1. Listing extraction
The pipeline fetches the listing page from the portal (Funda, Idealista, Rightmove, ImmoScout24, Immoweb, SeLoger, Homegate, Willhaben and 40+ others) and extracts asking price, size, rooms, property type, amenities and address. Extraction uses structured data where the portal provides it and a two-step language-model probe as fallback; every extracted field carries a confidence flag.
2. Location and market enrichment
The address is geocoded and enriched with distance to airport, beach and ski, walkability, sunshine hours, tourism growth and safety indicators. Short-term rental revenue inputs — average daily rate, occupancy, seasonality curve and comparable count — come from city-level open data (InsideAirbnb) where available, falling back to country-level benchmarks. The data source and a 0–100 confidence score are shown on every report.
3. Financial engine
Annual revenue = ADR × 365 × occupancy. Operating costs combine a variable component (platform, cleaning, management as a percentage of revenue) with fixed components (insurance, utilities, maintenance reserve, property tax). NOI = revenue − costs. Acquisition costs use a country-specific table (e.g. Belgium ≈14%, Spain ≈12%, Netherlands ≈9.5%, Denmark ≈2%). Net yield = NOI ÷ (price + acquisition costs).
- DSCR: NOI ÷ annual debt service on a 70% LTV, 25-year annuity at the market mortgage rate; a stressed DSCR adds 2 percentage points
- Break-even occupancy: (operating costs + debt service) ÷ ADR ÷ 365
- IRR (10 years): three scenarios — pessimistic (occupancy × 0.65, rate × 0.85), base, optimistic (occupancy × 1.25, rate × 1.15) — with house-price-index-based exit values
- No mortgage is assumed in yield figures; leverage metrics are shown separately
4. Tax and regulation
For each of 13 countries the engine holds named tax regimes (for example micro-BIC and LMNP réel in France, cedolare secca in Italy, Box 3 in the Netherlands, non-resident income tax in Spain and Portugal) and computes an effective rate on NOI for each, recommending the lowest. Tourist tax per night and a rough capital-gains estimate are included. City-level STR rules — licence requirement, night cap, primary-residence condition, enforcement intensity — come from a maintained regulations database.
5. Risk score
Ten dimensions are scored 1–10 by rules (higher = riskier) and combined with fixed weights: regulatory 25%, market 20%, liquidity 15%, seasonality 10%, maintenance 10%, currency 5%, natural disaster 5%, platform dependency 5%, interest rate 3%, damage 2%. The overall score is presented on a 0–100 scale where 100 is safest. Explanatory text per dimension is generated by a language model from the scored inputs; the scores themselves are not.
6. Narrative and quality assurance
A language model writes the executive summary, strengths, risks and negotiation points from a structured JSON context and is instructed never to introduce numbers that are not in that context. The buy/consider/avoid recommendation is overridden by server-side rules on net yield, overall risk, regulatory risk and DSCR. Before a report is marked complete it passes seven QA checks (valid price, net yield within 0–40%, geocoding present, scenario spread, risk present, financials present, narrative present); failures mark the report for review and no credits are charged.
Limitations
Revenue inputs are market estimates, not guarantees; confidence is lower where only country-level data exists. Tax calculations are indicative and ignore your personal situation. Regulation changes frequently — the report shows the last-verified date. Terrivio is analysis software, not financial, tax or legal advice.
Frequently asked questions
Does Terrivio use AI to calculate yield?
No. All financial metrics are computed by deterministic code. AI is used only to write the narrative and per-dimension risk explanations from already-computed inputs.
Where does the revenue data come from?
City-level open short-term rental data where available, with country-level benchmarks as fallback. Each report shows its data source and a confidence score.
How often are tax and regulation data updated?
Regimes and city rules are reviewed against primary sources on a rolling basis; each report displays the last-verified date for its regulatory data.
Can I see the assumptions behind a report?
Yes. Every report lists the ADR, occupancy, cost ratios, acquisition cost percentage and tax regime used, so you can adjust them in the scenario calculator.