Switch on your streaming service of choice or open the website for your preferred department store and a recommendation system is sure to kick in.
“You liked this TV series, so we think you’ll like this one!” Or: “As you’re looking at a pink linen skirt, think about buying these cream espadrilles to go with it!” They are key commerce drivers because they help customers see the products that they’re most likely to purchase. But they don’t fit neatly into existing machine-learning toolchains.
Some of the best-known recommendation engines are for content. YouTube’s eerie sense of what you might like to watch next is one example, and the ultimate champion of this game is TikTok: It’s deliciously addictive, precisely because the algorithms know what your little heart desires.
In some cases, however, there is more to a recommendation. For an online shop, there may be different margins for different product lines, and it has information that the engine itself does not; for example, people might not be buying ski gear now, but they damn sure will later in the year. Rubber Ducky Labs, a San Francisco–based startup, is looking to make it easier for teams to debug, analyze and improve their recommendation systems.
The team is working in a space that has a deeper trend: How do you know that the AIs are delivering good work? Increasingly, the algorithms do things that humans don’t fully understand — and without a feedback loop, it can get tricky.
“We call it like, operational analytics for recommender systems, because operational analytics is this emerging category where you’re plugging in on top of the data that someone’s already gathering, and helping them make sense of that and helping them make decisions and take action with it,” said CEO Alexandra Johnson.
Rubber Ducky Labs isn’t building new machine learning models; rather, it’s building systems to help people make better use of the information their existing systems have gathered. “We don’t build the machine learning models,” she said. “These companies already have existing machine learning models. We come in and plug on top of the data that they’re already gathering and help them to visualize it and understand it in a way that’s purpose built for recommender systems.”
In short, Rubber Ducky Labs is trying to make it easier for analysts to sift through the data acquired by machine learning models and turn it into something that companies can deploy effectively.
Recommender systems are commonplace, but they are also used, or set to be used, in areas you might not expect. “There are several big categories for recommender systems,” Johnson said. “There’s the traditional e-commerce category, streaming media — like your Netflixes and Spotifys, and all of the different streaming services that have popped up — consumer marketplaces, because you have to balance marketplace dynamics. And then the one that most people don’t realize is going to be really big is video games.”
It’s clearly a huge market that is ripe for disruption and further development. Rubber Ducky Labs has raised $1.5 million in seed funding, led by Bain Capital Ventures with participation from Cadenza Ventures and angel investors — including Brad Klingenberg (former chief algorithms officer at Stitch Fix), Patrick Hayes (co-founder of SigOpt), and Dave Aronchick (co-founder of Bacalhau and Expanso) — to help expand the company.
Rubber Ducky Labs is entering a space that is seeing a lot of movement. Just last year, Qloo raised $15 million to create an API-driven recommendation engine, and on the content side, Dable raised a $12 million round to expand content discovery.
“The next area for us to tackle after the analytics side of operational analytics is the operations side,” Johnson said. “Part of our vision when we go into improving recommender systems is tackling this business logic layer. When you ask people who’re working on recommender systems what’s the hardest part of their job, they’ll say that it’s the algorithm. There’s a lot of parts of machine learning engineers’ jobs that aren’t very fun to deal with and aren’t very fun to talk about. And this business logic layer on top of recommender systems is like one of those parts.”