case_study[03] · ABLab · Internal product · 2026

The CRO Dashboard, built for the team.

I built and shipped an internal A/B-testing platform for my growth team at Lenus. It replaced a weekly spreadsheet routine and became the team's testing workspace.

Initiator, designer & sole developer Shipped through Legal & Security review In daily use · 4 designers 2026

The ABLab Running tab: live experiments, smart alerts and recent activity

The Running tab, every screen in this case is the real, shipped toolablab · live

ChallengeA/B results scattered across three tools, shared as spreadsheet screenshots, learnings lost between markets.

My roleInitiator, product designer and sole developer, no engineering team behind it.

Team & usersBuilt for the Lenus growth & design team; results shared with key account managers and coaches.

Timeline2026, began as one question to Claude, evolving continuously in weekly use since.

ToolsClaude (design & code) · Firebase · Vercel · Google sign-in.

ConstraintsHad to fit the exports and tools the team already used, and pass Legal & Security review to run on real client data.

StatusShipped and in daily use by 4 designers across the Nordic markets.

OutcomeHundreds of tests run and decided through it; one shared language for hypotheses across the team.

01 · Context

The workflow before: three tools and a screenshot.

To report one test, I pulled data from several tools, rebuilt it in a Google Sheet and shared a screenshot. Every designer repeated the same routine.

KAMs received a spreadsheet screenshot when they needed a verdict.
Designers stitched the same exports together every week.
Old results stayed buried in a database that was slow to browse.
Four markets had no shared view of the hypotheses worth pursuing.

02 · Origin

It started as one question to Claude.

I was tired of rebuilding the same report. One small question to Claude turned into a product:

01"Can you make the results from this Google Sheet more visual?"
02If Claude could read the sheet, I could skip it and import the CSV directly.
03Each annoyance in my own workflow became the next feature.

ABLab's task board: seven tasks across the team in to-do, in-progress and done, each tagged with a coach and a due date

Every roadblock became the next feature: the weekly routine, inside the tooltasks

03 · Design decisions

Codify how the team already decided.

I encoded the Bayesian model, Monte Carlo draws and 90% threshold the team already used. The method stayed familiar; the result became easier to act on.

Three stages, one funnel

Each test is scored across the full funnel, so a sign-up win cannot hide a weaker close rate. Fitbroidit proved why that matters.

Hypotheses as first-class objects

Tests are tagged by hypothesis, so the team can answer a simple question quickly: "Have we tried this before?"

Written for the people who decide

One click gives the KAM a plain-language conclusion they can forward as-is.

The Patterns view: behavioural hypotheses ranked with win rates and linked tests

Patterns, where finished tests become team knowledgehypotheses[12+]

AI CRO insights: verdict, forwardable conclusion and suggested next experiments

The verdict, the summary and the next three testsai_insights

04 · Iteration

Built by its first user, then bent by its second.

I used ABLab every week. The CSV template and VWO import came from my own friction; ABC tests arrived after a colleague asked for them. That request made it feel like a team product.

The frictionWhat shipped because of itDriven by

Explaining the export format for the third time to the third person.The CSV template, a format anyone can fill without asking me.user_zero

The single most repetitive step of the old routine: assembling VWO numbers by hand.One-click VWO import, the stitching workflow, deleted.user_zero

A designer hit the two-variant ceiling on a test that needed three.A/B/C experiments, a structural change to the statistical model, requested, built and shipped.colleague_request

Test templates: eight CRO patterns with fill-in-the-blank hypotheses

Templates, friction I felt, removed for everyone after mefill_in_the_blank

05 · Shipping inside a company

From a side deploy to an approved tool.

I first shipped ABLab quietly on Firebase and Vercel. Once people wanted to use it with real client data, I took it through Legal & Security. The approval badge started as a joke for my teammates: we're actually allowed to use this now.

06 · Adoption & outcome

From my workflow to the team's.

100s

of tests run & decided through ABLab

designers using it in daily practice

person to design, build and ship it

Four designers still use ABLab. Hundreds of tests have run through it, results are shared directly from the tool, and finished experiments feed a library the team can search.