The COVID-19 pandemic accelerated the growth of e-commerce. People were left with nowhere to shop but the internet, aside from local grocery stores and drug stores. Macy’s recently announced it would overhaul its existing website to offer more personalized experiences to their e-commerce shoppers. Macy’s was just the beginning. Now, more companies are working to increase their conversion rate for online shoppers.
Customer shopping data can be used to improve e-commerce personalization. The question on many people’s lips is how? The answer lies in artificial intelligence automation. AI can be used to process customer data, ranging from purchase history to search history, to improve sales conversion. Hamish Ogilvy, CEO and co-founder of Search.io, an AI-powered e-commerce search platform, sat down with FashionUnited to offer his expertise on how AI is transforming e-commerce.
We work with various retailers and companies to do personalized recommendations and help with browser pages and filter pages. What we are really good at is putting products in the right order for each customer to maximize sales and conversion.
Say someone continuously picks a size of shoe. That data is kept, so when the customer returns, that information is added onto the search. What you find is when people search for shoes and get the result they wanted, they filter for their size, and then their size is out of stock. The percentage of people who will then bounce and leave the site is very high.
Now, with AI, if I know someone is a size 14, then it’ll put all the size 14s in front of the customer, even if it’s not their first choice of style. That has a huge impact on conversion and revenue.
Department stores collect a lot of data. If you’re browsing shoes or electronics, they know the interests you have right now. Then there’s the personal aspects. There’s the browsing anonymous information to understand the intent, then you have the ‘I know this person is a customer and I know how they buy things, let’s use an algorithm according to them.’ The main goal is to identify bits of information that are relevant, then there’s machine learning to maximize revenue or conversion, or whatever you want to do.
It’s very hard. Many components in search are not actually using artificial intelligence. They are using more tried, trued, and tested machine learning techniques. A lot of things, like what products to show you and in what order, are more like machine learning in things like gaming. AI models need a lot of data to train. They also need a clean data set to train. When you use techniques like reinforcement learning, you can know nothing about a situation and get better and better over time without explicit training steps.
For example, in Australia a sweater is called a jumper. We had a team in Australia doing work for a big U.S. retailer. The test crew was using the phrase X-mas jumper. They were expecting to see sweaters, and they saw sweaters because the artificial intelligence understood a jumper is a sweater, even though people in the U.S. don’t call it that. That’s where AI is having a huge impact. It understands the meaning and query of products, and understands how to search that effectively.
If you look at it as an opportunity for a retailer, it’s enormous. We added brand infinity, meaning if you buy a particular brand, you’re highly likely to keep buying that brand. That raised conversion by 2 percent, which equated to millions of dollars per year just by adding that small piece of personalization.
With things like groceries, you repeat buy the same things. If you buy dish soap online, you’re highly likely to buy that same dish soap again. The biggest conversions we made for one company through that personalization change resulted in 25 million dollars in revenue.
Companies like Fast and Bolt, which do payment experiences, are trying to tap into the personalized parameters from different sites. There’s a lot of things happening in the AI space, but the revenue impact is enormous
They don’t need that much data actually. When you think about it, when you go shopping on a site and you’ve filtered for a certain size and gender, then you’ve already indicated your personal preferences. You can use those preferences to bias your results for the rest of that shopping experience, and then keep that information for when customers return.
You do have to be careful because sometimes you’ll have people shopping for someone else, and you don’t want to remove yourself from other data. In terms of collecting data, it’s easy to collect browsing data and purchasing data. That data can then be used to personalize the whole experience.
Will there be more brands or less brands in the future? That’s a conversation that’s been going on a lot lately. I recently heard 48 percent of e-commerce purchases are done through marketplaces, but at the same time, you can start a brand tomorrow through Shopify and don’t have to go through warehouses and distribution overhead.
Look at Fashion Nova. That’s a huge business worth billions of dollars that started on Shopify. At the same time, you have companies like Amazon, who are pushing out smaller players. The challenge for brands is scale and speed of shipping, when companies like Amazon or similar expedite shipping. Small brands will need to figure out how to compete.
Subscribe to get the latest posts sent to your email.