I’ve been asked to write a blog about personalization - and how machine learning¹ serves personalization. But what I actually want to write about is rather different, at least at first.
I want to talk to you about shopping. Specifically, sari shopping.
Saris are glorious garments - there are dozens of ways to wear them, they’re made from a range of different fabrics and weaves, they seem to come in every color under the sun, with patterns ranging from plain to dazzlingly intricate - an impossibly large range of options to sift through.
All of this can make stepping into a sari shop rather overwhelming.
That’s where the true genius of sari shopping comes in.
When you go to a shop to buy a sari (in my experience!) they welcome you in, give you a cup of tea, and subject you to a minor interrogation of sorts. And not just your price range. You’ll be asked your* tastes, your interests and your preferences in colors, fabrics and patterns. You’ll be asked if it’s for an event or a particular occasion and if so when and where that event will be… the list goes on.
(*Or, if it is a gift, you’ll be asked all these questions about the person who’ll be wearing the sari - along with discreet questions about their appearance, and how impressive you’re intending your gift to be.)
And then the sari seller will take all you’ve told them, and from the thousands of saris in the shop, will chose a few for you to consider - bringing the choice down from mind-boggling to bearable.
OK, so I may have been talking about personalization all along.
Yes, personalization is about engaging your customers, and tailoring your interactions with customers, and applying techniques to improve business metrics. But an absolutely critical component of personalization is search efficiency: making it easy for the customer to get to what they’re looking for, from the myriad products available. For example, a brief survey of about 50 large fashion and cosmetics Qubit clients showed a total catalogue size of 20M products in 1 year! That’s a lot of product choices to narrow down.
And - like in the sari shop - to help customers find what they’re looking for, you need to understand them. On a website we can’t stop every visitor for tea and a twenty minute chat, but we can infer their preferences from their on-site behavior and history, as well as, on occasion, being able to ask them directly.
Where machine learning comes in is to combine a variety of signals ranging from general social behavior, to an understanding of the catalog, to granular, customer intent. It then uses all of these signals to attempt to deliver the most relevant and impactful products at the right time, through the right channel to the right customer.
So, in our digital sari shop, we’ll be able to get future visitors straight to the colors, fabrics, and patterns they’ll like the most: saving them time, and giving them an experience that’s hand tailored for them. And why does this matter? Because we've studied the impact of personalization and found that it leads to a 6% uplift in RPV for our clients.
¹ Machine learning is an interesting term - often the press uses it differently from the data scientists working in the field. I’ll be blogging about the terminology in future, but for now, what I mean by “machine learning” is a way for a computer system to learn without being expressly programmed - by looking for patterns in data.