Welcome to the 2019 Predictive Analytics World (PAW) Conference Series. Recorded live in Las Vegas, this series is bringing interviews straight to you from exhibitors and speakers at this year’s event. In this interview, host Jamin Brazil interviews James Taylor, CEO of Decision Management Solutions.
Find James Online:
My guest today is James Taylor. We are live at Predictive Analytics World, Marketing Analytics World. His business is Decision…
Got it. You are chairing a track today.
Yeah, I’m monitoring the business track. I’m kicking it off right after the keynotes, and then I’m monitoring it for the next couple of days.
That’s great. Obviously, you know who’s on the panel.
Yeah, there’s a panel; there’s a whole bunch of presenters. It tends to be the ones who are not so much talking about how to build a predictive analytic model so much as how to use a predictive analytics model: how to put it into production, how to get people to adopt it, what the challenges are, getting people to understand what a model does, and those non-modeling kinds of things.
Which actually the better that goes, the bigger the lever of the model, right? I mean…
Yeah, I’d go further. I would say that if you… There’s not difference between business value and analytic values. If your analytic, however good it is, isn’t actually being used, then it has no value.
That’s right, that’s right. The ROI on that’s really easy to calculate. Now, this is not your first show.
No, I’ve been, I think, to every Predictive Analytics World. I’ve been coming here a long time. I know Eric and the Rising Media folks well. And it’s been fun to watch it grow and watch it change from really a very niche kind of show where’s there’s people from financial services and credit cards to one where all sorts of people are here. It’s great.
Now, your business, tell us a little bit about what it is that you guys do.
So, Decision Management Solutions focuses on companies that are really trying to automate decision-making. So, there’s some high-volume transaction where it’s not obvious what to do. So, you have to decide what to do for each transaction. And, obviously, that’s driving a lot of analytics. A lot of people what they want to do is they want to use machine learning, analytics, AI to make a better decision about these transactions, but they’re high enough volume that you can’t just show stuff to people and hope the people can handle the transaction, you got to automate it. So we help build those automated systems, into which you can embed these kinds of analytic models.
Do you have like a favorite project or ideal customer-type?
Uh, my favorite project probably is around like “next best offer, next best action” kinds of things. If you’re a multi-line company, you’ve got lots of different products or your products have complex eligibility like insurance products, then it’s easy to say, “Oh, you should make the next best offer to this customer this moment in the customer journey.” But actually, figuring out what that offer is – given what they already own, what they’re allowed to buy, which products go with which other products, what the rules are – is a non-trivial problem for most companies. And those systems tend to be more fun ‘cause they’re not as heavily regulated as some other decisions. So you don’t have to worry about the law quite as much as you do; you have to worry a little but not a lot. The key concern is privacy but the kinds of systems we build don’t need to who you are, they just need to know things about you.
So, you’re not dealing with PII, that’s how you’re able to bypass.
No PII, exactly. It’s one of our sort of rules of thumb. We don’t care who you are; we just need to know what kind of person you are, what kind of products you own. We need that data, but we don’t need to know which particular customer we’re talking to. And that makes it… It’s great: you can use the cloud; you can use advanced analytics.
It opens up a lot of stuff as soon as you can divorce the PII from the actual….. It’s such a big issue.
It’s a big deal, yes, and particularly as companies are trying to move to the cloud, and particularly with analytics, and you need that horsepower. And they get very nervous when you start making decisions about customers ‘cause they all want to know about PII. So we have lots of meetings with infosec people, who come in, and we describe how it works and they go, “Oh, OK. We’re done.”
Oh, that’s a very nice shortcut.
Because you’ve been part of the ecosystem for a while, what has been like one of the big trends that you’re seeing emerge in the space?
So, I think there’s a couple of things that have really changed. One is, when we started, getting from what the analytics team built to something you could execute was often a huge production: a lot of coding, a lot of recoding. And so, we would talk about deploying the model as a big barrier. Nowadays, you look at the modern tools that are out there, and they got one-button deployment. And one of my colleagues has this great phrase: He says, “You know it’s not about deployment; it’s about employment.” However easy it is to deploy the model, are you employing it? Are you doing something with the model? And so, that’s always been our focus, but now we can’t talk about deploying the model as a sort of phrase ‘cause people go, “But I’ve got a button for that.” I go, “Yeah, I know you have a button for that. Now you have an API that calls your model.” Still, “Is anybody using it? How’s it going to be used? Can you explain how it’s being used? Can see if it’s being used appropriately? Can you see what the key factors were in the model when you used the model ‘cause you don’t care what the factors were when you didn’t use the model? You only cared when you used the model.” All those kinds of questions come up. That’s really shifted the conversation away from that.
The other thing, I think I would say, is I see more IT departments that care more about analytics. When we started, the IT departments were like, “Don’t talk to me about the predictive analytics guys. They run their own server; they do they own thing. I just try not to fret about it.” And now, I wouldn’t say that they’re totally in it. But we meet analytics teams that are part of IT; there’s a lot more integration. A lot of BI teams are trying to get into data science, particularly analytics. So there’s a lot more overlap. I think that’s a really good thing because I think IT in most big companies is the key barrier for a lot of analytics projects. If you don’t get IT to buy in, then you’re not getting there.
Yeah, absolutely. It’s a big budget and also a gatekeeper.
‘Cause they’re like, “You’re going to destabilize my system with this probability nonsense? No!’” So you have to find a way to get through that problem.
Yeah, yeah, for sure. You’ve got a lot of management consulting around the actual, not just the execution, but, as you said, the employment of those insights inside of the organization. What are you seeing as other trends? How’s the space going to be different in the next two years?
I think one of my pet projects is get business analysts, more generally, to include in their requirements the need for machine learning, the need for analytics ‘cause one of the things I see at the moment is the only people who are really thinking about how you might use machine learning are people who know how to build machine learning models. But in most big companies, they don’t write the requirements documents; they don’t write the specs. And so, by the time, they get involved the requirements document’s written; the IT budget is set; the project’s up and running. And then, you’re constantly trying to bolt things in and add things to dashboards. And it’s never integrated. And so, one of my pet projects is to get business analysts to be much more specific about saying, “This system is making this decision. I wonder if we could use data to make it better. What would that look like?” And get them to drive that into their requirements so that right at the beginning of a system’s life cycle, people are talking about, “What decision does this system make? How do we use analytics to get better at that decision?” We’re going to get scale in companies.
So you really have to move upstream.
Oh, yeah, right. My title of my talk today is “Doing It Backwards” ‘cause I feel like people say, “I’ve got the data. See what interesting things I can find out about the data. And then I see if I can use that.” And I’m like, “Well, that’s backwards.” You actually need to start by saying, “What problem am I trying to solve?” and then see what analytics you need to solve it, and then see what data you need to build the analytics. They’re like, “Well, but where’s the research? Where’s the hypothesis testing? And I’m like, “Yeah, if you’re the kind of company that’s good at research, you should do a little bit of research.” But most companies are terrible at that kind of stuff. And you’re going to be the same kind of company in ten years time that you are now. So focus on the things you have to do to be that kind of company and use analytics to get better at them. Rather than trying, “Maybe, there’s a new business in my data.” Well, maybe, but probably not. If you’re an insurance company today, in ten years time, you’re going to be an insurance company. The only question is whether you can use data and analytics to be a more profitable, more effective, more successful insurance company or not. So that means you got to start by thinking about, “What does it mean to be an insurance company? What decisions do I make?”
Isn’t that interesting how it’s like, it’s really helping frame the business as a whole beyond just… That was a very good answer.
What I find just fascinating is that we’ve seen this like democratization of insights or analytics across the organization. So, it used to be relegated to a few, and now it feels like it’s seeping through many, many different, even IT, which is amazing. So, as we continue to evolve… I loved the way you depicted it: moving upstream and being involved in the early stages of the product and then having that actually built into the key requirements. I mean it’s going to be an exciting world because really brands are at a spot where they get to decide if they’re going to win or lose on Day 0, based on how they’re incorporating analytics into their platforms.
Yeah, absolutely. One of my favorite phrases I often talk to my typical customer is I like to say a “big, boring company.” And people keep telling them to use machine learning and AI. They got to be agile and nimble and think like a startup. And I’m like, “No, being big and boring is actually part of their value proposition. Buying life insurance, you’re buying it from big, boring company. You’re banking with a big, boring company. That’s kind of the point, right? If you want to buy a tractor, you want to buy it from a big, boring company, right, ‘cause that’s the point, right? And so, it’s not enough if we say only small nimble startups, digital natives can use analytics pervasively. Big, boring companies have to be able to use it pervasively. And that means we’ve got to fit the way they build systems, the way they think about their business, which tends to be lots of documentation, lots of thinking, long lead times, requirements. If we don’t get data thinking into that process, they’re never going to be pervasive users of data. And they need to be.
My guest today has been James Taylor. If somebody wants to get in contact with you or your firm, how would they do that?
So, the easiest way to find it is DecisionManagementSolutions.com. It’s a really long URL but all the words are spelled exactly the way you say it.
It’s great for SEO, by the way.
That’s always good. And then, James@decisionmanagementsolutions.com gets me. And that’s an easy way to find us.
James, thanks so much for being on the show.
Thanks very much.
Everybody else, if you please take the time to screen shot and share this episode. Special thanks to Predictive Analytics World and Marketing Analytics World for hosting this particular episode. I hope you have a fantastic rest of your day.