Skip to main content

Work Work tracking and review dashboard

Planner, the system we run our own work on

Result
Tasks, reviews and builds across more than 60 repositories in one dashboard; at capture, 11,173 tasks completed and 88% passed review the first time.
Built with
Go, PostgreSQL, Alpine.js, Chart.js
Status
Our own work
Quality metrics. First-pass rate across 10,723 reviews, rejection and escalation rates, and review latency for recent completed tasks, with reconcile-shortcut tasks left out.

The project

The problem

We build and run a lot of software at once: websites, business systems, security tooling and the infrastructure under them. Work was spread across issue trackers, code hosts and build logs, and nothing showed whether a change had actually been reviewed, tested and shipped.

What we built

Planner is a Go service with a PostgreSQL store and a server-rendered dashboard. Work is tracked as tasks with written requirements. The planner refuses to let a task be claimed until it has requirements or go to review until each one is marked met, and its review gate checks each change for failed tests and merge conflicts.

The dashboard reads the same data: work in flight, stale tasks, review latency, rejection reasons by category, and test and build status for each repository it tracks.

What changed

Review is no longer a matter of trust. Reviews are recorded against each task and rejections are grouped by reason code, so the weak spots show up in the numbers instead of in production.