Google tried to get ahead of web cookies in 2020, announcing that its Chrome browser would stop supporting them in 2022. That’s been pushed to 2023, but the end of web cookies is coming, and for experts in recommendation — the point of personalization — that’s fine.
Innovators are cracking the code on using artificial intelligence (AI) to read purchase intent from a few telltale datapoints, converting browsers to buyers just by figuring out what’s on their mind.
Alexandre Robicquet, co-founder and CEO at aptly-named recommendation engine Crossing Minds, told PYMNTS’ Karen Webster that in a quest for conversions, operators have long relied on the old-timey search engine optimization (SEO) strategy: “What can I get from this kind of outlaw world of cookies that existed before for me to just start an understanding of who that person is?”
The problem is that this bucketing of lookalikes into cohorts is “usually extremely detrimental because it makes you feel like you’re not special,” Robicquet said.
“It makes you feel like you have the same experience as another 33,000 people,” he continued. “There’s usually a ton of things that make these experiences feel like, ‘It’s not for me, it’s for who you think I am.’”
That’s a shortcut left over from Web 2.0 days, and we’re on the cusp of Web3 now. As Robicquet said, it is more useful to know someone’s recent browsing and purchase history than it is to know their age and location. He added that these improved data points greatly improve the output of Recommendation-as-a-Service platforms.
He painted a hypothetical situation where a store, either online or offline, welcomes a new customer — and then a genie appears offering two types of information about this stranger to make a sale.
“I can tell you where they live, their age range, their gender, things about the person, [or] I can tell you the three first pieces they look at and the one they have at home,” Robicquet continued.
It doesn’t take an expert to understand that what consumers buy and own is more revealing.
“Everybody would pay for that information, and yet 99% of the Fortune 500 have been brainwashed to think you need to have a ton of information about your users to understand who they are,” he said. “The answer is no; it’s not true.
“Instead of focusing on acquiring as much of those intrusive data about where you live, who you are and all, why don’t you focus on the first three clicks of what they’re looking at and what they’re spending time on? You end up with a ton of implicit feedback, completely anonymized, that gives you a sense of what is that person here for.”
AI-powered recommendations remove the old school idea of giving websites data just for the pleasure of using them, transforming digital engagement.
“People used to think that if they give you [a free online service], the only way for you to repay them is giving your data or something that they can sell,” he said. “That is revolting. We’re in 2022. You should not have to pay with your personal information, you should not have to pay with your time and your attention.”
See also: Google Analytics Violates Privacy Law
A Step Beyond Search
Digging into how AI is making these insightful jumps without a website cookie, Robicquet said the creepiness of searches that blindly follow us misses the point.
“Recommendation is one step further than search,” he said. “It’s what can I show you without you having to type what you’re looking for.”
The way Crossing Minds does it, a new user arrives on the site or in the store, and the platform serves the best matches based on behavior the AI is tracking. This can happen on an eCommerce site or with a sales associate in a store accessing the recommendations.
“This is the concept,” he said. “Send us the live interactions, and we’ll send you back the 10, 20, 64 items that you should probably showcase or display to your users.”
Webster pointed out that the same consumer may behave very differently depending on the type of items and purveyors they’re searching, which is a key weakness of cookies. Describing contextual commerce, where “session-based” recommendations shine best, Robicquet said the current state of recommendations needs an upgrade — badly.
“There is a caveat that most of the services do that — actually, I thought was revolting to some extent, where people tell you, ‘We provide recommendations for brand new users,’ or, ‘We provide recommendations that are cookie-less,’” but they really don’t, he said.
There’s been “an awakening around the importance of personalization online,” Robicquet continued. “Most of those business that provide recommendations now never put the words ‘personalized recommendation.’ Showing you the most popular, showing you the most recent, showing you the same as other people is still considered as the recommendation, but too many businesses fall into the trap of assuming that those are personalized.”
Calling for a new lexicon around personalization for the Web3 era, he added that the point of platforms like Crossing Minds is to proactively use real-time signals, not demographics.
“Start cherishing your first-party data,” he said. “By definition, first party is yours. It’s what happened on your website. Cherish a click, value a scroll. All those things are absolutely key. After that, the idea is if you start sending us this data … and if you start sending us your item catalog and what makes each of those items unique — what’s the image, what’s the material, what’s the story behind it — then we can start building models that learn from those patterns.”
Doing the Tech Cha-Cha
Taking recommendation engines to the next level with AI has implications beyond what to buy. Estimating that 3 billion hours are wasted each year due to the paradox of choice just around what to watch on TV, smarter recommendations could end a lot of movie night bickering.
“You absolutely know you want to watch a movie, [but] the average time for a user to pick a movie on Netflix is 19 minutes,” Robicquet said. That gets longer adding partners and friends, and a real-time recommendation engine could save a lot of relationships with streaming alone.
Conceding that he thought cookie-less commerce would be further along two years into the pandemic, Robicquet said it’s here and requires some perceptual realignment to catch on.
Likening tech and policy as a cha-cha — two steps forward, one step backward — he said, “I think in three years, people will start realizing that there are alternatives to what they’ve been offered until now, both on the solution basis for business and on the experience side. People will start asking for more when it comes to personalization and digitization in the eCommerce.”