September 5, 2025
The accidental audit that changed everything
I ran a project audit expecting bad news. Instead I discovered 20+ story points of completed work I forgot I had built. The Jira board was lying.
The board said I was behind

Five weeks into building HomelabARR, I did what any PM does when the dread sets in: I audited the backlog. The board was grim, "To Do" stamped everywhere, stories estimated at 8 points sitting untouched. The Go architecture rewrite was supposedly not started. The React migration was listed as "needs work," and the SnapRAID implementation was marked "To Do."
By Jira's account, HomelabARR v2 was maybe 40% complete. I had been building so fast that I forgot to update the tickets, which is the one thing a PM is never supposed to do, and I was doing it to myself.
What the code actually showed
I started comparing the board to the codebase, ticket by ticket, and the gaps were big.
, Native Go Storage Implementation (8.0 SP). Jira said "15% complete prototype with mock data." The actual code was 1,061 lines of production storage management. pkg/storage/storage.go ran 596 lines, pkg/storage/snapraid.go another 465. No mock data, no prototype scaffolding, and it was running.
HL-141, React/Next.js Migration (5.0 SP). Jira said "To Do: frontend migration needed." It was already a full React 19 app: 54 components, shadcn/ui, WebSocket hooks, routing, a theme system, all shipped and working.
HL-137, Go + Docker Architecture Rewrite. Jira said "To Do: architecture needs rewriting." The entire v2-poc was the rewrite: 3,000+ lines of Go backend with full Docker SDK integration. The ticket was describing work I had finished weeks earlier.
, Go CLI with Web GUI (8.0 SP). Jira said "In Progress." It was a 95% finished NAS dashboard, the kind of thing Unraid and TrueNAS charge for. Container management, storage monitoring, file sharing config, app store, all working.
, SnapRAID Implementation. Jira said "To Do: need SnapRAID integration." It was done in native Go, with automatic drive detection and categorization.
20+ story points of ghost work
The audit turned up over 20 story points of completed work that was either tracked wrong or not tracked at all. So the project was never 40% done. Real completion sat closer to 85%. I had been so deep in the building that I never went back to move the cards.
Solo, nothing forces the board update. When I was managing twelve teams at Neptune, a daily standup did it, because someone was going to ask. Alone, there is no standup and no one asking what I worked on yesterday. There is just the code, then the next feature. Dragging a card from "In Progress" to "Done" does not feel like work, so it kept getting skipped.
What the audit fixed
I spent the rest of that day updating every ticket: moving finished work to Done, adding completion evidence, linking the Confluence docs to the Jira stories, and writing the real closure comments I had left blank.
I closed duplicates, linked related work, and found a few stories that had been quietly superseded by others. Normal backlog grooming, except I had never once done it to my own backlog.
I am aware of the irony. I spent a decade teaching teams to keep their backlogs honest, ran the grooming sessions, gave the talks about accurate sprint tracking. Then I spent five weeks ignoring all of it because I would rather build.
The audit changed my head more than any feature did that month. I walked in thinking I was drowning and behind on everything, and walked out knowing I had built most of a NAS operating system in five weeks. The codebase never changed; I just finally had an accurate picture of it.