Online shoppers are a fickle bunch. Research has shown that e-commerce websites have as few as 30 seconds to impress customers or lose those shoppers to greener digital retail pastures.

But a new set of algorithms, developed by a team of researchers, may help. While many e-commerce sites display products using a ranking mechanism that chooses items based on their past popularity, the researchers found customer engagement rose by as much as 30 per cent when rankings were determined through a dynamic blend of product popularity and variety among products.

“We were surprised by the magnitude of the results,” says Shreyas Sekar, an assistant professor of operations management at the University of Toronto Scarborough and the Rotman School of Management. “We thought that maybe we’d get a five percent improvement, which is not bad because we’re dealing in the order of hundreds of thousands of customers, if not millions on some of these platforms. But this was amazing.”

Sekar and his colleagues focused on hedonic browsing, the online equivalent of window shopping. These are consumers who are not looking for anything specific but who are casually checking out sites for something that might catch their eye. The retailer’s task is to present an initial set of products appealing enough to get customers past that crucial first impression window so that they are hooked and will stay on the site to look some more, instead of leaving.

It’s a tough sell; about two thirds of online shoppers never move beyond that pivotal first page of listings.

Working with Wayfair, a multi-billion-dollar home goods online retailer, the researchers designed and refined a set of algorithms to help Wayfair’s website determine and continually adjust for the optimal set of product rankings for any given shopping event, be it Black Friday, Halloween or another special event. The algorithms kept the typical attribute of product popularity but added diversity of items, leading to rankings with broader customer appeal.

Using historical data from Wayfair, the researchers were able show that across six different events, their algorithm hooked an average of five to 30 per cent more customers than an algorithm based on popularity alone. Researchers also designed the algorithms to be efficient learners, so that they did not need massive amounts of customer data to determine the best product mix.

This is the first study to consider online retail ranking decisions in the context of browsing shoppers versus those with specific purchase intentions. Its findings extend beyond shopping platforms to any ranking system or display targeted towards maximum user engagement, such as news media top stories lists and email marketing campaigns.

“We tend to think that revenue is the only objective that online retailers care about, but increasingly platforms care about growth and retaining their user base,” says Sekar. “If a customer gets hooked today, the platform may get to keep them as a repeat customer over and over again.”

The study was co-written with co-written with Kris J. Ferreira, of the Harvard Business School, and Sunanda Parthasarathy. It appears in Management Science.