# Round 1: Skeptic (unit economics, competition, regulation, risk) ## 1. Verdict Not fundable as written: the plan is a services business (audit, ad management, transcript review) priced like software, and it asks dealers to pay for leads Facebook Marketplace gives them free. The real asset is the dealer inventory app plus booking and attribution; sell that, and treat ads as a later add-on. ## 2. Top 5 risks 1. **Cost to serve exceeds the fee.** Hands-on labour per dealer is about 20 h/month (ASSUMPTION), roughly C$1,000 at C$50/h, against a C$299-499 fee. Test: log real hours per dealer for weeks 2-4. Kill if above 6 h/month steady state after week 4. Mitigation: self-serve onboarding, ad templates, transcript sampling (10%, not 100%) after week 4. 2. **Marketplace is free, so why pay?** The only answer is incremental buyers beyond the dealer's own posting. Test: ask each dealer in the demo visit how many leads/week they get from Marketplace and what they would pay for 10 more. If they won't commit a number, stop. 3. **Dealer won't mark cars sold, so attribution and stock freshness both fail.** The plan admits this is "make-or-break" but tests it only after building. Test: before week 2, have the dealer use a spreadsheet/phone prototype for 5 days; require 90% of sales marked within 24 h. Kill if under 70%. 4. **Regulatory exposure lands on us.** Ontario requires all-in prices in any ad (OMVIC; only HST and licensing may sit outside; asterisk fine print is non-compliant). Mistakes lead to complaints, fines or licence action for the dealer, and they will blame us. BC, Alberta and others have their own regimes (not checked: ASSUMPTION that they are similar). The agent can also bind the dealer: a BC tribunal held Air Canada liable for its chatbot's wrong statements (Moffatt v. Air Canada, 2024 BCCRT 149). Mitigation: price field is all-in only and validated; agent answers by templates from dashboard fields, never free-text price, availability or "deal" language; every ad and page has dealer sign-off on a template; contract has the dealer approving prices; get a lawyer to read the template once (~C$1-2k, ASSUMPTION). 5. **Defensibility is thin.** AI BDC chat is commoditised, lead-gen ads are a service, and AutoTrader owns buyer traffic. Our only moat is being the dealer's system of record. Mitigation: lead with the inventory app (stock, photos, sold marks, booking); the data and habit are the lock-in. Also: CASL implied consent for an inquiry is 6 months, but marketing texts after that and the 10-day buyer check ("Did you buy a car?") should be treated as transactional only; have counsel confirm. PIPEDA/provincial privacy law covers buyer phone numbers and call recordings. ## 3. Unit economics (all ASSUMPTION unless sourced) Per pilot dealer, per month, steady state: - Revenue: fee C$399 (mid of plan range). Ad spend passes through at cost (C$25-40/day, about C$750-1,200); it is not our revenue. - Direct non-labour cost: SMS ~C$10, LLM ~C$15-30 (100 chats x ~20k tokens), tracked number ~C$5, hosting share ~C$5. Total about C$40. - Labour: ads management 3 h/wk (12 h), transcript review 2 h/wk (8 h) = 20 h x C$50 = C$1,000. One-time audit: 10-12 person-hours = ~C$550 plus travel. - Contribution now: 399 - 40 - 1,000 = about -C$640/month, before the audit. - Funnel at plan targets: 20 leads -> 6 booked -> 3 shows -> ~1 sale per 2 weeks, so about 2 sales/month. Ad cost per sale ~C$500. If dealer gross is C$2,500-3,500 per car (ASSUMPTION), they net ~C$2,000 on 2 sales minus C$399 fee and C$1,000 ads: only about C$600 ahead, and only if every sale is truly incremental. - CPA: 2 sales x C$275 = C$550/month plus a lower fee. Revenue about the same as the fee-only case; the dealer's win shrinks. - To make C$399 work we need labour at or under about 4 h/month (contribution ~C$160 after non-labour cost, rising as hours drop). - Break-even on a 2-person team at C$12k/month (ASSUMPTION): 12,000 / ~160 = about 75 dealers. At 25 dealers this loses money. Day one is a loss maker; scale only works if hours per dealer fall about 5x. ## 4. Competitors | Name | What it does | Price | Source | |---|---|---|---| | AutoTrader.ca (Trader) | Dealer listings, packages, Deal Builder, inventory tools | Not published. Third-party claim C$450-2,000+/month (unverified) | https://go.trader.ca ; https://cardog.app/blog/autotrader-canada-review | | DealerAI | Chatbot, voice and lead tools, quoted in CAD | Chatbot C$599/rooftop/month; Reach from C$1,199; Voice from C$499 | https://dealerai.com/ | | BDC.AI | AI BDC chat/lead handling | From US$595/month (currency unclear) | https://bdc.ai | | Impel | AI sales assistant for dealers | Not published, per-rooftop quote | https://www.voiceaispace.com/tool/dealerai (comparison listing) | | Nobi | Website AI assistant, no phone/SMS | From ~$25/month | https://nobi.ai/blog/best-ai-assistants-for-car-dealerships | | Facebook Marketplace | Free vehicle listings, DM | Free | general knowledge (not searched) | | Kijiji Autos, dealer website/chat vendors | Listings, sites, chat | Not found: not researched | n/a | Takeaway: AI chat is already C$500-1,200/month and sold per rooftop. Our fee is not out of line, but it does not stand out; nothing here sells a lot audit plus a cheap inventory system to a paper-based independent. ## 5. What to keep - Phone-first inventory dashboard, sold mark and "who bought it" flow: this is the product. - Public car pages with all-in price and a booking code. - Agent limited to dashboard fields, with handoff for anything else. - Buyer-check text as a second attribution source (after legal review). - Fee first, CPA later; the pilot month at cost. - Week-4 leading metrics. ## Kill criteria (add to plan) - Dealer will not mark sold: under 70% in a 5-day trial. - Labour above 6 h/dealer/month by week 4. - Fewer than 10 leads or 2 shows in the first 2 weeks of ads at C$25-40/day. - No dealer will commit in writing to a C$299+ fee after the pilot. - Any all-in pricing or wrong-answer incident not caught by templates.