← Back to work CASE STUDY

Turning Client Taste Debates Into a Reusable Research Asset

A synthetic customer panel, built once from real Jobs-to-Be-Done research, gave a marketing agency a permanent way to test creative work. Every campaign since has drawn on the same panel instead of restarting the argument from scratch.

A data-focused marketing agency with a national client reach had a recurring problem: clients second-guessed creative work based on personal opinion, leading to endless revisions. Affordable creative testing tools didn't exist at their scale. Rather than solve this project by project, we built a permanent asset: a custom GPT backed by JTBD-based customer personas that the agency could return to on every future campaign, not just this one.

"Clients used personal perspectives to provide feedback, leading to multiple revisions and wasted account hours."

The agency needed a more data-driven approach to creative testing that was both effective and affordable. Traditional testing tools were out of reach at their budget. Without objective data, every creative review became a debate based on taste rather than evidence, and the agency had no way to defend their work.

STEP 1

Build the Customer Archetypes

We selected a pilot brand and worked with the account strategist to understand the brand's customer base. Using this foundation, we built a Jobs-to-Be-Done map for each identified customer archetype: main job, sub-jobs, emotional and functional criteria, and circumstances.

STEP 2

Create the Synthetic Panel

Using the JTBD maps as a foundation, we worked with the LLM to develop two detailed personas per archetype, creating a synthetic panel of nine customers. The panel was grounded in real job structure, not invented demographic profiles.

STEP 3

Deploy the Custom GPT

The panel, persona maps, job maps, and additional guidance were built into a custom GPT the agency could return to on every new campaign, not rebuild for each one. The tool produced a percentage desirability score and customer quotes the agency could present to clients alongside the creative work.

04 — THE RESULTS
Fewer client revisions per project
Client confidence in creative decisions
9 Synthetic personas in research panel

Clients had greater confidence in the future performance of their ads, because the creative work was now backed by data-driven insights from a synthetic customer panel rather than gut feel. The agency experienced a measurable reduction in client-requested revisions and gained a repeatable, affordable process for validating creative work, one they could run again on the next campaign without rebuilding it from scratch.

A research asset built once and reused compounds in value every time it's used again, the opposite of a one-off deliverable that's spent the moment the project ends.
Grounding synthetic personas in real Jobs-to-Be-Done research, not invented demographics, is what makes the output credible enough for a client conversation, not just internal use.
The hardest part was never the AI. It was building institutional judgment into a form the team could reuse, the same problem Agency AIOS solves at the scale of a whole agency instead of one research panel.
Next case study

Four People, Running Like Eight

The same method applied at the scale of an entire creative team, not one research panel.