The Company That Reverse-Engineered America's Closet

Most retail advertising is a confident guess. A brand decides its customer is a woman, 25 to 45, in a mid-size metro, and buys her attention by the thousand - knowing full well that a large share of the people it just paid to reach would never have bought the jacket anyway. PreciseTarget, a data-science company tucked into an office on Wisconsin Avenue in Bethesda, Maryland, was built on the opposite instinct: that the useful question isn't who a shopper is, but what they actually like to wear.

To answer it, the company assembled something unusual - a profile of the retail buying taste of roughly 99% of U.S. adults, inferred from more than five billion SKU-level transactions flowing in from thousands of brands and retailers. Where a demographic model sees a zip code, PreciseTarget's system sees, in effect, what is already hanging in your closet. And it sells that difference to fashion and apparel brands that are tired of paying to reach the wrong people.

01 / The founderA third act, aimed at your wardrobe

PreciseTarget is the work of Rob McGovern, an entrepreneur best known for creating CareerBuilder, the job-search company that went public in 1999 and was acquired for around $200 million in 2000. His pattern is consistent: take a messy human decision - finding a job, buying clothes - locate the data exhaust it leaves behind, and turn that into a product.

This company started life in 2013 under a different name, Cobrain, and rebranded to PreciseTarget in 2016. The old handles still linger online as "@cobrainco," a small fossil of the pivot. The retail data products didn't ship until 2019. That gap - years of quiet infrastructure work before anything looked like a launch - is the least glamorous and most honest part of the story.

The technical spine matches the ambition. Turning five billion transactions into individual taste profiles is a heavy-machinery problem, and the company leans on the tools you'd expect at that scale - distributed data processing, gradient-boosted and deep-learning models, and cloud infrastructure to run daily assortment feeds against a constantly moving picture of what people are buying. The output isn't a dashboard for its own sake; it's a prediction that has to survive contact with a real ad campaign.

02 / The productTaste, not demographics

The core asset is what the company calls Consumer Taste Profiles - individualized read-outs of apparel and footwear preference built from real purchases and refreshed against daily assortment feeds from brands and retailers. Stitched together, those profiles form a Retail Audience Taste Graph: a map of how shoppers, brands and products relate, so a retailer can find audiences whose taste genuinely fits its assortment rather than its imagined demographic.

On top of that sit two working products. ConsumerInsights, launched in 2021 and pitched as "data science in a box," gives a brand a deeper read on both its existing customers and its best acquisition targets. AcquisitionAI, launched the same year and billed as the first AI-based customer-acquisition product for retail, studies the customers a brand has just won, then automatically builds audiences that resemble the most valuable ones - and pushes them straight into Meta, Google and other ad platforms.

03 / Why it worksFewer wrong people

The interesting thing about PreciseTarget isn't that it promises more reach. It's that it promises less waste. In a market drowning in cheap impressions, subtraction is the product - the value is in the people the system tells you to skip. The company has reported conversion improvements of 50% or more across channels for participating retailers, with claims of a 50% to 100% lift in ad performance and purchase conversion.

There's a second, quieter argument in the design. The profiles are built to be privacy-safe - taste is inferred from aggregated transaction data rather than by tracking individuals around the web. As third-party cookies crumbled and platforms tightened their rules, that approach aged well: a way to target that never depended on following anyone.

What a brand does with this in practice is mundane in the best way. A D2C apparel label runs its existing customer file through ConsumerInsights and learns which segments actually drive repeat purchases, not which ones merely bought once. It hands AcquisitionAI the list of customers it won last month, and the system builds a fresh audience of shoppers who resemble the profitable ones - then delivers that audience to Meta and Google so the next campaign starts from a better guess. The work that used to be a media buyer's intuition becomes a data feed. For a category where a single high-lifetime-value customer can be worth many times a one-and-done buyer, sharpening that input at acquisition is where the money is.

04 / The marketRenting a taste graph nobody wants to build

A brand could, in theory, assemble its own view of customer taste. In practice, almost none can - the data spans thousands of retailers no single company sees. That's the wedge PreciseTarget sits in. It competes with retail and audience-data platforms like Bluecore, Amperity and Zeta Global, and, more pointedly, with the native lookalike-audience tools built into Meta and Google. Its bet is that a purpose-built taste graph across 3,000-plus brands beats a platform's guess drawn only from its own walled garden.

The business model follows the logic. PreciseTarget licenses its taste data and audience products to apparel, footwear and fashion retailers and D2C brands, enriches their existing customer data, and supports the media buying itself - selling improved audiences and campaign performance rather than raw media. A 2021 partnership with Equifax paired its taste data with a much larger audience footprint.

05 / The honest caveatsWhere the pitch gets tested

The claims are the company's own, and the conditions matter. Taste-based targeting leans on a specific category - apparel and footwear - where past purchases are genuinely predictive of future ones. Stretch it into categories where taste is thinner or buying is need-driven, and the edge narrows. A brand with a tiny customer list, a brand-new aesthetic, or a shopper base outside the data's coverage gives the model less to work with. And a 50-100% lift is a range, not a guarantee - it depends on the baseline you're measuring against and how sloppy the old targeting was.

What's copyable here, for anyone building a data business, is the framing. PreciseTarget doesn't sell a feeling of sophistication. It sells one measurable outcome - the wrong people you no longer pay to reach - and it built years of unglamorous infrastructure to be able to make that promise credibly. The team stayed lean, roughly 30 people, pairing retail veterans with machine-learning engineers rather than chasing headcount.

06 / Where it sits nowA quiet piece of retail plumbing

The customers are apparel, footwear and fashion retailers and the D2C brands chasing them, along with the agencies and media teams that buy on their behalf. The data underneath spans roughly 99% of U.S. adult shoppers and draws on more than 3,000 brands - a footprint wide enough that a mid-size label gets a read on its market it could never afford to build alone. That breadth is the moat: not a clever algorithm anyone could copy, but a data set that took years of retailer relationships to accumulate.

PreciseTarget isn't a consumer name and doesn't try to be. It's the kind of company that shows up as a line item on a growth team's stack and a lift on a performance dashboard - infrastructure that a brand rents instead of builds. Backed by around $32 million from investors including Moonshots Capital and the University of Maryland's Dingman Center, it has kept its aim narrow: know what America likes to wear, and sell that knowledge to the people trying to sell America clothes.

If you sell apparel and you're still buying "women 25-45" audiences, the company's existence is a mild provocation. The question it keeps asking - would this person actually wear your brand? - is the one most ad targeting skips. PreciseTarget's whole business is refusing to skip it.