1. Verdict Idea Score: 51/100 — Test cheaply. The underlying problem is real, but the proposed magic trick—one fridge photo → genuinely useful meals for a family for seven days—may be weaker than it sounds because the photo cannot see the freezer, cupboards, quantities, expiry dates, preferences, or what the children will actually eat (assumption). 2. Scorecard | Criterion | Score | Why | |---|---|---| | Pain intensity | 6/10 | "What are we eating this week?" is recurring friction, but families already solve it imperfectly with familiar meals, shopping lists and takeaway (assumption). | | Willingness to pay | 4/10 | Meal planning competes with free solutions. I would test £3.99–£5.99/month rather than assume parents will subscribe (assumption). | | Reachability | 6/10 | Parenting communities, UK family-food creators and local groups give you cheap channels for the first 100 testers (assumption). | | Competition | 4/10 | SuperCook already recommends recipes from ingredients you have; Samsung Food offers weekly meal planning; SideChef combines ingredients-on-hand with meal planning. | | Defensible differentiation | 3/10 | Photo input is convenient, but computer vision/API access is easily copied (assumption). Your moat would need to become family preference/history data. | | Unit economics | 5/10 | At ~£5/month (assumption), AI serving costs could plausibly be manageable, but paid acquisition could destroy the economics (assumption). CAC needs testing. | | Founder fit | 6/10 | £2,000 and 10 h/week are enough for a narrow prototype (assumption), not enough for an overbuilt polished consumer platform. | | Speed to first revenue | 7/10 | You can charge for a crude web-based version before building native apps (assumption). | Total: 41/80 = 51/100. 3. Three riskiest assumptions 1. Parents actually want the fridge-photo workflow. Cheap test: £0–£30. Build a landing page with fridge-photo upload; manually produce plans behind the scenes. Recruit 20 UK parents within 7 days. Pass: ≥10 upload a photo, ≥6 request another week's plan, ≥3 attempt to pay £4.99. 2. The generated plans are useful enough to trust. Cheap test: manually/AI-generate 10 plans from real fridge photos. Pass: ≥7/10 users rate the plan 8/10+, and ≥5 cook at least three suggested meals. 3. Families will pay repeatedly. Cheap test: Stripe/payment-link preorder, ≤£50 promotion (assumption). Pass: ≥5% of qualified landing-page visitors buy at £4.99/month or ≥20% of completed testers convert. 4. Pre-mortem - Image recognition misses ingredients or quantities, so users spend time correcting it. - Children reject unfamiliar meals, making technically good recommendations useless (assumption). - Free competitors are "good enough"; Samsung Food already offers weekly planning, while SideChef and SuperCook cover adjacent ingredient-driven discovery. - Retention collapses after the novelty of photographing the fridge wears off (assumption). - You spend your limited 10 h/week building OCR, authentication, mobile apps and recipe infrastructure before proving anyone pays. 5. Sharper versions A. "Use it before it expires" family food-waste assistant — estimated 65/100 (assumption). Photo ingredients, identify what needs using first, then suggest 3 dinners, not an entire week. Much clearer immediate value. B. "5 dinners for picky families under £X" — estimated 69/100 (assumption). Parents specify budget, children's disliked foods, dietary constraints and existing ingredients. The product generates five realistic dinners plus one consolidated shopping list. This attacks a stronger combination of planning + budget + family preferences. 6. Seven-day validation Day 1: Create one-page proposition: "Photograph your fridge. Get five family dinners using what you already have." Day 2: Add photo upload, family size, dislikes, allergies and email. No app. Day 3: Recruit 10–20 UK parents through relevant communities and personal outreach. Day 4: Produce recommendations manually with AI. Track corrections required. Day 5: Ask users which meals they would actually cook and why they rejected others. Day 6: Offer the next week for £4.99 (assumption). Take real payment, not "would you pay?" Day 7: Calculate upload → useful-plan → second-use → payment conversion. You can send this today: "Hi! I'm testing a simple idea for busy parents: take a photo of your fridge and get a 5-dinner family meal plan based on what you already have. I'm looking for UK parents with 2+ kids to test it free this week. You'd send one fridge photo, get the plan, then tell me what was useful or useless. Would you be willing to try it?" 7. Kill criteria Stop or materially pivot if, after 30 qualified parents: - fewer than 30% complete the fridge-photo flow; - fewer than 50% say they would cook at least 3 recommendations; - fewer than 25% return for a second week; - fewer than 5% pay £4.99 when genuinely offered; - or producing an acceptable plan still requires substantial manual correction. Do not spend the £2,000 yet. Spend the first £100 proving repeat usage and payment, because the technical app is much easier to build than the habit.