Is Statsig Down?

This incident has been resolved --- [2026-08-07T21:00:38.962-07:00] (monitoring) Thank you for your patience. The main query-intensive jobs are through now, but it also took quite a bit longer than before. Catch up for subsequent jobs may therefore take up to 3am PT to complete. Our sincere apologies for the delay and we will continue to share updates as we monitor the job completion. --- [2026-08-07T20:55:40.653-07:00] (monitoring) We are still making progress on re-running the exports. We should start to see some exports completing but the expected completion time for the full system to rerun has extended to 9pm PT. We will continue to keep you updated and aim to have the next update at 9pm PT! To clarify: 1. This only impacts customers who use Statsig warehouse native. If you are on Statsig cloud, you can ignore this status page. 2. For WHN customers, once we finish rerunning the final jobs, Statsig will automatically take care of refreshing the data for all experiments that are configured to incrementally refresh metrics. --- [2026-08-07T16:10:54.261-07:00] (monitoring) This issue only impacts customers on Warehouse Native on the echidna pipeline. Around 11 am PT, we have identified that WHN accounts will have event and exposure export delays from August 6. No data is lost as the job rebuilds the full day from source so there will be no gaps once the post-revert run lands. Everything through August 5 is intact. Around 1pm PT, we have implemented a fix that should have mitigated this issue. We expect exports to complete around 7pm PT.

Between approximately 20:56 UTC on July 8 and 01:40 UTC on July 9, customers using the V2 Config Delivery API may have experienced incorrect evaluation of launched Feature Gates. During this window, launched Gates could return false instead of their expected default. V1 consumers were not affected. We identified the root cause as a recent change to config-spec generation. We rolled back the affected services across all regions and regenerated config payloads. We have verified that launched Gates are again returning their expected defaults, and all Config Delivery API components are now operational. No action is required to restore correct evaluation going forward. However, if your systems persisted data or took actions based on evaluations returned by the V2 Config Delivery API during the impact window, we recommend reviewing and reconciling that downstream data, as affected launched Gates may have returned false during that time. --- [2026-07-08T19:15:14.749-07:00] (monitoring) A fix has been implemented and we are monitoring the results. --- [2026-07-08T18:39:38.496-07:00] (identified) We are continuing to work on a fix for this issue. --- [2026-07-08T18:35:38.831-07:00] (identified) We are continuing to work on a fix for this issue. --- [2026-07-08T18:29:00.264-07:00] (identified) We are continuing to work on a fix for this issue. --- [2026-07-08T19:59:42.274-07:00] (identified) Since approximately 1:56 PM PT on July 8 (20:56 UTC on July 8), customers using the V2 Config Delivery API may have experienced incorrect evaluation of launched Feature Gates. Specifically, launched Gates may have returned false instead of their expected default value. Our team identified the root cause as a recent code change and started a rollback of the affected services. We were actively working to confirm that evaluations returned by the V2 Config Delivery API were again showing expected defaults for launched Gates. V1 consumers were not affected. At the time of this update, no immediate customer action was required. We would provide the next update as more information became available.

Warehouse native exposure forwarding has caught up. --- [2026-06-19T22:36:18.235-07:00] (monitoring) Data ingestion has caught up and is operating normally. We are continuing to monitor warehouse native exposure forwarding delay. --- [2026-06-19T21:45:37.598-07:00] (monitoring) We have identified an issue from a recent change impacting BigQuery connectivity since 4:30pm PST. No data has been lost. We have rolled back the change and data ingestion is catching up. Experiment results for some customers may be delayed.