Early ecommerce personalization tools significantly increased retailers’ conversion rates when they were first introduced, displaying relevant products based on obvious triggers from customers’ browsing activity and past purchases. Today, however, these tools are showing their age. In particular, they limit the customer’s ability to shop the full depth and breadth of a catalog on a merchant’s website – leaving a lot of incremental revenue untapped.
PSYKHE AI is a recommendation engine designed to solve this exact problem. Alongside standard personalization functionality, it also personalizes ecommerce site grids, based on a wide range of data, including the customer’s psychographic traits. It works in real time, reranking the selection every 10 seconds with each user interaction.
In our latest episode of Martalks, PSYKHE founder Anabel Maldonado describes how PSYKHE aims to make more human-like recommendations based on a wide range of data, and knowledge of the user’s personality – rather than simply showing visually similar items:
If you were in the store at Bergdorfs, they’d bring you a bag, a clutch, a jacket, shoes. If you really felt that they understood you, that they were really hitting the notes and got your style, you’d probably buy more than one thing.
In this conversation we discussed:
- How PSYKHE AI compares to previous personalization tools
- The concept of ‘relevant adjacency’ – showing products based on real-time data about the customer and their style preferences, and not just their most recent search.
- How PSYKHE establishes the psychographic profile of both the items and consumers, and maps these together.
Listen to the full podcast here:
The limitations of standard ecommerce personalization tools
Darrell: “Most product discovery tools are limited in their ability to discover the depth and breadth of the catalog on a diverse merchant’s website. Why is that? Why don’t you see most of the stuff they have buried in the back – and how does Psykhe AI improve on that?”
Anabel: “There’s been this idea of, ‘what’s good enough? What’s a scrappy, scalable way to increase conversion by a good-enough degree?’
Other personalization tools have approached this with a very simple CatBoost model, only in the carousels but not the product categories, and assuming the customer’s intent to buy visually similar items.
So it’s been a very rote mechanical expression, which has been important in paving the way, but I like to call that ‘personalization 1.0’.
Now, we’re going beyond the carousels and visually similar items, to what we do at Psykhe AI, which is re-rank the entire catalog and the whole category in real-time. This surfaces more of the catalog as you describe.
If you think about it, if that category is static or not personalized, or you’re floating the same few best-sellers to 20m people, then the thing that might convert you or I might be on page 212 and we’d never see it. It cannot convert because it’s just never been seen.”
What’s the lift of Psykhe AI?
Anabel: “We’ve been seeing 3-5x conversion rate increase, and as of last week, we saw up to 168% increase in engagement metrics.”
Darrell: “So people really do buy based on their personality?”
Anabel: “We’re looking at a lot of inputs. I love to talk about the personality angle, and it’s part of what gives us our depth. But make no mistake, we’re not saying one thing is more important than the other. There just feeding the algorithm allthe information it needs.
Sometimes you’re just looking for the cheapest pair of Airforce 1s in a size 9, and we can pull those out too.”
About The Rosenstein Group
The Martalks Podcast is published by The Rosenstein Group, the leader in martech executive search. For over 20 years, we’ve been recruiting heads of sales, channel sales leaders, and other members of the commercial team, across martech, supply chain, ecommerce, sales enablement and systems integration.
Find out more here.
Apple Podcasts
Spotify
Google Podcasts
