Incident Report: Pinterest publishing failures
Date: August 5, 2026 Duration: Approximately 1 hour (08:46 to 09:50 UTC) Severity: Major
Root Cause
The incident was caused by a degradation of Pinterest's API, not by any change on our side. Pinterest began returning HTTP 500 errors and timeouts on requests to create pins and to fetch board data. A small volume of posts continued to go through, so the degradation was partial rather than a full outage.
We confirmed this was upstream by checking that all other connected channels were publishing normally, that no recent deploys touched Pinterest or shared publishing code, and that our publishing infrastructure was healthy throughout. Reports from third-party monitoring sites lined up with the same window.
Customer Impact
Customers publishing to Pinterest during this window saw posts fail. At the peak, only about 5.5% of Pinterest publishing attempts succeeded. Failed posts surfaced an error that included Pinterest's own message, "Something went wrong on our end. Sorry about that."
No other channels were materially affected. Publishing success rates for all other networks stayed above 91% for the duration.
Steps to Resolution
We were alerted to a spike in Pinterest publishing errors and opened an incident within minutes. Our team ruled out deploys, worker capacity, and infrastructure issues, then segmented publishing success by channel and confirmed the failures were isolated to Pinterest. External signals corroborated that Pinterest was experiencing broader issues at the time.
Because the failure was upstream, no code or infrastructure changes were required. Error rates returned to normal as Pinterest recovered, and we verified this with a real end-to-end Pinterest post at 09:50 UTC before moving the incident to monitoring. The status page was updated throughout and marked resolved at 10:20 UTC.
Key Learnings
We can't prevent an outage on a partner's platform, but we can reduce how much it costs our customers. Three things we're taking away:
The error message customers saw was a raw pass-through of Pinterest's own text, which isn't useful when you're trying to understand what happened. We're adding a clearer message for this case so it's obvious the problem is with the platform and not your post.
We're reviewing our retry behaviour for posts that fail with provider side server errors, so more of them recover automatically once the platform is healthy again.
When a platform is failing at a sustained rate, continuing to send requests helps nobody. We're looking at backing off automatically during these periods and resuming once a test request succeeds.
