3. Feature-by-Feature Scrutiny (Your Stated Features)
Categories: TS = Table stakes · D = Differentiator · X = Distraction · L = Liability.
3.1 Classification
| # | Your feature | Category | Reasoning |
|---|---|---|---|
| 1 | One app for business, short-stay, vacation and tourism travel | L (positioning) | Business travel has a different buyer (the company), requirements (policy, approvals, expenses, duty of care) and competitors (Navan, Perk, Concur). Mixing it with leisure is the all-in-one mistake again. Pick leisure. |
| 2 | Requirement- and days-based itinerary generation | TS | Free in ChatGPT, Gemini, Google, Booking, Expedia, Tripadvisor and MakeMyTrip. It earns nothing on its own. It's a differentiator only if it handles hard constraints better than the others: budget caps, dietary rules, mobility, kids, visa rules for your passport, opening days, and realistic travel times. |
| 3 | Flight options | TS (search) / L (selling) | Search and comparison is commodity (Google Flights, Skyscanner). Selling flights requires IATA accreditation or a consolidator, and brings refunds, schedule changes and chargebacks. Flight data APIs are paid or partner-only (Google Flights has no public API). Deep-link out only. |
| 4 | Stay options | TS | Affiliate deep-links (Booking.com, Agoda, Expedia, Airbnb where possible) are standard. The commission is small (estimated at a few percent of booking value). Direct hotel inventory needs bed-bank contracts and payments, which means liability. |
| 5 | Food recommendations | TS → D in niches | Google Maps covers general food discovery. Dietary-specific recommendations are a genuine gap: pure-veg, Jain, halal, kosher, allergy-safe, "Indian food near me in Prague". Hard to do reliably without good data. |
| 6 | Destinations and "what to try and explore" by location | TS | Commodity content: Google, Tripadvisor, travel blogs, YouTube, Instagram. |
| 7 | End-to-end "everything required" | L | Bundling flight + stay (+ activities) can legally make you a package organiser under EU/UK rules (with UK ATOL for flights), liable for the whole trip, insolvency protection included. Even without bundling, users will hold you responsible when anything goes wrong. |
| 8 | Match with a guide | L | A two-sided marketplace with safety risk (solo travellers meeting strangers), licensing (many countries and heritage sites require licensed guides), insurance, payments, disputes and no-shows. GetYourGuide, Viator, Klook and ToursByLocals already do this. Link to them via affiliate instead. |
| 9 | Be like a guide: info and alerts during the trip | D (if done well) | The strongest consumer idea on the list. A companion that knows your itinerary and adapts: flight delayed → reshuffle the day; rain → indoor swap; museum closed Monday → move it; "leave now for your 3 pm tour". Most AI planners stop at trip start. Risks: data costs (flight status from FlightAware, Cirium or OAG is paid; weather; closures), battery and offline behaviour, and AI hallucination. |
| 10 | Safety and advisory alerts | D / L | Useful (government advisories, strikes, weather events). If you miss or misstate one, you face trust and liability exposure. Source only from official feeds and label them as informational. |
| 11 | Inputs from the internet (reviews, likes and dislikes) | L | Scraping Google, Tripadvisor or Booking reviews violates their terms. Google Places data has strict caching and display rules. Licensed data costs money. |
| 12 | Other people's likes, dislikes and history (crowd personalisation) | X / L | A cold-start problem: you need millions of trips before the "people like you" signal beats Google's reviews. Storing travel history, locations and passport details is sensitive personal data under GDPR, DPDP and CCPA. |
| 13 | Personal preference learning from the user's own history | D (long-term) | Genuinely valuable ("you prefer boutique stays, hate early starts, eat vegetarian"). But with 1–3 trips a year, it takes years to learn anything. It works far better for an advisor who sees many trips. |
| 14 | Mobile app + web | TS | A mobile app is required for in-trip use. Planning happens on desktop and mobile. For the advisor wedge, the advisor uses web and the traveller uses a mobile web app, installed as a PWA, not a native app, at first. |
3.2 Scoreboard
| Category | Count | Features |
|---|---|---|
| Table stakes | 6 | Itinerary generation, flight search, stays, food (general), destination content, apps |
| Differentiator | 3 | In-trip adaptive companion, niche hard-constraint planning (dietary, accessibility, visa), long-term preference learning |
| Distraction | 1 | Crowd likes and "other people's history" |
| Liability | 5 | Business + leisure in one, selling flights and packages, end-to-end promise, guide marketplace, scraped internet data |
About 20% of the idea is a reason to choose you. The rest is free elsewhere, or a legal or operational risk.
3.3 Feature bloat and building to avoid selling
- The booking engine. It feels like the "real" product but takes months, brings refunds and support, and earns thin commissions.
- The guide marketplace. Building both sides before one side pays.
- Crowd-sourced personalisation. An ML project that needs data you won't have for years.
- Business travel. A second product for a second buyer.
- Every destination worldwide. Content quality collapses at global scale. AI hallucinations about opening hours and closed restaurants are the top complaint about AI planners.