Integration · Experimentation
GrowthBook
Audit-mode connector for the open-source experimentation platform. Reads via the REST API for experiment metadata + per-experiment warehouse SQL templates (Snowflake / BigQuery / Postgres) for assignments and metrics. Propensity is assumed_weighted against your declared variation weights. Co-marketing path with the GrowthBook community — same MIT spirit as offpolicy.py.
The contract
What flows in
- REST API: experiments, metrics, dimensions, variation weights
- Warehouse SQL: assignment events + metric events via the templates you already configured in GrowthBook
What flows out
- DR + ESS readout on every imported experiment, refreshed on the GrowthBook metric cadence
- Per-cohort CATE grid that GrowthBook's default analytics doesn't surface
- Disclosed assumed_weighted provenance — the readout is advisory until decision-mode upgrade
Setup
- 01Connect: paste your GrowthBook API token + warehouse credentials (we reuse your GB warehouse connection if exposed)
- 02Confirm which experiments to mirror — defaults to all running
- 03Sync interval defaults to GrowthBook's update cadence; overridable
- 04Optional: enable side-by-side comparison view that shows GB's frequentist number next to our DR estimate for the same experiment
Open-source meets open-source
GrowthBook is MIT-licensed; offpolicy.py is MIT-licensed. The connector is the natural bridge between two OSS-led growth stacks — same audit ethos, deeper math.
We don't write back into GrowthBook. The customer's GB instance stays the source of truth for assignment; Metapolicy adds the doubly-robust causal layer on top.
Other experimentation integrations
Wire GrowthBook in a 30-minute call.
We pair with your engineer, ship the integration live, and run the first decision on your stack.