Integration · Experimentation
Statsig
Audit-mode connector. We read your Statsig Forward stream (assignment rows) and Console API (experiment metadata + bucket weights), then translate every assignment into an ExperimentAssignment in our pipeline. Propensities are tagged assumed_weighted — derived from the published bucket weights — and the readout is rendered as advisory, not billable. Upgrade individual experiments to decision-mode (our SDK) when the cohort math warrants outcome-priced billing.
The contract
What flows in
- Forward stream (S3 export) of assignment rows + event rows
- Console API: experiment metadata, bucket weights, layer config
- Optional: gate evaluation log if you want feature-flag holdouts read alongside
What flows out
- Doubly-robust lift readouts on your existing Statsig experiments (CATE grid included)
- Propensity-quality badge on every readout (assumed_weighted) so the CFO sees the math's honesty
- Connector status + sync history surfaced live over the dashboard's job-run timeline
Setup
- 01Connect: paste your Statsig Console API key + S3 forward-stream bucket / role ARN
- 02Pick which experiments to import — defaults to all active
- 03Map reward events: which Statsig events count as reward, attribution window, sign
- 04First sync runs immediately; subsequent syncs every 15 minutes (configurable)
- 05Upgrade-to-decision-mode CTA appears on every imported experiment once you've reviewed the audit number
Why audit-mode is the right first step
Statsig's t-test on assignment buckets is what your CFO can't sign off on post-ATT. We don't ask you to migrate — we read the same log and produce a DR + ESS readout the CFO will accept. The audit number arrives without an engineering ticket.
When the cohort breakdown convinces you a specific experiment is worth outcome-priced billing, one click cuts that single experiment over to our SDK. Your other experiments stay in Statsig.
Other experimentation integrations
Wire Statsig in a 30-minute call.
We pair with your engineer, ship the integration live, and run the first decision on your stack.