War-game a research readout before sample size, method, and recommendations leave the private room and become roadmap truth.
Research readouts fail when quotes are treated as prevalence, when recruiting bias is hidden, when recommendations outrun the evidence, and when product already decided the answer before the study. A tidy slide deck is not evidence that the insights survive a careful room.
Freeze the readout package
Lock the question the study answered, the method, the sample, the limitations, and the decisions the readout is allowed to change. Attach the discussion guide, screener, raw notes or coded themes, and the decision memo that will use the findings. If research, product, and leadership tell different stories about what the study can claim, stop and reconcile before the seating pass.
Write the decision you will make if the war game finds nothing new, and the reverse criteria: the bias, sample gap, or overclaim that would reopen the readout within a stated window.
Seats that matter
- Method skeptic. Sample, recruiting, and analysis choices that weaken prevalence claims.
- Product. Recommendations that outrun what the evidence can support.
- Customer or user. Paths and pains the study never heard because of who was recruited.
- Finance or ops. Cost or capacity claims the readout implies without numbers.
- Prior-belief seat. Where the pack confirms a decision that was already made.
Give every seat the same pack. Secret "we already know" notes for one chair create fake consensus.
Private failure loop
Feed Pingpong the readout, method notes, and decision memo. First pass steelmans the findings. Later passes attack from skeptic, product, customer, and prior-belief seats. Final pass turns surviving objections into narrower claims, clearer limits, or a hold. Delete invented prevalence from a handful of quotes and dual-counted themes from the same participant.
If quiet roadmap commits are the real risk, force the review to invent the decision rule before the readout meeting. Write which claims can change the roadmap, which need a follow-up study, and who owns the next method. A readout that only works with silent overclaim is not research. It is a pitch with footnotes.
Separate insight from recommendation in the model. If product blurs them to clear a feature, ask the skeptic seat to price both paths and the evidence each implies. Pair with war-game a customer interview plan, war-game a product launch, pretend you are the CPO, war-game a feature request triage, and the war-game decisions hub. See how to run a Pingpong.