And so I was very curious about my own experience and very conscious about what I was feeling, and it felt like the products were at the center of the entire experience. It wasn’t really about the shoppers. The stores are saying, “Hey shopper, come into the store. I have this product. If you can fit into it, great. Otherwise, go to the next store.”
I’m a shopper. I'm willing to dedicate my money, attention, time, frustration, emotion, and everything that goes into the shopping experience. But the retailer is saying, “I can't do anything with it.” How is this supposed to be a good experience? Why is retail not really trying to learn about me and my body when I can potentially spend a lot more money with you if you did?
So,there you are, frustrated with the shopping experience. What was your next move?
I also wanted to ensure it wasn’t an immigrant problem I had. Had American women figured out a smooth shopping experience where they felt great about their bodies? So, I incubated this hypothesis. I'm a Yale spouse, so the university incubator gave me access to a lot of American women of all different age groups and professions. Over the next few years, I spent time asking them one question: Why did you buy the last item of clothing you bought?
And from those thousands of conversations that I had with women, I noticed two key things which are the founding pillars behind what we’ve built at Lily AI. First, I was really surprised at how much detail they used to describe these products. The second thing we learned was that some of these details they were describing were basically their perceptions of how a certain fabric or a certain color makes them look taller and thinner and more fancy or less fancy.
Each shopper’s perception of the same attribute or same detail is very different. And that emotional context plays a role in that. So if the need is a delightful shopping experience that is relevant and personalized, then emotional context must be understood. Without understanding that context, you can do all the personalization that you want to do in the world, and its till is not going to be relevant.
Through this customer development exercise that I did, I was able to understand the problem very well.Fundamentally, it came down to: If I have to make the shopping experience relevant for somebody, two things are required. I need to understand all the product details that could be a reason for a shopper to like or dislike that product. And I need to understand how each shopper reacts to those things. And these were two fundamental pieces of data that retailers had not leveraged.
It feels like a huge undertaking to create something that understands individual shoppers’ emotional context. What were the first steps in creating a solution?
My non-technical background worked in my favor because there were no boundaries of what AI can do or not. We never restricted ourselves to what experts in the space said is possible or not possible. Instead, we were so focused on solving the problem for the consumer; we worked backward from that and created our own standards of what is possible.
I’ll give you an example. The majority of teams trying to do, let's say, image recognition, focus on acquiring a few specific types of datasets that are already out there. And they are massive datasets, but they're not like very good training datasets. Even large companies, at the time we first started, were working off of that readily available data.
We also looked at that data, but we really quickly understood that it wasn’t good training data for our AI. We actually knew almost from day one that we needed to create our own training data. And today, there are a billion-plus training data points that we have used to train our AI models, which are doing a fantastic job. We have some of the deepest capability in the space.
Tons of brands are trying to do this thing called personalization. What makes Lily AI so different, and how are you guys able to bring something new to the table?
To make a shopping experience relevant for the shopper, you need to understand the products in detail and you need to understand how each customer reacts to those details. So how do you get access to this data? How do you create what is very incredibly hard to create? We had to build it, grow it in-house to train our AI. We've built all of this in-house ourselves over time,which has given us an edge.
But we also had to understand how brands could leverage the enriched product catalog data and the psychographic information. We've spent a lot of time trying to uniquely solve the problem of how to attain better customer data and how to apply it. We've used the meta model of the end consumer, not just centering the experience around the products but also the shoppers, as our North Star. Whatever we have built, we have kept empathy at the center.
How are brands able to take your psychographic profile to see business outcomes?
As we spoke to brands and retailers, we wanted to understand what different teams and applications use products and consumer data. It turns out almost every team uses product and consumer data in some form. But the most popular use cases of enriched product and consumer data are with site search recommendations; it contributes to a good percentage of the overall revenue. We can prove an incremental lift in less than six weeks once the retailer switches to this new methodology of personalization, leveraging true product and consumer context. (Read more about use cases here).
Is there anything else that you say to brands as they're looking to improve their digital experience?
Retailers are at a point with e-commerce where they cannot treat it like a stepchild. It is an important part of their business now,if not the most important part of their business. It's so much more important for brands and retailers to make the shopping experience relevant if they want to be successful. Shoppers have so many choices. The sooner brands can leverage enriched product and customer data, the more ready they will be for the future. It’s the next wave and evolution of e-commerce.