Improving discovery for NFTs will amplify digital creators and marketplaces

Four years ago, my team set an ambitious goal: to build the smartest recommendation platform on the market.

Since then, our algorithms have empowered companies of all kinds to provide tailored product and content recommendations, all without using cookies or invasive strategies.

Our algorithms are so precise, that they can leverage only a few on-site actions from a consumer to predict what else they’ll like. Ostensibly, that would seem to be the happy ending.

But NFTs are different.

The past two years have shown us that NFTs’ influence is growing, and fast. NFTs hold a tremendous amount of potential not only for buyers but for
artists as well.

The purchase of an NFT doesn’t only reflect a buyer’s taste, it also tells us what they think might be a prudent financial investment

Yet, that potential isn’t being fully tapped. The inability to predict what NFT buyers want represents one of the main issues preventing NFTs from going mainstream and digital artists from being amplified. The challenge lies in solving the “cold start” problem for NFTs, where it’s nearly impossible for brands to provide accurate recommendations for new and anonymous users.

The challenge for NFT marketplaces is to provide accurate recommendations to new and anonymous users.

First and foremost, shopping for NFTs isn’t like shopping for anything else we usually buy online.

Thousands of new NFTs are bought and sold every week, buyers use multiple wallets, and they tend to be completely anonymous.

Each of these attributes contradicts what we expect with online shopping. We expect a finite number of pairs of jeans, for instance, not a never-ending supply of choices. We tend to link one credit or debit card, and while most online consumers today are, in fact, anonymous, there’s still usually the option to let a site remember your preferences.

But NFT marketplaces don’t know who their customers are and what they’re interested in. Customers struggle to parse a seemingly infinite list of available options and find what they really want, and creators struggle to be discovered. This reflects the cold start problem, but on an exponentially larger scale.

It’s also important to understand that prediction models in use today assume that there is a clear target metric to optimize for. Recommendation models aim to promote products and content that will increase a specific metric, like engagement or conversion.

An NFT purchase could reflect the buyer's aesthetic taste, or what they think is a prudent financial investment. Or both.

Typically, it’s safe to assume that such metrics are directly correlated to an individual’s taste. So, it follows that the simpler the connection between a user’s behavior and the metrics, the simpler it is to build an efficient recommendation engine. For most online marketplaces, the main target metric is profit, so the more an algorithm can link profit to an individual’s historical patterns and behavior, the more efficient it will be.

The problem is, the purchase of an NFT doesn’t only reflect a buyer’s taste, it also tells us what they think might be a prudent financial investment.

Most online shoppers don’t tend to select a pair of jeans based on its potential resale value 10 years down the line. Many NFT buyers are, however, considering ROI when they make a purchase and potentially weighing that return more heavily than their appreciation of the art itself.

Put simply, with NFTs, it’s uniquely difficult to discern the motivation behind a purchase. And given the hype around NFTs today, most recommendations would likely end up being weighted more toward investment-driven choices rather than preference. That would lead to an exponentially larger and more abstract amount of data for an algorithm to take into account.

Ultimately, a successful NFT recommendation algorithm would have to be able to parse through the nuances of not only what a buyer’s preferences are right now, but what their future intentions might be as well.

These parallel purchasing motivations add yet another layer of complexity to the cold start problem and ultimately lead to less potential for intuitive discovery when it comes to NFTs. This is a major factor keeping them from moving beyond a relatively niche audience; one where digital creators aren’t prioritized celebrated and compensated as they should be.

If a recommendation algorithm can ensure that buyers can meaningfully discover NFTs they love, or think there’s investment potential in or both, and if it can keep buyers coming back for more, then artists will reap the benefits for years down the line.

It’s not surprising that the rise of NFTs has led marketplaces and creators into the cold start problem currently encountered by other industries. After all, it’s a new technology, and that means new problems. But it also means that there are new, exciting possibilities.

Recommendation algorithms pose a huge opportunity for the growth and expansion of NFTs. Many people have speculated that NFTs will reach their potential once they’ve been regulated, or once online NFT ownership is blended with offline use cases, but investing in recommendations — and subsequently meeting the immediate needs and desires of consumers — is the first step to reaching that potential.

Originally published on TechCrunch, 3/25/2022.

Alexandre Robicquet
CEO & Co-Founder

Alexandre earned his first two Master’s degrees in Mathematics and Machine Learning at the age of 23 from ENS Paris Saclay. Alexandre had his work published at the age of 21 and held two 4-year positions as a researcher under Pr. Sebastian Thrun (founder of Google X) and Pr. Silvio Savarese. He received a third Master’s degree in Artificial Intelligence and started on a path for a Ph.D but chose instead to work on building Crossing Minds full-time.