When I introduce myself at a cocktail party as the CEO of a consumer data company I can almost feel the other person’s anger begin to well. He or she is thinking ‘I didn’t know stalkers were so well dressed.’ The first minutes of the conversation often include a discussion, really a complaint, about retargeted ads. This is the industry term for the ads that haunt you after you click on a product. You looked at a pair of shoes on Nordstrom’s web site, and suddenly those shoes are appearing in ads you see across the web. You’ll see them when you visit CNN and Facebook, or even when you visit non-profit sites like NPR and PBS. My new detractor will rail about the deceitful behavior of “they” or “them.” They’re selling information about me, and I wish someone would stop them.
I take a deep breath before attempting to describe how my company is different. During the exhale I note that the host is playing music by Radiohead, a hot
indi-band. Its playing on Amazon Alexa speakers, which are probably listening-in on our conversation; the music stream is no doubt coming from Spotify which uses massive amounts of consumer data to curate its playlists; and I’m overhearing a neighboring couple talk about Glow, the hot new show on Netflix. The couple probably doesn’t realize the show came as a result of Netflix mining its massive user data trove about TV and video watching tastes.
Politicians are doing what politicians do by creating laws to stop companies from exploiting consumer data. The laws have names like GDPR in Europe, or CCPA in California. The laws are trying to lock-down consumer data with severe punishment for the abusers. FYI, punishing Google or Facebook with a $1 billion dollar fine would be as severe as me imposing a $5 tax on you. It’s not my nature to be a doubting Thomas, but the politicians deserve what’s coming. The data exploitation problem won’t be solved with new laws. The tech industry moves at light speed compared to the tortoise-like state houses. It’s Usain Bolt sprinting against grandma in the wheel chair. The Texas state house, governing 28 million citizens, only convenes every two years. I can assure you two years is plenty of time for the tech giants to conjure up and deliver awesome new deceitful methodologies between legislative sessions.
When I started my company, I began with a question. Is there a way to use consumer data for a positive good? When your doctor gives you a medical prescription you have some level of confidence the medicine is safe. When you read the bottle you’ll see dosage information, along with warnings about potential side effects. This information is the output of a data project, aka a clinical trial. Tens of thousands of people took the drug and the results were turned into a large data set. The data is being used for good purposes, like stopping headaches, pimples, or cancer, despite most drug trial panelists never reading the
agreement saying the drug company now owns data about every cell in his body. People seem to like Spotify and Netflix, which run their own versions of algorithmic clinical trials every day. They’re billion-dollar companies using and monetizing your data to sell more stuff. And, your Google search results are heavily dependent on “collaborative filtering”, which in layman’s terms means your search data and online behavior is being used by others. You’re right, the large tech companies are using your brain, free of charge, to build their products.
Where’s the privacy line? I’m talking about the line between exploiting people and empowering people. If I jumped to the side of my new cocktail party friend, he wouldn’t want me saying we’re against the personalization in Spotify. We like personalized music, right? And, surely he wouldn’t want me giving a thumbs-up to stopping all clinical trials of new drugs. We want personalized experiences; but we don’t want to become the product. I love that Spotify has learned my taste is 1970’s rock (apologies to my children for this dad reveal), but I wouldn’t want everyone who plays Elvis’s Blue Suede Shoes on Spotify to see a notation that says ‘Rob McGovern always listens to this song when he’s in the shower.’ Empower me with personalized music, privately please.
My company set-out to build a system that would empower consumers with better shopping results. Making a satisfying purchase is so much better than staring at the why did I buy it? dress in your closet. We started with a mantra that using the collective knowledge of millions of people is good, exploiting even one person is bad. A central thesis of our business plan was that in the long-run our company will be much more valuable as a pro-consumer data company rather than the flavor-of-the-month data exploiter. (Trust me, you don’t want to know about the number companies trying to sell me your location data.) Our early results have been overwhelmingly positive. Retailers and consumers like the idea of a pro-consumer data service.
We based our system design on three principles. First, our design had to be based only on aggregated data. When viewing an Amazon product you’ll also see “Customers who viewed this item also viewed these products”. You’ll see an array of products that are based on aggregated data. It doesn’t tell you the names of the other viewers, nor anything about them. The second principle is that we ‘d never possess personally identifiable information (“PII” in industry jargon). This means we’d never be able to identify a consumer in our system. Ever. The sneaky big players try gloss over this saying their identity information is encrypted. My take is that encryption is only as strong as the most powerful brute-force hacking system, and given the relentless pace of Moore’s law, encryption doesn’t stand a chance. We had to design a system that was so lock-tight that even we couldn’t view a consumer’s identity. The third principle was the system needed to be aligned with desires and wants of the consumer shoppers. We have to be a consumer first company. Retailers and advertisers using our data should focus on the tastes of the consumer, not their inventory clearance needs.
I can assure you, the faces of my engineering team looked like dogs hearing a high-pitched noise when I described what I wanted them to build. You want an aggregated data product delivering product results that match a consumer’s personal tastes, who we can’t identify? Well team, if it was easy it would pay minimum wage.
A critical step was finding a trusted party who could help. We need an intermediary who knew all adult consumers. We selected one of the largest credit reporting companies who have a complete identity graph of all US adults. Their business requires this knowledge in order to provide credit information to banks and credit card companies. We asked this company to create a persistent synthetic identifier for each adult consumer and provide it to PreciseTarget. We also signed an agreement for them to be our trusted match intermediary. If a retailer wants to provide data to us, they must pass it through the match intermediary. For example, if a retailer wants taste information on 100 customers, they must first submit the customer information to the match intermediary. The trusted intermediary that makes a request to us using the synthetic identifier. Our system is like a mixing bowl filled with 220 Million nameless grains of sand. Only the match intermediary can determine the name and other personal information about a granule. The intermediary is the only holder of the decoder ring; thus, it’s the trusted exchanger of information. We elected to use a credit reporting company for the principle reason they’re already trusted with the complete financial information on nearly every person. Plus, they have developed the most advanced information security procedures ever developed.
Our system is similar to Spotify or Netflix in that we cluster and aggregate people with similar tastes. Spotify uses this capability to predict the next song or playlist, where they are essentially saying “this is what people with your music interests are doing.” In our case, we are aggregating synthetic identifiers that have similar product tastes to predict what the aggregated data cluster would like. We’re not interested in making predictions for you, as we only make predictions for an aggregated set of identities. A retailer can learn about the tastes of one of their customers, which means the retailer wants to know the jeans taste of the aggregated cluster that includes one of its customers.
We believe using aggregated data with a pro-consumer intent is the right way to build a system. The data privacy crisis won’t be solved with new laws and regulations. A better idea is for companies to build systems where privacy is part of their fundamental system design.