A research system for 22 brands

1The problem

Each team runs A/B tests across its brands, but test ideas weren't always grounded in a shared understanding of who was using the site, what they were trying to accomplish, or where they were getting stuck. Across about 22 brands, I needed a repeatable way to turn the behavioral data we already had into research teams could actually use.

2What I built

A standard set of four research pieces per brand, built from the behavior data we already had:

PersonasGrouped by what people are trying to do on the site, not by demographics.

Journey mapsWhere each group gets stuck between landing and ordering.

Empathy mapsWhat users said, kept separate from what I inferred.

Friction listsProblems ranked so teams can turn them straight into tests.

Full sets are finished for three brands, along with test backlogs for three brands and full UX audits for two.

Source: finished sets dated July, August, and September 2026, stored in Confluence.

3The rule that makes it work

Every line in every deliverable is tagged with where it came from: Looker, Clarity session recordings, GA4, a live page check, or input from another team. If the data doesn't support a field, it says "not evidenced." I don't fill it with a guess.

A persona with blank fields beats a persona with made-up ones.The blanks tell you what to research next.

4What it turned up

While building personas for one brand, the data surfaced a visitor group we hadn't been accounting for: existing customers arriving on our authorized sales site looking for customer support.

Their behavior could look like friction in the sales journey, but they weren't actually there to buy. They had landed in the wrong experience for what they needed. That distinction changed how we interpreted their behavior — and the opportunity. Instead of trying to convert them, the better experience was helping them get to the right support destination.

Source: persona work using Looker exports, Clarity, live page checks, and input from the paid search team.

The personas also started feeding real decisions. One brand's persona set became a list of test ideas within about four weeks.

Source: persona set July 22, test ideas doc August 19, 2026.

5What's still missing

The system is strong at showing what people do, but behavioral data can't fully explain why they do it. The next step is adding focused interview rounds to validate the behavioral patterns, uncover motivations, and fill the gaps the quantitative and behavioral data can't answer.

Gap: no direct user interviews behind these sets yet.