Validation · drive to a verdict
Validation to-do
The ordered, cheapest-first test to reach a confident go / park / kill on Cited — the AEO/GEO + agent-commerce swing. Tick items as you go; the bar and verdict save in this browser. Rule: no full product build until every gate is green.
◆ The one make-or-break
Can we reliably move a brand's presence in AI answers, prove the lift decays without us (so it's recurring), attribute revenue to it — and get businesses to pay outcome-priced?
Four linked bets, tested in order: Move (method works) → Decay/Retain (recurring, not one-off) → Attribute + Pay (real money) → Economics (margin + delegable). The method proof is cheap sweat; the paid-pilot gate is the real one.
✕ Falsification — any one of these kills it
- We cannot reliably move presence in ≥2 of 3 test brands vs controls.
- Lift persists without us — one-off SEO-style work, no recurring value to sell.
- We cannot attribute any revenue/traffic to the work, and no business will pay for it.
- Delivery can't be made delegable/automatable at a margin — it's a bespoke consulting hourly job, not a scalable product.
Verdict
After the gates: pursue seriously, park, or kill?
▲ GO — pursue
All gates green. One of the 2–3 we commit to.
⏸ PARK
Signal mixed / too early vs peers. Revisit later.
✕ KILL
A falsifier tripped. Stop, free the slot.
~6–8 wks
to a confident verdict
< $2k
tools + a small traffic test; mostly sweat
0
product code before Gate C passes