Built by analysts who ran out of patience
We built AB-Labz because we kept doing the same manual work — switching between Python scripts, Jira, Confluence, and Excel — for every experiment. There had to be a better way.
The problem we kept running into
Every team we knew ran A/B tests across a fragmented stack: one tool to manage hypotheses, another to run statistics, a shared doc to write conclusions, and a spreadsheet to track wins and losses over time. Results got lost in Confluence. Conclusions were rewritten from scratch every time. Nobody had a clear view of whether the testing programme was actually working.
The deeper problem: most analysis was done incorrectly. Ratio metrics were treated as simple proportions. Outlier handling was inconsistent or absent. Multiple testing correction was applied blindly or not at all. The tools available were either too generic (Excel), too academic (R packages), or too expensive for most teams (enterprise A/B platforms).
AB-Labz is our attempt to solve this: a platform that handles the full lifecycle of an experiment — from hypothesis to archived result — with a statistically rigorous engine underneath and a workflow designed for real product teams, not just enterprise data science departments.
The platform was never a planned commercial launch with investors and a roadmap drawn on a whiteboard. It grew gradually, experiment by experiment, as we solved our own problems at work. We use it ourselves — every day, in real production environments. We know how it behaves not because we designed it in Figma, but because we live with it in the field and build features when we hit our own pain points.
What we believe
Statistics should be correct by default
If the tool makes it easy to do the wrong thing, teams will do the wrong thing. The engine enforces correct approaches — you don't have to know to ask for delta method on ratio metrics, it just happens.
Process and analysis belong together
Separating "where we manage experiments" from "where we analyze them" creates friction and lost context. Everything should live in one place, from the first hypothesis to the final archive entry.
Small teams deserve good tooling
Enterprise A/B platforms cost more than a small team's entire analytics budget. We've made the core platform free for solo analysts and priced the corporate tiers to be accessible to companies that are actually doing serious testing.
We're early — and building in public
The product is actively evolving. We ship updates regularly and genuinely want feedback from analysts who use it for real experiments. If something's wrong, broken, or missing — tell us.