Retail media · Case 03

The portal the evidence said not to build

Role
Design Manager, paired with a product counterpart — the study run by the two of us, on our own initiative; the method design, the sample, and the external phase were mine
Timeline
~2–3 months (internal + external phases)
Result
bet redirected before the build · portal cut to less than a quarter of the original scope · evidence became a gate in the prioritization process

01 — My role

Retail media is one of the company’s central B2B revenue bets: getting the industry’s biggest advertisers to invest, keep investing, and invest more.

The area’s leadership came to tech with the solution already designed, an on-demand platform where the industry would see:

  • what was planned and what was live
  • campaign results and assortment data
  • as a bonus, it would free up the commercial team’s time

My role: build this.

The case, in the order it happened
02What was at stake

Months of a squad, the top players, and a risk no one had put on the table.

Building the portal meant months of a squad on a product with no adoption validation whatsoever, touching the relationship with the advertisers who concentrated most of the category’s investment. And there was a third risk, invisible in the brief:

03From brief to the right questions

I didn’t accept the brief as a feature request. And I didn’t hand back an opinion-based “no” either.

I turned the pre-baked solution into three testable questions:

1. Does it need to be digital?
2. Would there be adoption to justify building it, and maintaining it?
3. What’s the industry’s real pain?

And I designed the study in two waves, to answer them with evidence:

Wave 1 · in-house
Commercial, planning, operations, brand studio, leadership, plus a benchmark with a global retail media operator. A front run by my counterpart, on the design I did.
Wave 2 · the industry itself
With me leading: eight large advertisers, deliberately segmented by media maturity.
04The study, with the craft of someone who signs it

A “no” to an executive doesn’t hold up on opinion. It holds up on evidence.

Leadership’s questions stayed as the study’s spine: tested, not replaced. The sample, designed by maturity, armored any cut against accusations of bias. And every finding stayed traceable to its source quote.

this isn’t a case of directing a team: my counterpart and I ran the study with our own hands. the seniority is in the method design and in the stakeholder negotiation from data.

The study plan that held up the “no”:

From a pre-baked solution to the right questions: the study planThe sample was designed so no conclusion could be accused of only listening to whoever confirms the hypothesis.From a pre-baked solution to the right questions: the study planThe sample was designed so no conclusion could be accused of only listening to whoever confirms the hypothesis.01 · INTERNAL WAVESix fronts heard inside the companyoutput: process pain ≠ digital product painsales / account management · category managementmedia planning and operations · brand studiothe area's own leadership02 · EXTERNAL WAVEEight advertisers, sample designed by maturityfour categories · from the most sophisticated to beginnersexternal benchmark: a global retail media operatorplus a former executive of the previous model03 · SYNTHESISDiagnosis and prioritizationevery key message anchored to a traceable quote in the repository04 · OUTPUTIdeas and solutionsnot the platform that arrived pre-bakedThe three principles that held up the "no"the executive's hypotheses became the study's backbone (tested honestly, not swapped out) · sampling by maturity shielded against selectionbias · every finding traces back to its sourceanonymized: fronts described by role · no brands, names or internal toolsFrom a pre-baked solution to the right questions: the study planThe sample was designed so no conclusion could be accused of only listening to whoever confirms the hypothesis.From a pre-baked solution to the rightquestions: the study planThe sample was designed so no conclusion could beaccused of only listening to whoever confirms thehypothesis.01 · INTERNAL WAVESix fronts heard inside the companyoutput: process pain ≠ digital product painsales / account management · category managementmedia planning and operations · brand studiothe area's own leadership02 · EXTERNAL WAVEEight advertisers, sample designed bymaturityfour categories · from the most sophisticated to beginnersexternal benchmark: a global retail media operatorplus a former executive of the previous model03 · SYNTHESISDiagnosis and prioritizationevery key message anchored to a traceable quote in therepository04 · OUTPUTIdeas and solutionsnot the platform that arrived pre-bakedThe three principles that held up the "no"the executive's hypotheses became the study'sbackbone (tested honestly, not swapped out) ·sampling by maturity shielded against selection bias ·every finding traces back to its sourceanonymized: fronts described by role · no brands,names or internal tools
What this diagram proves: the "no" was armored by design (waves, a maturity-based sample, traceability), not by opinion.
05What the evidence showed

The pain wasn’t access to data. It was analyzed insight.

The top players didn’t even use the portals they already had from other retailers: “I don’t miss it”; “I have access and no bandwidth to analyze it.” What they were asking for was a reading that would help them decide, and defend budget inside their own companies.

Selling the direction happened in key-messages sessions at each phase, framed as questions instead of verdicts: “do we really want to show all the results to everyone?”

The matrix that turned into the decision, with the original bet in the nearly empty column:

Maturity × how results actually get consumedThe artifact that turned the decision: the pre-baked bet sat in the near-empty column.Maturity × how results actually get consumedThe artifact that turned the decision: the pre-baked bet sat in the near-empty column.BRAND PROFILESELF-SERVICE PORTALthe original betREPORT VIA ACCOUNTINSIGHT DISCUSSED IN ACALLthe real painHigh maturitytalk KPIs as equalsaccess, no bandwidth to useitMid maturitydigital team recently created—Low maturityneed the basics translated—dominant (the real pain)current behavioroccasional / no real use—absent“We don't miss it —sometimes it's easier tojust ask our accountrep.”— global food brand“I have platform accessand can do anything — Ijust have no one toanalyze it.”— global beverage brand“Looking at the detail ishard — that's what theweekly calls are for.”— frozen food brandanonymized: no brand names · quotes lightly edited to remove identifiers · sizes and values omittedMaturity × how results actually get consumedThe artifact that turned the decision: the pre-baked bet sat in the near-empty column.Maturity × how results actually getconsumedThe artifact that turned the decision: the pre-baked bet sat inthe near-empty column.High maturitytalk KPIs as equalsSELF-SERVICE PORTALaccess, no bandwidthto use itREPORT VIA ACCOUNTINSIGHT DISCUSSEDIN A CALLMid maturitydigital team recently createdSELF-SERVICE PORTAL—REPORT VIA ACCOUNTINSIGHT DISCUSSEDIN A CALLLow maturityneed the basics translatedSELF-SERVICE PORTAL—REPORT VIA ACCOUNTINSIGHT DISCUSSEDIN A CALLdominant (the real pain)current behavioroccasional / no real use—absent“We don't miss it — sometimes it'seasier to just ask our account rep.”— global food brand“I have platform access and can doanything — I just have no one toanalyze it.”— global beverage brand“Looking at the detail is hard —that's what the weekly calls arefor.”— frozen food brandanonymized: no brand names · quotes lightly edited toremove identifiers · sizes and values omitted
What this diagram proves: the original bet sat in the nearly empty column. The dominant pain, insight analyzed in conversation, wasn't even software.
06The redirection

I redirected the bet before the build.

The investment went where the real waste of hours was: automating execution and proof-of-delivery, and recurring data with a layer of human analysis. The platform vision became automated reporting, at less than ¼ of the original scope.

The bet, before and after:

The bet that arrived pre-baked — and what the evidence pointed toThe implicit premise was that the brands' pain is access to data. Discovery came back with a different question.The bet that arrived pre-baked — and what the evidence pointed toThe implicit premise was that the brands' pain is access to data. Discovery came back with a different question.BEFORE · THE PRE-BAKEDSOLUTIONAn on-demand platform for brandssee what's planned and what's livesee campaign resultsassortment and pricing datasecondary hypothesis: free up the salesteam's timeDISCOVERY CAME BACK WITHIs the pain about data — or aboutdeciding?the question that became the casePRODUCT PRIORITYAutomating execution and proofof deliverywhere the real hours were being burned —this becomes a tech buildSERVICE, NOT SOFTWAREAnalyzed insight + a humanrelationshipthe customer's real job: to decide, and todefend a budget internallyWHAT SURVIVED OF THE PLATFORMRecurring data + an automatedreportresidual scope — a fraction of the originaltestedRESULTMonths of squad time saved on a product with no validated adoptionplus the opportunity cost re-prioritized — and the discovery became the evidence bar for every bet that followedinternal tools genericized · no advertiser brands · no absolute figuresThe bet that arrived pre-baked — and what the evidence pointed toThe implicit premise was that the brands' pain is access to data. Discovery came back with a different question.The bet that arrived pre-baked — and whatthe evidence pointed toThe implicit premise was that the brands' pain is access todata. Discovery came back with a different question.BEFORE · THE PRE-BAKED SOLUTIONAn on-demand platform for brandssee what's planned and what's livesee campaign resultsassortment and pricing datasecondary hypothesis: free up the sales team's timetestedDISCOVERY CAME BACK WITHIs the pain about data — or about deciding?the question that became the casethe evidence pointedPRODUCT PRIORITYAutomating execution and proof of deliverywhere the real hours were being burned — this becomes atech buildSERVICE, NOT SOFTWAREAnalyzed insight + a human relationshipthe customer's real job: to decide, and to defend a budgetinternallyWHAT SURVIVED OF THE PLATFORMRecurring data + an automated reportresidual scope — a fraction of the originalRESULTMonths of squad time saved on a product withno validated adoptionplus the opportunity cost re-prioritized — and the discoverybecame the evidence bar for every bet that followedinternal tools genericized · no advertiser brands · noabsolute figures
What this diagram proves: the redirection was a product swap, not a cut: the build went where the real waste was.
Months of a squad avoided
on a product with no validated adoption, plus the opportunity cost reprioritized
Evidence became a gate
bets of that magnitude now require discovery before build: a formal prioritization gate