At a time when “data scientist” is akin to “rock star” in the tech world and omnichannel has become a central theme for even mom-and-pop stores, data can be both a gift and a curse.
Jonathan Beckhardt, general manager of insights and cofounder of DataScience Inc., helps e-commerce players such as JustFab and Tradesy sort through the noise to make sense of data.
“There’s a lot of nebulous conversation about data,” he said, “but what is it that retailers are actually doing with data? Nine out of 10 people talking about data won’t have a good answer for you. How you actually use it is where the rubber meets the road.”
Beckhardt gave WWD a primer on what retailers should know about working with a data scientist — the potential untapped goldmine that resides in those clicks and wires.
WWD: When we hear the word “data,” what are we really talking about?
Jonathan Beckhardt: There are two things you can do: analytics and data science. They are closely related, but definitely different.
Analytics is knowing what is happening in your business at a point in time. Knowing, for example, that I sold 250 sweaters last week in Miami, or knowing revenue is down 10 percent from two months ago. It’s tracking your business.
Data science is about using data to understand the fundamental drivers of your business — using data to develop an understanding of how your business works and what makes what happen. It’s using it to model your business, formalize the cause and effect of your business. It’s knowing if I do more of “x,” I will get more of “y.” So while analytics is, “I sold 250 sweaters,” data science is about saying all these things happened in Miami last week and mining those events so if I do more of this campaign, I will sell more of these types of sweaters.
It’s a tool with a larger goal of understanding how to better drive value in business.
WWD: Are retailers focusing too much on analytics?
J.B.: They focus on information like average order value and conversion rates — single points. Part of that is a necessity of business, but it’s really a way undervaluing an opportunity. Essentially, you’re dividing that customer into more segments.
One of the things that a business should optimize for is true customer lifetime value.
That’s not the only thing that data science can do, but it’s one of the few things it can do. Very few businesses have the capability to predict where every customer is going to be. It can become powerful, like, “I know this is where my customer is going to be in three years, so I can do more of what’s making them long-term customers.”
WWD: What else are retailers missing out on?
JB: Knowing the status of a customer. Imagine if I have a customer who I see every day and then I stop seeing that customer for a week. I’m going to say, “That was my best customer. She bought three things, and then I didn’t see her for a week, I wonder what happened to her.”
That happens all the time. If you have a million customers, and it’s happening with a certain percentage every week, if you are not looking at the data in a rigorous way, you’ll miss that. And one of your best customers disappeared. By understanding the patterns, you can take out who had a bad experience.
The other thing you can then do is predict when that will happen.
WWD: That sounds easier said than done.
JB: My personal pitch would be to hire a data scientist to provide both products and services.
For products and services targeted to specific problems, it’s really about leveraging data scientists’ expertise. A lot of people find the first person comfortable with data and call them a “data scientist,” but they need to deal with uncertainty and noise; with statistics but also engineering, in a rigorous way, and that’s very rare.
WWD: How do you get all this information? Does it just happen automatically if you have an e-commerce site?
JB: You have to make a decision to actively track and store it. There are a number of ways to capture the information. The number-one thing is just capturing it. Capturing event stream data in a data lake is the first thing. Typically, this can be set up by a software or database engineer. There are systems that allow a retailer to decide what they want to track, whether it’s an event-tracking system or just an internally developed mechanism.
WWD: We hear about “predictive data models.” What does that mean?
JB: People are not creating models; they are creating point-in-time understanding. So, going back to Miami last month: If you sold 10 fewer than last week, it’s saying, “Did I do something right or wrong, or is that random noise? Is that thing A or thing B?” Just taking a finger-in-the-wind approach misconstrues one of two things: Patterns that influence the future and forecasting the future, so what does the future actually look like?
WWD: Anything else that retailers are missing?
JB: It takes time to get the most out of it. The return on investment is exponential depending on the amount of time you put in. At first, the returns are slow to come in, but over the course of a year, it becomes significantly impactful. I think a lot of people get discouraged, but after you get over that hump, they see results, and the model gets stronger. After one to three years, they develop the key capability. Now we are seeing the companies that invested in this two years ago, we are seeing them climb up the Internet Retailer (IR) 500. Now it is “make it or break it” time.