Selection used to be partly a reading exercise. That signal is gone. The only place it moved to is the interview, and the interview is the one thing you cannot scale.
Same five-claim protocol, same signals, same bar — so two companies can be read against each other rather than against whoever reviewed them.
| Company | Examined | Held | Did not | Flag |
|---|---|---|---|---|
| Tarn Logistics | 5 | 1 | 3 | no sources |
| Novaplan | 5 | 3 | 1 | one contradiction |
| Halden Robotics | 5 | 4 | 0 | clean |
| Kestrel Health | 5 | 0 | 4 | deck is not theirs |
Run it on the shortlist before interview slots are handed out. Interview the founders whose claims held, not the ones whose decks read well.
Run the whole cohort on arrival. You learn where each team is thin before mentoring starts, instead of in week six.
Run it two weeks out. The delta against the first run is the only honest answer to whether a team is ready to be on stage.
Failures cluster by type. Eight capability gaps in go-to-market need a different mentor than eight unsourced market figures.
Three years of PitchBob, 21 accelerator programmes, 60+ ecosystem partners. Every answer typed by a founder describing their own company. It is why a verdict can say how thin an answer is, rather than that it feels thin.
Nobody buys this from a page. Run one cohort, keep the dashboard, and decide afterwards whether selection week was easier. If it wasn't, there is nothing to discuss.