, so you can give 25 points if it comes with a particular VP title. It could give plus five points if there’s consumption of a download, if a white paper, but these are static scoring mechanisms and they are at the end of the day based on human based hunches and they don’t append lots of the external firmographic, demographic, and other types of signals that machine learning can bring in. So, we provide a deeper level of statistical significance and precision and we do take the hunch work out of some of these exercises.
Glenn: You mentioned firmographic and demographic, can I assume that you’re also pulling in behavioral data?
Jim: Sure. The answer is yes, you can and it does depend on what the marketer or the sales shop is trying to achieve. So, one of the types of models, the primary type of model that we build we would call “Fit Scoring.” This is a score that is applied at the moment in time when a lead does come into the system and we use these firmographic and demographic, and again, the appended signals that the network can bring in, and that’s a moment in time score. If something changes in the database, and say a rep changes the title from “Empty” to “VP of Marketing,” and that actually has turned out to be a signal that is important to them, the model of the score can then go up, but that’s all called fit scoring.
Then there is behavioral scoring, think of it as, again, an example would be consumption of a market and asset, attended a webinar. There are some nifty things happening with company intent data that firms like our and players in this predictive space are able to bring in. So, for example, you have a lead of a particular person or contact in your nurture database. There’s ways for us to get information about other people from the same company, not necessarily identified but anonymous, bringing in that company level, sniffing around, similar types of assets with similar types of keywords, that might be an indicator to pop a score up and send it over to marketing.
So, there is a play between what we would call a fit model and a behavioral model. We do have companies that use those two together. A lot of marketers that might be listening to this podcast would certainly be familiar with serious decisions—
Jim: —and got a relatively well built out frame for how you want to think about the fit score and the behavioral score, and then thinking about automation mechanisms that a company would use to make the right decisions about how to optimize and take best advantage of the moment in time when a lead is hot to be spoken to.
Glenn: So, I’d love to step back and have you tell us an example of what companies are doing in the real world and how they’re taking advantage of predictive analytics.
Jim: Sure, sure, sure. Well, there’s an easy example I can speak to. I was the first customer of this company Infer when I worked at Box and when I was running sales at Box and I had the luxurious challenge of dealing with very large lead flows. We had a freemium aspect to the business. We also had a very vibrant free trial aspect to allowing people to get into our service pretty quickly. So, very large flows and leads, we’re talking 10s of thousands and I could never afford to apply a lead qualifier to plow through all of those leads systematically over time. I needed a way to find the proverbial needle in a haystack and I started working with Infer and we were able to build a predictive model together and use that to score leads coming in.
And immediately we more than tripled our conversion rate because before we just had reps going through long lead lists and trying to pick out the ones that had a company domain or maybe we could tell had an activity in the account. Infer automated all of that and just automatically shot leads into qualifier hand without them having to do any research on the front end. We immediately tripled the revenue that we were getting from these particular sources of leads, freemium and web trial abandoners. I’ll give you another example.
Glenn: Great, great.
Jim: A company in the customer success space, they actually are a good example of a company that has married a fit score with a behavioral score and theirs is a particular type of behavioral score which is, again, application usage, right? So, we actually came into create that fit scoring. They actually used another vendor and we actually worked with that vendor to append the scores together. We were able to find that we were able to identify 95 percent of the opportunities that would close in their top cluster of leads and it allowed them to get increased close rates faster, but it’s a great example of combining fit and behavioral model usage.
Glenn: Good, so this sounds like a very powerful technology for marketers and I know there are a lot of companies that have adopted predictive analytics but there are many that haven’t yet. So, tell us a little bit about what’s prevented the adoption of predictive.
Jim: Sure, there is a number of factors that have prevented the adoption of predictive. I think the first, just to acknowledge the category is really only about three-years-old. Predictive has been happening in the B to C world longer than the B to B world, that has now changed, but it’s still, relatively speaking, about three-years-old. So, although it is new I would certainly argue that awareness has rapidly increased. The category is pretty well known and acknowledged these days.
The second factor in play, I think, companies historically never really had the ability have an in-house data scientist who could work on these types of challenges for sales and marketing.
Glenn: Oh, right, right. Great shortage of data scientists in marketing.
Jim: No doubt about it. So, predictive as a service, again, relatively new, new answer to the challenge. I think the third issue about what’s prevented this is a bit more complicated and it ties to some of the politics that you’ll see between sales and marketing in some companies, and one might actually say in many companies. So, when you think about a predictive model which is a data driven sort of statistically truthful view on what’s happening with the flow of leads coming in, there are definitely marketers in the world who don’t want the quality of their lead flows exposed to other folks. They really would rather just have sales process everything and keep the truth under the covers so-to-speak. Flip side of the same coin, the sales side, when a stack ranked prioritized flow of leads is coming into a sales engine, it really puts the burden on sales to apply the right levels of resource to the right clusters of leads.
Glenn: Interesting. So, in both cases machine learning is either going to make us better marketers, better sales people or it’s going to expose where we’re not quite as good as we thought we were.
Jim: No question about it and, again, it puts the burden on sales to really apply the right levels of activity to the right leads or else you can’t get the value out. And the other thing I would say, this interesting, by no means would a stack ranked predicted flow of leads, should that imply to any sales or marketer that’s listening to this who are out in the world that this would allow a company to avoid great sales qualification, to avoid sending classes of leads into really well thought out executed nurture campaigns. There is nothing about the fact that, hey, just because it score highly it’s absolutely going to close. You still have to do all the hard work to sell the value of your company’s service or product or whatever it is.
Glenn: Of course, right.
Jim: It’s not that, let’s just put it that way. So, and I think just the fourth issue, just back to the original question about prevention, what’s prevented adoption. You mentioned this a bit earlier and it’s got to do with the volume and the history of a lead flow into a company. Like, we have to have a certain number of outcomes that we can train against with the models and if we don’t have enough of those outcomes the concept of statistical significance gets eroded and companies need to have roughly speaking certainly well into the hundreds of leads a month and have to have been doing that for some period of time. But, yeah, I think that you’re also going to see some innovation in the predictive space that’s going to allow predictive to come down to S&B companies that may not have the very, very large flows or the historical flows of outcomes.
Glenn: Oh, that would be very interesting!
Jim: That would be very interesting, it’s currently something I’m thinking about.
Glenn: And that would greatly expand the market opportunity for predictive analytic companies and for companies that can take advantage of it because not everybody has 100,000 leads a month.
Jim: It’s been a mid-market and above play thus far, but I think all of us are kind of jockeying for how we get at that volume, right? And that’s a great place to be doing business.
Glenn: Well, Jim, this had been very helpful. I really appreciate you sharing your insights with us.
Glenn: And I’ve learned about predictive marketing and I hope our listeners have learned a lot as well.
Jim: As do I. Thanks very much for having me on your podcast.
Glenn: It’s a pleasure.
Jim: Right on.
Glenn: Talk to you soon.
Jim: Take care, bye.
Glenn: If you like this podcast please subscribe and rate us on iTunes and tell your friends about us. You can also go to our website, CrimsonMarketing.com, and sign up for our free monthly newsletter featuring the very best of our marketing insights, featured Moneyball for Marketing podcasts, and one of our favorite features called, “Bad Marketing,” or email me at info@CrimsonMarketing.com. Thanks for listening to Moneyball for Marketing from Crimson Marketing. Have a great week and let us know if we can help you in any way.