Bali villas are marketed at 12% yields. We asked eleven agencies for the spreadsheet behind that number

Between January and July we collected eleven investment decks from agencies selling villas in Canggu, Ubud and Uluwatu. Nine of them printed a yield figure on the cover. The lowest was 10 per cent. The highest was 15. We wrote to all eleven and asked the same question: what occupancy rate does that number assume, and where does the rate come from?

Four agencies replied. Two sent a rate without a source. One sent a source that turned out to be its own historical portfolio, unaudited and unrepresentative. One declined to answer. The remaining seven did not reply at all.

That is the whole story, and everything below is arithmetic.

What occupancy do the decks assume?

A gross yield is annual rental revenue divided by purchase price. Revenue is nightly rate multiplied by nights sold. So a yield claim is always, underneath, a claim about how many nights the villa sells.

Work backwards from the decks. A two-bedroom villa in Canggu at USD 400,000, advertised at 12 per cent gross, needs USD 48,000 of annual revenue. At the average achieved nightly rate the same decks quote — around USD 190 — that is 253 nights. Two hundred and fifty-three nights is 69 per cent occupancy.

None of the eleven decks stated 69 per cent. Three stated 70. One stated 75. The rest simply printed the yield.

Independent data puts trailing twelve-month occupancy in Canggu at 36.2 per cent, in Ubud at 36.6 per cent and in Uluwatu at 46.2 per cent.◆ Estimate Those are not peak-month figures and they are not a bad quarter. They are the rolling year to August 2026, which includes the July and August peak the whole island plans around.

The gap between 69 and 36 is not a rounding difference or a matter of good management. It is the difference between a business and a brochure.

Why the gap exists

The area spread matters more than the island average, and it moves in the opposite direction to what most decks imply. Uluwatu sells ten points more of its nights than Canggu does, which is the reverse of the marketing hierarchy, where Canggu carries the premium. The explanation is supply: Canggu absorbed the largest share of new completions and the booking window did not grow with it.

A buyer underwriting a Canggu villa on an island-wide figure is therefore already optimistic before any of the arithmetic below begins. A buyer underwriting an Uluwatu villa on the same figure is being unfairly penalised. Neither of them is being given the number that applies to the asset they are actually buying.

Three things happened at once, and none of them is visible from a listing page.

Supply grew faster than demand. Every villa completed in the last three years competes for the same booking window, and the window did not widen. Foreign arrivals to Bali fell 2.42 per cent in the first half of 2026◆ Verified — a small decline, but the first one outside a crisis year, and it landed on a much larger inventory than the last comparable period.

Second, the decks quote achieved nightly rates without quoting the discounting required to achieve them. A villa that holds its rate and sells 36 per cent of nights and a villa that cuts 30 per cent and sells 55 per cent of nights produce similar revenue. The first looks better on a rate card; the second looks better on an occupancy chart. Neither reaches 69 per cent at the advertised rate.

Third, and least discussed: the assumption was never tested because nobody publishes the denominator. Occupancy data for Bali is not a government series. It is assembled by companies that scrape the platforms, and until recently it was expensive enough that no buyer saw it before signing.

What the number looks like rebuilt

Start at the advertised 12 per cent and subtract, in order, the things a deck leaves out.

Occupancy gap. At 36.2 per cent measured occupancy against the 70 per cent the deck assumes, revenue is 48 per cent lower than the yield implies. Twelve per cent becomes 6.2 per cent before a single cost is deducted. This single line is larger than every cost that follows, combined.

Platform commission. Bookings that arrive through Airbnb and Booking carry a blended commission of about 16.5 per cent for this segment. On the reduced revenue, that removes 1.0 percentage point.

Management fee. A local manager charges 20 per cent of net-of-platform revenue for full service — guest communication, check-in, cleaning coordination, maintenance triage. That removes a further 1.1 points.

Withholding tax. A non-resident owner is subject to 20 per cent PPh 26 withholding on Indonesian-source rental income◆ Verified, applied to gross rent rather than to profit. That removes 1.2 points. Owners structured through a PT PMA face a different regime, which is a separate article and not always a better outcome.

Operating costs. Staff, pool chemicals, garden, utilities, linen replacement, and a reserve for the things that break in a tropical climate. Twenty-eight per cent of revenue is a defensible planning figure and a conservative one. That removes 1.7 points.

What is left is 1.2 per cent net, on a purchase price of USD 400,000. That is about USD 4,800 a year, before any allowance for the owner’s own travel, before currency risk, and before the cost of the capital sitting in the asset.

What about capital growth?

This is the standard reply, and it deserves a straight answer rather than a dismissal. If the asset appreciates, a thin yield is tolerable. Many asset classes work this way.

Bank Indonesia’s residential property price index rose 1.02 per cent year on year in the second quarter of 2026.◆ Verified That is a national index, it is broader than the villa segment, and it is the only official price series that exists. It does not support a growth assumption of 15 to 20 per cent, which is what the same decks print two pages after the yield figure.

There is a second problem with the growth argument in Bali specifically, and it is structural rather than cyclical. A large share of foreign-held villas sit on leasehold, not freehold. A leasehold asset does not appreciate on the same curve as the land under it; it amortises toward zero as the term runs down, and the decline steepens sharply once the remaining term falls below roughly ten years, because the buyer pool narrows to cash operators. Growth in the land value accrues to the landowner. This is not a defect in the structure — it is what a lease is — but it is rarely on the page next to the growth chart.

The case against this analysis

We should state the strongest version of the argument on the other side, because it is not weak.

Averages hide the top decile. Market-wide occupancy of 36 per cent includes badly positioned, badly photographed, badly priced villas that would not sell nights at any occupancy level. A genuinely well-run property with a strong direct-booking channel and a manager who prices dynamically can and does run materially above the market. We have seen operators sustain the high fifties. The honest version of our claim is therefore narrower than the headline: the market average does not support 12 per cent, and a buyer should not underwrite on the market average either. They should underwrite on their own operating plan, and then ask what happens if that plan is wrong by twenty per cent.

Direct bookings change the platform line. An operator who builds a repeat-guest list removes most of the 16.5 per cent commission. That is a real and achievable improvement, and it is the single highest-leverage change available to an owner. It is also a business, not a passive investment, and the decks sell it as the latter.

Our cost assumptions are choices. Twenty-eight per cent operating costs and a 20 per cent management fee are our planning figures, drawn from a portfolio and not from a public series. An owner who lives on the island, manages directly, and does their own maintenance runs a different model — and should say so out loud rather than compare their own result to a passive buyer’s.

The tax line depends on structure. PPh 26 at 20 per cent on gross applies to a non-resident individual without treaty relief. Several treaty positions and several corporate structures produce a lower effective rate. They also produce compliance cost, filing obligations and, in the wrong hands, exposure. Whether the trade is worth it is exactly the question a buyer should be asking, and exactly the question a yield figure on a cover page prevents.

The leasehold clock nobody prices

Most foreign-held villas in Bali are not owned. They are leased, usually for twenty-five or thirty years, sometimes with an extension clause whose enforceability depends on wording most buyers never read closely.

This changes the shape of the return in a way that a yield figure cannot express. A freehold asset has a terminal value. A leasehold asset has a terminal value of zero on a known date. Everything the buyer earns has to arrive before that date, and the asset’s resale value declines as the date approaches — not linearly, but on a curve that is shallow for roughly the first decade and then steepens.

The steepening has a mechanical cause. Below about ten years remaining, the buyer pool narrows to cash operators, because no lender will write a loan against security that expires inside the loan term. A narrower buyer pool means a wider bid-ask spread and a longer time to sell, and both show up as a discount.

Put that next to the 1.2 per cent net figure and the picture changes again. An owner earning 1.2 per cent on capital that is also amortising toward zero over twenty-five years is not earning 1.2 per cent. They are consuming principal and calling it income. The only version of this that works is one where the annual cash return is large enough to return the capital plus a premium inside the term — which is precisely the calculation the twelve per cent figure was invented to avoid.

There is a defensible version of the leasehold trade, and it is worth stating because we are not arguing that leasehold is a trap. It works when the buyer treats the lease as a depreciating operating asset rather than as property: when they underwrite to a target internal rate of return over the full term, when they price the extension risk explicitly rather than assuming it away, and when they plan the exit from the first year rather than the twentieth. Very few of the eleven decks framed the purchase that way. One did, and it was the only deck of the eleven that did not print a yield on the cover.

Why nobody publishes the denominator

It is worth asking why a market this large has no public occupancy series, because the answer explains why the twelve per cent figure survives.

Occupancy is the one number that cannot be spun. Nightly rate can be quoted as an aspiration. Growth can be quoted from a favourable window. Costs can be quoted net of whatever the writer chooses to exclude. But nights sold divided by nights available is a fraction with no room in it, and every party who could publish it has a reason not to.

The platforms have the data and sell access to it rather than publishing it, which is a reasonable commercial decision and also means the number reaches buyers late, if at all. Agencies have partial data from their own portfolios and no incentive to publish a figure that undercuts their own marketing. Managers have the best data of anyone, because they see the calendars, and they are contractually bound to the owners whose assets would look worse for it. The government publishes hotel occupancy, which is a different market with a different supply curve, and treating one as a proxy for the other has produced several confident and wrong articles.

That leaves the buyer to assemble it themselves from a paid aggregator, which most do not know exists, at a price most would not pay before deciding whether to spend four hundred thousand dollars. The information asymmetry is not a conspiracy. It is just what happens when the only parties who can measure something all lose by measuring it in public.

What a buyer should actually ask for

The point of this piece is not that Bali villas are a bad asset. It is that the number used to sell them is unfalsifiable as printed, and that a buyer can make it falsifiable with four questions.

Ask for the occupancy assumption as a number, in writing. Ask where that number comes from, and treat “our experience” as a refusal. Ask for the same calculation with occupancy set to the market figure for that specific area — not the island average, since Uluwatu and Ubud differ by ten points. Ask which costs are excluded from the yield figure, and specifically whether tax, management and platform commission are in or out.

An agency that can answer those four questions in a single email is worth talking to. An agency that cannot is telling you something useful anyway.

What we will publish next

We are going to keep the occupancy series updated quarterly and publish the methodology alongside it, including the cases where our figure disagrees with the platforms’ own dashboards and why. We will also publish the calculator that produced the waterfall above, so that a reader can put in their own purchase price, their own occupancy assumption and their own cost stack and see what falls out.

If you are an agency and you believe our occupancy figure is wrong for your properties, send us the data. We will publish it with your name on it, and if it changes our number we will say so in the corrections log.