categories

HOT TOPICS

Thought Leaders in Artificial Intelligence: Josh Sutton, Data & Artificial Intelligence Global Head at Publicis.Sapient (Part 3)

Posted on Wednesday, May 3rd 2017

Sramana Mitra: As I’m talking to you, I’m making a note in my head to invite Mark in this series once you have finished getting to a point where this is all real.

Josh Sutton: That was the first front around insight generation.

Sramana Mitra: Could you actually throw some light on what kind of products this retailer sells?

Josh Sutton: This was an electronics retail company.

Sramana Mitra: You said you already knew a lot of data about customers coming to their site. What was the source of that data that you were working into the learning algorithm?

Josh Sutton: In that particular case, the source of the data was their own internal organizations as well as what we could pull. Initially, the use case was much more around leveraging their own internal data. We use that to drive the recommendation engine for the most relevant product categories that we would show to them on their home page.

I’ll give you a different example of some of the things we are doing today, which are a bit more advanced than that. Over the past two years, we’ve built a very significant platform that enables us to aggregate individual-level data on everybody that has a digital footprint in the US and different countries as well. The US tends to have more lax data laws than some of the other parts of the world. I can tell you, with a fairly high degree of precision, for about 90% of the country, what somebody does the moment they wake up and the moment they go to bed, where they’ve physically been, what they’ve looked at online.

We had to renegotiate about a hundred different contracts with different data providers to shift our ability to collect our data from a segment level down to an individual level. Now we’re using this in the exact same model with some other very large retail organizations to take that same model of being able to provide recommendations and provide targeted information to a much higher degree of precision, because we’re not just relying on the first-party data that’s owned by the company. We’re able to augment that with third-party data that we’ve collected from a tremendously wide variety of sources.

When someone comes to a site, we’re able to identify what exactly is going to be the best message to show to them. More importantly as they move through the journey of exploring a purchase or thinking about doing something, we’re able to optimize the message that we deliver at each point along that consumer journey based on what we’ve seen work and not work. This really gets to the power of machine learning optimization and being able to take insights that we can extract from massive datasets, and optimize on an individual level. In some cases, we’ve seen increases in efficiency to the tune of multiple hundreds of percentage points.

Sramana Mitra: Give me a couple of examples of what kinds of recommendations or what kinds of decisions you can make based on availability of this level of precise data and the learning algorithms that you have access to.

Josh Sutton: This is an example from a very large retailer. They initially started with a very hierarchical recommendation model and standard navigation models. What we then started looking at is how people navigated through their site. This is a clothes retail store. We were able to drive out an entire shift in how they’re structured to where we’re able to show them, in real-time, what we think they’re going to be interested in.

As they behave in a certain way, we’re able to take that based on machine learning models that we’ve already trained on and know where to steer them for incremental recommendation and even get to the point of how to show what bundles we might want to bundle together and what offers we might want to make to maximize the likelihood of sales and profits. Depending on the quarter and what they’re looking at, they might be looking to optimize revenue or profits. Those are two different optimization models for how you steer people.

This segment is part 3 in the series : Thought Leaders in Artificial Intelligence: Josh Sutton, Data & Artificial Intelligence Global Head at Publicis.Sapient
1 2 3 4 5 6 7

Hacker News
() Comments

Featured Videos