# Round 2: Skeptic All figures ASSUMPTION, carried from r1-skeptic unless noted. ## Does either idea fix the -C$640/month? No. - **Product's lean build** cuts build cost, not cost to serve. Doing attribution, reports, buyer checks and ads by hand adds about 4-6 h/month, so labour goes to ~24 h (C$1,200). Contribution: 399 - 40 - 1,200 = about -C$840. - **Dealer's cross-posting** helps retention, not cost. Pricing the dealer's way (C$175 base plus ~3 shows x C$37) brings in about C$285, so contribution falls to about -C$955. - **Growth's funnel** (C$1,200 ad cost per sale, midpoint) means C$30/day buys under one sale a month. The dealer is net negative on ads alone, so he will not pay a fee on top. - The only lever that moves the number is hours: 24 h to under 5 h by week 8. Hours fall through text-the-car intake, templated ads (Meta catalog feed from stock), and transcript sampling (10%, plus 100% of handoffs). ## Positions **(a) Interface.** Concede to product: SMS is the dealer interface, plus one mobile add-car form. I withdraw "dashboard is the paid product". What he pays for is the sold record plus someone answering after hours. Build the web dashboard only after the pilot shows what he touches. **(b) Financing.** Later. Concede a minimal capture: cash or finance, monthly budget, down payment. No income, employer or credit self-rating in the MVP: it is sensitive personal data (privacy law exposure) and edges toward arranging credit, and I have not checked whether that needs a licence in any province. The deal sheet is a v4 test, once a dealer asks for it. **(c) Week 4.** Process and cost, not demand. Decide week 4 on: sold-lag, answer time, wrong answers (zero), and our hours per dealer. Demand verdict at week 8, on a burst (C$60-80/day, 10 days), read as ranges. Growth is right that 2 weeks at C$500 proves nothing; I now drop "6 booked, 3 shows" as a gate. **(d) Pricing.** - The dealer pays ad spend at cost from day one, in his own ad account with us as partner. We never front or run money in our account. - Month 1: no fee. Months 2-3: C$199 flat only if sold-lag is under 24 h. No per-show fee (shows are easy to dispute and the owner will resent counting them). - I accept the dealer's stop-loss: no credited sale by week 8, no fee. It is cheap, at most C$400. **(e) Attribution.** Take the dealer's narrower rule for the pilot: our sale = a buyer booked or chatting through us before first contact with the lot, 30 days. Product's manual Friday call is right at low volume. Buyer-check texts go out only with dealer-approved wording and an off switch. The 60-day any-car rule waits for CPA, if ever. **(f) Intake.** Text-the-car wins; it is the one idea that cuts both the app-login problem and our hours. Concede the audit to one visit, top 15 cars, done once. Risk: photo-read VIN or price errors. Control: the dealer confirms price and km by reply before a car goes live. ## Concessions - Growth: week-4 demand gate was empty. Product: build less. - Dealer: sold-lag and Marketplace-copy pain are the real adoption test; the missed-call text-back (cheap) is a better lead-in than ads. ## Final top 5 must-haves for v3 1. **Pre-build test:** 5 days of the dealer marking sold by text (90% within 24 h) before any ad spend. 2. **Hours budget:** log hours per dealer weekly; plan to be under 5 h/month by week 8. 3. **Grounded agent:** prices and facts rendered by code, stale cars take no bookings, handoff by rule, plus an all-in-price field validated per province and dealer sign-off on templates. 4. **Text-the-car intake with dealer confirmation**, SMS sold flow, and manual attribution with the narrow rule. 5. **Pricing:** ad spend paid by the dealer at cost, C$199 only after proof, a week-8 stop-loss. ## Kill criteria - Sold-lag: under 70% marked within 24 h in the 5-day trial (stop; segment fails). - Labour above 6 h/dealer/month at week 6. - Week 8 burst: fewer than 20 leads and 4 bookings, or no credited sale. - No dealer commits in writing to C$199+ after the pilot. - Any all-in-price or wrong-answer incident that a template did not catch.