From payments to healthcare, the teams below run Forge on production every day. These are their numbers, measured on their own systems.
At 14:02 checkout started timing out. By 14:09 Forge had grouped 214 alerts, traced them to a pool limit in one deploy and had a canary of the fix at 64 percent. We approved it from Slack.
52%
fewer repeat incidents
Ledgerline moves money for 3,000 businesses on a platform of 380 services. Every outage was a customer escalation, and the same handful of failure patterns kept coming back because the lessons lived in old review documents nobody reopened.
Forge mapped the platform in its first afternoon, read two years of incident history and began grouping alerts the same week. Within a month, every recurring pattern had a guardrail check that runs before each deploy, and every new incident arrived with its likely cause attached.
Over two quarters, repeat incidents fell from 31 to 15, weekly pages fell by 93 percent, and median time to mitigation dropped from 48 minutes to 7. Engineers approved 212 Forge fixes, and none of them needed a rollback.
Kestrel routes freight across 40 countries. Forge replaced a wall of per-team alerts with grouped incidents that name the owner, so the right on-call engineer is paged first.
Lumen runs Forge in its own VPC with no patient data in scope. Capacity and certificate watches now catch the slow problems that used to surface as weekend outages.
Forge learned Parcelwise's review rules, flag policy and deploy windows from 1,240 past pull requests. Approval rates on Forge fixes rose from 61% to 89% in the first quarter.
The first week it found three services we had forgotten we were still paying for.
Our quietest on-call rotation in four years, and nothing got missed.
It says when it is not sure. That alone made our senior engineers trust it.
Reviews take minutes because every pull request explains itself and links the evidence.
We approved our first Forge fix from a phone in an airport. It held through the full canary.
Incident reviews used to take a day to write. Now the draft is waiting when we close the channel.
Start with a 30 day pilot on one production environment. A Forge engineer works alongside your team the whole way.