Around your experiments — not instead of them
AB-Labz doesn't replace your split system. It sits next to it for planning, analysis, monitoring, and archive — connected to the data you already collect.
Where AB-Labz sits
Three parts. You keep the first. You connect through the second. AB-Labz is the third.
Split & data
Your app and split system stay as they are — GrowthBook, Statsig, or something you built. Events land in your warehouse, and you prepare an experiment data mart the way you already query other product tables.
Nothing here changes. AB-Labz doesn't touch it.
Connect
Push a CSV per experiment on a schedule through the API, or run the self-hosted Runner so files stay on your servers and only aggregates leave. Manual upload is always available on the Solo plan when you need a one-off.
Pick one path. Both feed the same workbench.
Workbench
This is where the team plans experiments, watches them live, runs analysis, and keeps the archive. Portfolio analytics and digests sit on top of the same cards.
This is the layer you add.
Works with any split system
If your data can be queried with SQL, AB-Labz works with it.
No split system yet?
We recommend GrowthBook — free and open source. We've put together a complete guide to get you from zero to your first experiment.
Read the setup guide
GrowthBook
Statsig
LaunchDarkly
Unleash
PostHog
Firebase A/B
Amplitude
Optimizely
Your own
How a test moves through the workbench
Split launch stays in your system. Everything around it — planning, monitoring, analysis, archive — stays in AB-Labz.
01 · Hypotheses
Hypotheses
Product, design, or marketing drops an idea onto the board as a hypothesis card — metric, MDE, audience, and owner. The backlog fills without waiting for an analyst to open a ticket.
02 · Schedule
Schedule
An analyst picks a card from the backlog, calculates sample size, and parks the test on the Gantt calendar. Overlaps and free traffic are visible before anything launches.
03 · Run
Run
You launch the experiment in your split system as usual and link the experiment ID on the card. Data arrives through the API or Runner. SRM, sample progress, and win forecast update on the dashboard without anyone opening a spreadsheet.
04 · Analyze
Analyze
When the sample is ready, one click runs the engine. It handles preprocessing, picks the right method for each metric type, and applies FDR where needed. AI writes a structured conclusion from the full experiment context.
05 · Archive
Archive
The finished card becomes a searchable archive entry you can share or export. Over time those results feed Portfolio Analytics for the whole program, and digests keep the team in the loop.
Start with Solo — connect later
Use the workbench free. Bring in team features and a data connection when you're ready.
Quick start guide