Buying Data Science, Instead of Hiring it

A new approach to solving the sparse customer data problem

by Rob McGovern, CEO of PreciseTarget

It’s likely that the early 2020’s will be memorialized as the data science era of retail. Fashion shows, merchandising, and client loyalty is being replaced by algorithms, massive data sets, and machine learning. How important is this change for brands? It’ll be as important as off-shore manufacturing, mobile commerce, and social media marketing.

The advertising industry provides a close parallel which is instructive to brands. In the ‘old days’ (pre-2010) advertising was programmed almost entirely by humans. Agency creative departments developed ad campaigns, slogans, brand imagery, and taglines. Over the past 10 years nearly all of this has been replaced by a digital advertising ecosystem run by algorithms, and some of the big data companies in the center of it, like LiveRamp, TradeDesk, Epsilon, and Acxiom, are essentially large computer systems which happen to have employees.

We’re entering the Programmatic Retail era, which will be equally transformational. In advertising, the battle ground has shifted from the best creative to the best data and algorithms. The same will happen in retail, with the winners and losers determined by who possess the best data and partner with the best data scientists. Apparel brands enter the battle at a disadvantage because they typically possess sparse data about their customers. Most brands know customers by one or two transactions per year, often without even knowing their name. The bad news is that brands need far more data. The good news is that brands can now purchase new classes of off-the shelf data to level the playing field.

Data science can be purchased?

If your house includes young adults you’ve almost certainly heard of TikTok. While this new video sharing platform gained attention in the political arena, it’s more notable for something else: it has broken new ground in using AI to make recommendations. TikTok users describe it as engaging — okay, addictive — user experience. They say its uncanny how brilliantly it recommends videos users love, keeping them watching for hours at a time. TikTok is using a new break-through data science model that changes the game in user experiences. In reality, this data science concept is available for retail brands, without the requirement of having an in-house data science team.

The previous generation of personalization was based on the social graph. Facebook, Instagram, Twitter, Pinterest, and Snap aggregate users by social connections. The premise is that you’ll like the same content and products as your friends. Like all technologies, the social graph was vulnerable to the next big thing. TikTok is pioneering it, clearly leap frogging the social graph. They’ve implemented an “interest graph”, which scientifically clusters people who share the same interests rather than clustering consumers based on their social connections. They use advanced machine learning to analyze videos at scale, creating an intelligent machine that can understand the content of a video. To be sure, machine learning a video without spoken words is pretty incredible. Does it work? More than a few users have described it as the crack cocaine of web surfing.

How does this apply to brand marketing?

PreciseTarget has developed an interest graph for apparel, footwear, and other soft goods categories. Much like TikTok, it receives ‘signals’ from the user community. The company has received over 5 billion opt-in retail transactions, and it uses machine learning to understand a consumer’s fashion interests across the entire soft goods universe.  Rather than being bound by a single brand data set, it takes a cross-brand view of the world.

The system has learned that consumers do not have strong affinities to individual product brands, but it has learned that consumers are very consistent in purchasing patterns. For example, the jeans and dresses in your closet may be from different brands although they likely share many commonalities. These commonalities would be classified by a machine as an “interest”. Your jeans interest may be straight fit, $50 to $70, in dark colors. Your dresses interest might be A-line style, cotton fabric, $80 to $100, machine washable.  This knowledge is extremely valuable in personalizing web, ad, and email experiences, or ensuring that assortments include the products that reflect their customers’ interests.

Competing in the new era: Mastering data and algorithms

In order to compete in the current era, brands must address two challenges. First, they need mountains of data that go far beyond an existing customer data base. If your data only represents one or two items in your customer’s closet you only partially know your customer. Second, in order to harness the power of massive data sets, brands need access to data science expertise with experience in machine learning and artificial intelligence.

PreciseTarget has created an off-the-shelf solution to address this two-pronged problem. It has created an interest profile on each consumer, which articulates their taste profile in every category. These profiles provide new insights on virtually every US adult shopper, and they can easily be appended a customer file, CRM, or CDP. The idea is to enable brands and retailers to leverage a vast cross-merchant data set, providing expansive customer insights data unattainable if sellers are bound by only their own data. The data is exchanged using CCPA compliant pseudonymized identities using a trusted third party, with PreciseTarget having no way of identifying any customers. 

It’s precomputed data science, enriching you with actionable data for the new era.

Interested in an informational meeting or a demo of our products? Contact info@precisetarget.com