Preferabli, Co-Founder and CEO: All Our Algorithms Are Homegrown

Preferabli, Co-Founder and CEO: All Our Algorithms Are Homegrown

(US and Canada) Pam Dillon, Co-Founder and CEO of Preferabli, speaks with Robert Lutton, Vice President, Sandhill Consultants, about the driving thought process behind the platform and how it fulfills its novel technology requirements.

Dillon describes Preferabli as a solution for wine, beer and spirit businesses to market and sell to consumer preferences. It is driven by an assumption that people like to buy things that they like the taste of, and with an overwhelming variety of wines, beers, and spirits available, it was a problem worth solving.

“We built the platform from the very beginning to imagine sensory consumer products broadly. So, from the very earliest days, I had imagined all of the sensory consumer products. We started with wine and spirits — full disclosure, because I love them. The team loved them. It's better to work with something that you deeply love because you are going to be around it for a long time,” she says.

Because Preferabli was imagined as a platform to help merchants sell more precisely to their consumers, “We imagined to ascertain the presence and the absence of hundreds of different characteristics that define preference,” Dillon further explains. “So, it's the core of our technology and our software. Where a merchant is building its journey for its consumers is really important, and it's really important for us to meet them where they are in their journey to give them the modules that they need to build their journey.”

The modules are in three basic groups of software. First, onboarding software brings someone into a personalization platform. It is expert with respect to a specific product. This can be used to not only onboard existing customers into a platform, but also new customers.

The second group is about one-to-one personalization, beyond segmentation, collaborative filtering, etc. “It was always one-to-one, one person at a time, one product at a time,” Dillon notes. “The largest group of modules sits in that second group because they are the most valuable, not only to sales and marketing but lifetime consumer value.”

The third group — platform technologies — allows businesses to focus on what they do well. It includes platforms for marketplaces and elements for platforms, from labels to technical information. It caters to larger or medium-sized, and even smaller-sized, businesses that are not looking to build a platform on their own.

Regarding the application of AI and ML, Dillon says that the technology was almost non-existent in the wine and spirits industry before Preferabli (formerly known as Wine Ring) entered the scene. And so, everything had to be created from scratch. While it was difficult, it was also an opportunity, she points out.

There are two kinds of recommendation systems — collaborative/crowdsourcing and content-based — that are rooted in data. Most existing recommendations were collaborative, using one person's opinion as a reference point for another person's recommendation.

“That made no sense,” Dillon continues. “We knew we had to work toward something that was content-based. The center of our intellectual property is content-based with our own data. These were algorithms that were homegrown, each line of code written by us. This is not an off-the-shelf neural network because, even today, the neural networks don't approach the problem in the way that we wanted to solve it.”

She further explains that other recommendation systems used data reference points only, and they did not consider individual preference profiles. They rationalized data patterns on their own. “Starting with the preference profile not only was novel, but it was also extremely useful because it allowed us to understand preference from the get-go, which made all of the recommendation technologies even more accurate,” Dillon concludes.

More from American Express

Related Stories

No stories found.
CDO Magazine