War-game a priority score model before weighting rules, override budgets, and publish steps harden into what every backlog promote will quote.
Score models fail when the rubric invents precision the data never had, when overrides dual-count the same ask as scored and as exempt, when money lanes still lack a named evidence owner, and when ops cannot show who owns the decision after a partial score misfire. A neat spreadsheet dump is not evidence.
Freeze the model
One sentence for why the score exists, which lanes and ticket types it covers, who owns weighting, evidence fields, and override, and the abort trigger if gaming or capacity miss past a named threshold. Attach the draft rubric, sample scored tickets, override log, and the measured path from score change to published backlog order. If product, sales, and engineering disagree on which fields are truly required, stop and reconcile first.
Name the decision you will make if the war game finds nothing new, and the delay criteria if any money lane still lacks a named override owner or a verified publish cadence.
Seats that matter
- Product ops. Where weighting invents fairness or hides shared scores across lanes.
- Engineering. Where capacity math still leaves the queue and becomes standing fiction.
- Sales or CS. Which customer promise breaks first when a legitimate ask scores low.
- Requester seat. How decline language trails the customer-visible score.
- Skeptic. The claim that looks strongest and is least evidenced by prior scoring cycles.
Attach the same source pack to every seat. Secret score boosts for favorite accounts only create fake calm.
Loop the review
Feed Pingpong the draft rubric, sample scores, and open risk list. Early passes steelman the design. Later passes attack from product ops, engineering, sales, requester, and skeptic seats. End with a pass that turns surviving objections into clearer owners, a timed publish, or a hold. Delete invented "we already score cleanly" claims and dual-counted success rates.
Force a month-after narrative: what happens if a large account escalates a low score, if sales quotes a feature still below the promote threshold, or if an operator widens an override under launch pressure. If those stories are stronger than your mitigation plan, fix the package before you publish the model.
Pair with the product ops lead seat, a backlog grooming cadence review, a feature request triage review, and the war-game decisions hub.