← All articles

Merge Velocity: Where Work Stalls

· By Mamer Alomari · AI-assisted

#SoftwareEngineering · #DeliveryMetrics · #WorkflowDesign

A week of pull-request data shows merges are fast and the review queue is not growing. The constraint is the supply of ready work.

A week of pull-request data makes one pattern plain: work is merging quickly, and it is not piling up in review. The constraint is earlier on the board.

The window runs from November 7, 2026 through 9:30 AM ET on November 14, 2026. In that span, 228 pull requests were merged, about 29 a day on average. Daily counts ranged from 5 to 53.

Once a pull request was opened, it did not wait long to land. The median time from open to merge was about 3 hours 20 minutes. The mean was about 5.6 hours. Ninety percent of those pull requests merged within 15.3 hours.

Open-to-merge time measures the life of a pull request, not the life of an issue on the board. A short merge interval can sit beside a long wait for work to become ready, and the two should not be treated as the same clock.

The review queue is not growing. There were 8 open pull requests at the end of the window, after swinging between 4 and 14 since November 11. Issues marked in review fell from about 47 on November 9 to 34. Most of those 34 are already merged and waiting on an owner verification line.

The ready column tells a different story. It held 1 issue, against 22 blocked and 31 parked.

The constraint is not merge speed, and it is not review throughput. It is the supply of ready work. A fast merge path cannot keep a board moving if almost nothing is sitting in ready to be picked up.

One measurement is still missing. This snapshot does not track how long an issue sits in ready, in progress, or in review. The next step is to log each status change so time-in-column can be measured. Until that log exists, column dwell time should be treated as unknown, not inferred from the open-to-merge figures above.

More writing →