Thought Leaders in Artificial Intelligence: Grant Ingersoll, CTO of Lucidworks (Part 2)

Grant Ingersoll: In the case of one of our large telecom providers that we power ecommerce for, if a user comes in and searches for iPhone and they’ve never done business with that company, then you want to show them the latest iPhone. If you happen to know that that user is already logged in and they’ve already bought the latest iPhone and the query is iPhone again, the chances are they’re looking for support or accessories. It’s that kind of behavior that machine learning can really help move the needle for with retailers.

The same data that makes for more relevant search is also the same data that makes for recommendations, personalization, and queries. That is all of that user feedback loop. Then these days, a lot of the work done is shoving all of that into tools and essentially creating a much smarter ranking function that then returns those results out to the user.

Sramana Mitra: I’m going to double-click one more level down on what you talked about. In any personalized online store, the store that I want to see and the store that you want to see are different stores. Could you talk about the trends in personalized stores? By the way, I did one of the first ever companies in online fashion in 1999. The whole concept was around the personalized store.

Grant Ingersoll: It’s really hard. You don’t actually know who your users are. You may have a cookie. You may have third-party sources that have tracked that user across other sites. If you’re lucky, they bought from you in the past and they haven’t deleted their cookies. Personalization is really a spectrum between anonymity all the way through to cohort segmentation and all the way down to personalization.

To your point of what you like on the site and what I like on a site is often a number of different factors. This doesn’t just apply to retail but we can double-click into retail. If I can put you in a bucket with other people who are like you, then that shows an improvement. The next level then is along the lines of that example of the iPhone. If I know your prior behavior, then I can use that to customize the results that I showed to you. We’ve all had this experience of retailers where they’re showing you a hundred versions of a thing you already bought.

I remember in the early days of Amazon. I would go buy a book. It would show me books of the exact same thing. It can go all the way down even to what you’ve returned, what you rated, what you disliked. You think of this search and machine learning engine as a big mashup engine that takes in all of these different features that I know about you.

At the end of the day, it’s this ranking for you at that point in time based on what our current state of catalog is. It’s completely customized for this situation. Of course, you can extrapolate your browse activities, your recommendation activities but it all comes from core mashing together of all of these different features.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Grant Ingersoll, CTO of Lucidworks
1 2 3 4 5