'What looks cool isn't necessarily what's best for the customer'

Technology inspires wonder. But in order to power the brands that drive an economy, feats of engineering also have to make business sense.

There are few disciplines where the push and pull between the high promise of innovation and the bottom line are more visible than ecommerce. A new feature may display a product in a more vivid and interactive way, or make it more easily discoverable based on user preferences. But in the end, its utility to a brand or retailer lies in whether it drives conversions or other key metrics.

Eli Finkelshteyn saw how this plays out in ecommerce teams up close. While working as a search engineer at Shutterstock a decade ago, he saw how tools that improved relevance in search results drew cheers on the engineering team. But when the business team saw the same results, it was met with something less than enthusiasm.

“It was a realization that all of the stuff that we cared about on the engineering side, and all of the tools that we were using were not really aligned to what people care about within ecommerce,” Finkelshteyn said.

For Finkelshteyn, entrepreneurial inspiration came when this insight combined with a goal to level the playing field for brands and retailers standing up their own ecommerce sites in an Amazon-dominated world. The result was Constructor, which launched in 2015. The product search and discovery platform aims to balance advances in relevance and personalization with the necessity to generate revenue on ecommerce sites.

Now, The CB Insights Retail Tech 100 honoree is working with clients including Sephora, Birkenstock, Bonobos, Backcountry and Serena & Lily. Against macroeconomic headwinds that have led to a consumer pullback, the company doubled revenue in its fiscal year 2022, while expanding both in consumer categories like home goods and furnishings, as well as geographically in EMEA.

Constructor’s site search and discovery tools employ AI. But the company’s work with brands and retailers is oriented around the very human question of how success is defined. This changes how technology is used.

Take search relevance. In web experiences as a whole, relevance can be defined in a variety of ways that range between matching the search phrase with aligning to what the person wanted. Success is also dependent on a number of factors, including whether a user clicked through and remained on a site.

But in ecommerce, the definition is more narrow. It all comes back to driving business results. Ultimately, that means revenue.

“For both the retailer and the shopper, if you're giving them something that they want to buy, then that's successful search,” Finkelshteyn said.

The tools employed around search to drive this success must align with business goals, as well.

To power personalization, Constructor is joining the industry as it is increasingly moving to first and zero-party data that is collected directly from customers through purchases or loyalty programs. This not only allows brands to better understand their customers, but does so in a way that aligns with privacy standards that are top-of-mind in the wake of Apple’s App Tracking Transparency and the coming deprecation of third-party cookies.

To collect zero-party data, which describes data that is given willingly by customers, Constructor recently launched a new product called Quizzes. The product poses questions to consumers throughout a digital shopping journey, much like an associate would in-person. Then, the data can be used to personalize the shopping experience for the user. So if a person says they are gluten free or a size medium, the results a user sees can be more reflective of these attributes.

Finkelshteyn said engineers are still in the “infancy” of tying this data to search results. “But as time goes on, that's the sort of personalization that builds trust with the user,” he said. “That's going to be the future.”

As development continues, it’s important for businesses to remember: There’s a fine line between personalization that drives business results, and personalization for personalization’s sake.

This lesson can apply across a number of new tools that stand to change search. Google and Etsy are testing visual search that can produce results from a picture. Generative AI tools like ChatGPT could introduce a more conversational element to the experience of finding the right product. But in each case, it’s important to evaluate how new use cases align with ecommerce goals. What wows in a demo may not make the trend line go up when you turn to the metrics.

“It's really important for those of us building to recognize that what looks cool isn't necessarily what's best for the customer,” Finkelshteyn said. “We need to think really carefully about how we make this the best thing for the shopper.”