Thought Leaders in Artificial Intelligence: Debjani Deb, CEO of ZineOne (Part 2)

Sramana Mitra: Let’s do a couple of other use cases. Give me a use case that is online. Let’s do an e-commerce use case.

Debjani Deb: Let’s take an e-commerce vendor. You are shopping on Overstock. You have taken the first click. You landed on the first page. You’ve taken the second click, and now you’re on the product page.

The machine learning algorithms that ZineOne has deployed on the website of Overstock is looking at your individual browsing right now. By the fifth click, it’s determining with about 90% accuracy whether you are going to checkout in that session or not.

If that algorithm determines by the fifth click that you are what we call the on-the-fence audience, it can nudge you with something that will incentivize you to go further in your customer journey. That nudge could be a loyalty nudge, a shipping nudge, or it could be an offer.

Essentially, this is very cutting-edge because thus far, there is no competition. Some of the big companies are doing it on their own. They are deploying what we call ML algorithms on a stream. Data hasn’t been stored yet.

Think about how fast this needs to be. ZineOne determines if the customer is not going to complete. That enables some sort of an incentive or a nudge. We’ve seen 12% increase in revenue on these kinds of things between test and control groups because you’re able to nudge them to completion. That’s an e-commerce use case.

Sramana Mitra: At the heuristics, what is driving the conclusion that you’re making? Out of five clicks, you’re determining if this person needs a nudge or is going to complete. What is the input?

Debjani Deb: This is a patent-pending technology. We’ve filed five patents on this. We call it the Customer DNA. Essentially, it’s the consumer genome mapping. I’ll talk a little bit about it. You know that the human genome has four nucleic acids that make up 32,000 gene sequences.

We’re using that analog as a framework to look at the clickstream data. We are mapping millions and millions that are coming into our cloud to understand what set of sequences or what set of behaviors displayed lead to what set of outcomes.

We are training our ML model through this notion of this consumer genome mapping to understand what displayed behavior will lead to what outcome. 

Sramana Mitra: So you’re doing pattern matching. You’re tracing the logs and looking at the results of those logs on what happens. Then you’re drawing conclusions and learning based on that.

Debjani Deb: Essentially, the only nuance here is that it’s being done as a real-time stream. You’re seeing past patterns and sequences.

Sramana Mitra: You need the past data to do the pattern recognition and the heuristic tracing. Even before learning, you’re sorting the real-time data into buckets and then the learning algorithm is learning more on top of that.

Debjani Deb: Exactly.

Sramana Mitra: You said banks are also part of your customer base. What do you do for banks?

Debjani Deb: We think of ourselves as direct to consumer. The same principles apply, which is to say that I’m looking at somebody applying for a car loan. I’m looking at the behavior displayed. I’m making a call with regards to how I can add value to that consumer journey.

I am looking at somebody who is trying to pay a bill and they look confused. They’re about to call the call center. I can pop up. The algorithms are running in the background. This is very different from the popup technologies that exist today.

Your most important desire is not to spam the user. You will do this maybe once every day. There are controls in regards to how often you can interject.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Debjani Deb, CEO of ZineOne
1 2 3