The Bali property research methodology: how the numbers get built.

1. Listing scrapes. Bali agency platforms, scraped systematically for asking-price comparables and time-on-market signals. The listing population is biased — only what an agent has chosen to publish — but the bias is observable and adjustable.
2. Booking-platform occupancy data. Airbnb and Booking.com scrapes for ADR ranges, occupancy variance by region, seasonal patterns. The data is noisy at individual-property level but tight at regional-aggregate level.
3. BPN title-office records. Parcel boundary verification, encumbrance checks, zoning classification. The authoritative title-layer source.
4. Notaris registries. Leasehold-encumbrance verification on existing assets. Confirms whether a parcel sits under a registered lease with a third party.
5. On-the-ground operator interviews. Fee-stack levels, staffing model norms, operating-permit reality by region. The piece that catches what the data alone cannot.

Each source is partial. Each is biased in a known direction. The reconciliation across all five is where the credible number emerges. The Thailand parallel sits at the Thailand data source — same multi-source logic, different market.

The amateur reads one source. The agent quotes one number. The analyst reconciles five and looks at the disagreement. Five sources that converge on a single number are giving you a credible signal. Five sources that diverge sharply are giving you a different signal: that the market is in transition, that one source is biased, or that the headline reading is wrong. The analytics method behind every Bali number.

Worked example: a Canggu villa listing quotes 12% gross occupancy yield. The booking-platform occupancy scrape shows ADR compression. The operator interview confirms staffing intensity higher than the brochure model. The BPN check shows leasehold structure with 22 remaining years. Five sources, one resolution: the 12% headline does not survive contact with the operating reality. Research vs the agent: what the data actually says.

The net-yield walkout shows the reconciliation logic applied to a specific deal. The foreign-buyer view of the same data.

1. Single-source headlines. A brochure number with no cross-source verification does not enter the model.
2. Optimistic-only scenarios. A yield model that only closes the math under perfect conditions is not a model; it is a brochure with a spreadsheet attached.
3. Borrowed Thai benchmarks. Bali operates under a different legal regime, different operating model, different cost stack. Importing Thai condo benchmarks into Bali villa underwriting produces a wrong answer. The Thai methodology is parallel, not transferable.