Every defect ledger already ranks its causes by annual cost, and the ranking is wrong for the decision it gets used for. A callback that costs the same as another one this year is not the same problem if it sits inside a statutory warranty tail that closes in eighteen months versus one that runs five years — and dollars-only ranking treats them as identical.
Every cause traced back to the trade, the plan detail, the supplier or the inspection stage that actually produced it, then re-ranked by how much of the annual cost is still inside its statutory tail. A callback is a symptom. The plan revision, the hold point or the supplier lot number is the fix that removes it once.
Rough opening framed out of square, repeated on every home built to the same plan. The single largest cause in the book.
Truss uplift never detailed on the plan that uses it. A drawing fix, not a drywall fix.
Slab moisture test skipped at handoff. Caught at a stage that already exists in the schedule.
Grade tolerance never verified before backfill — the largest five-year warranty exposure in the book, caught by a ten-minute check.
The rest of this model — the data sets it runs on, the cause split, the working behind the reference figure and a sample output — goes out by email. Two fields, nothing else.