Will Brands Ever Get Product Recommendations Right?

Consumers today are more accustomed than ever to personalised content — except while shopping online.

On Instagram and TikTok, they’re served up exactly the right content to keep them glued to the app. Netflix’s programming is calibrated to maximise how many hours its subscribers spend on the service.

Online retailers use reams of customer data and deployy the latest AI-empowered software to determine how much of any given item to stock, and in which sizes and styles. But few are taking that critical next step to offer individualised product recommendations and search results on their e-commerce websites — what experts call true personalisation.

In a survey of more than 100 brands and retailers, only 20 percent of them said they customise product recommendations based on a customer’s purchase history, according to research from the retail and supply chain software and services firm Manhattan Associates.

Adoption has been slow because even with current technological advancements, implementing personalisation in e-commerce is complicated. Brands must either build algorithms that can predict customers’ behaviour from scratch, or use software platforms that can do the heavy lifting for them. But the necessary tech talent can be hard to recruit and AI-driven software platforms are often costly. Still, some brands are finding workarounds, such as offering surveys, to collect first-party data that can be used in personalisation.

“Customers are expecting a personalised experience,” said Anabel Maldonado, founder and CEO of software company Psykhe AI, which provides personalisation as a service to retailers. “Shopping pages still being static is out of alignment of what we know we can do with AI.”

Common Barriers

There aren’t many examples of retailers that have successfully managed to offer personalisation, even among established brands.

In 2019, Stitch Fix, which is known for its styling service and monthly subscription boxes, began pushing customers to buy items recommended by algorithms based on their style preferences directly from its site. But these efforts ultimately caused a decrease in sign ups for its core service, resulting in a sales growth slump by 2022, company executives said on earnings calls. Multi-brand retail start-up The Yes spent years developing algorithms to create a marketplace that offered users the most relevant brands and items suited for their individual tastes, only to be shut down after being acquired by Pinterest last June.

For smaller brands, implementing personalisation features is especially cumbersome. Platforms with built-in AI solutions can charge a minimum of thousands of dollars a month, Maldonado said. By contrast, e-commerce software firms that power brands’ online storefronts, such as Shopify and BigCommerce, charge as little as $30 a month.

Building and maintaining custom algorithms in-house is even dicier. Fashion brands often struggle to lure data scientists away from more lucrative jobs in the tech industry. (Although recent layoffs at top tech companies could make this easier.) Many brands are not adept at gathering and interpreting the customer data they already have on-hand.

“If [brands] don’t fundamentally understand their customer, personalisation of the website is not priority number one,” said Sona Abaryan, a partner at Ekimetrics, which delivers AI-based solutions. “There’s a lot of work to do on the fundamentals. Understand your customer and where you want to go before you start deploying a bunch of tools.”

Getting Scrappy

Fashion and beauty start-ups that can’t afford sophisticated AI solutions can start by introducing services to help them in data collection.

“Leverage what [customers] do with your brand in that first-party environment to personalise their journey on the website,” Abaryan said. “If they’ve left something in the cart, if you know their size, these are the features that you can personalise to make their experience easier and more relevant.”

For example, skin care brand Fig.1 offers a consultation service. After customers answer questions about their skin concerns, products they currently use and upload a photo, they are sent an email from a licensed esthetician with a customised skin care routine full of Fig.1 products.

This service was primarily developed by one in-house engineer. The brand also employs one full-time licensed esthetician, along with a few freelance consultants in the same field, who selects items to recommend to shoppers who complete use the service.

Similarly, five-month-old performance footwear seller Hilma developed a fit finder to help shoppers navigate its 45 sizes, which range from 5 to 12 and include several widths and lengths. First-time customers answer questions about their body type, the shoe brands they normally wear and the volume and dimensions of their ideal running shoe and are then shown the Hilma running shoes mostly likely to fit their feet.

Brooke Torres, Hilma’s founder and CEO, wrote the algorithm for the fit finder and worked with a developer to code it into the site’s backend infrastructure.

“We never considered a running shoe that wasn’t personalised,” Torres said. “What I was trying to mimic with it is creating the best shoe salesperson you could imagine.”

However, brands are hopeful that AI software will be more affordable in the coming years. Andrew Bernstein, co-founder and CEO of clean cosmetics subscription box brand Kinder Beauty, said the company will be quick to invest in such software when it’s more cost-effective. Until then, Kinder is gathering more of its own data, allowing customers to choose two out of the five items they receive in their monthly boxes to learn more about their preferences.

“Where we are now is a starting point. We aren’t anywhere close to the final form,” Bernstein said. “As soon as we are able to dip our toes into those customised and personalised waters, we’ll be the first ones to jump.”