Personalization is a top agenda item for most brands, but data management technology still leaves something to be desired.
Machine learning could provide the answer, said Kabir Shahani, CEO and co-founder of Amperity, a Seattle-based startup that came out of stealth on Thursday.
Since raising $9 million in Series A in February, Amperity has been investing heavily in machine learning and AI, including active learning, a form of semi-supervised machine learning.
“Every brand has so much data, terabytes of Omniture data, click-through data – and there’s no place to put it,” Shahani said. “That’s where machine learning comes in to help manage identity across all of these discrete data sources.”
But there’s still an “enormously frustrating gap that still exists between the promise of technology and the reality on the ground, said Matthew Biboud-Lubeck, Amperity’s VP of consumer engagement strategy.
Biboud-Lubeck joined Amperity three months ago after seven years at L’Oreal, where he most recently served as VP of data strategy.
“I realized that I had the opportunity to help solve a problem I’d been trying to tackle for a decade with 2012-era technology and, frankly, a lot of manual lifting,” said Biboud-Lubeck, who was first introduced to Amperity when the company came to pitch his team at L’Oreal. “I also realized that it didn’t make sense to try and solve for it for one company when I could work to solve it for the marketplace.”
Rather than using an algorithm to apply existing rules to data sets, Amperity is teaching its algorithm to develop its own rules without human intervention based on information being brought on board in real time.
“The system is training itself,” Shahani said. “And that makes the overall model more effective.”
Brands integrate with Amperity the same way they’d integrate with a data management platform. As it continuously ingests raw data, Amperity applies machine learning to dedupe records, create identity matches and develop customer profiles that can be activated for targeting, segmentation, creating lookalike audiences, media buying – all the usual stuff.
What’s unusual is Amperity’s approach to identity in the open ecosystem.
“Systems like those of Facebook, Google or GitHub establish a universal identity first through authentication to open up an ecosystem of tools for that identity,” said Dave Fetterman, Amperity’s VP of engineering and a former mobile engineering exec at Facebook.
By contrast, Amperity creates identity matches “across disparate databases” within an organization, Fetterman said. In a sense, it’s the inverse of how a walled garden approaches identity.
Although Shahani didn’t have the authority to cite specific brands, he said Amperity is working with a number of large advertisers, including a well-known airline and several luxury retailers, that gave Amperity access to their data sets when the company was first launching in order to help train its models.
They did that because they want to take better advantage of their data and serve more personalized experiences, for which they need what Shahani called “real data science.”
“What I call ‘fake data science’ is really just analytics, building reports and garnering insights,” he said. “Real data science is actually using machine intelligence to do data science well.”
Founded in 2016, Amperity has a headcount of 40 with plans to hit roughly 100 employees by late 2018. The company is also considering raising more capital before the end of the year to bankroll a larger sales team and accelerate its sales funnel.