Across 449 bookings at our own three buildings and six rooms, over the twelve months from August 2025 to July 2026, the average booking ran 3.3 nights (measured 21 August 2026, published with the calculation method at operating data).
Occupancy over the same window was 83.6%. In a 30-day month that is about 25.1 nights sold per room, and split into stays of 3.3 nights it means 7.6 changeovers a month. Every argument about length of stay eventually reduces to that count.
What is average length of stay, and why does it reach the P/L?
The mean number of nights in one booking, which for us was 3.3. It reaches the P/L because the larger costs are incurred per booking, not per night.
A clean happens once per booking, not once per night. Linen, restocking and the post-stay check work the same way: whether the guest stayed three nights or five, the work is one event after they leave. Revenue, meanwhile, broadly scales with nights.
So selling the same 25.1 nights as 7.6 bookings or as 5 bookings leaves revenue roughly unchanged while changing the number of cleans. That is the entire reason length of stay is a financial question.
How does cleaning as a share of revenue move with stay length?
Converting from our own 17.5%: a 2.5-night average puts it at 23.1%, a 4-night average at 14.4% (the line itself is broken down in cleaning, the largest cost line).
The arithmetic is 17.5 x 3.3 / N. Changeover count is inversely proportional to stay length, and the cost share follows the same curve.
| Average stay | Changeovers per 30-day month at 83.6% | Cleaning as a share of revenue (converted from our 17.5%) |
|---|---|---|
| 2.0 nights | 12.5 | 28.9% |
| 2.5 nights | 10.0 | 23.1% |
| 3.0 nights | 8.4 | 19.3% |
| 3.3 nights (ours) | 7.6 | 17.5% |
| 4.0 nights | 6.3 | 14.4% |
| 5.0 nights | 5.0 | 11.6% |
| 7.0 nights | 3.6 | 8.3% |
What assumption is buried in that table?
That the nightly rate does not change. Break that and the table stops being usable.
In practice long bookings arrive in exchange for a weekly or monthly rate, so as nights go up the revenue in the denominator comes down with them. The table shows only what happens to the cost ratio if stay length moves on its own, and says nothing yet about whether lengthening stays is worth doing.
To answer that, the changeover being saved has to be expressed in the same unit as the discount being offered.
What is one changeover worth, expressed in nights?
About 0.58 nights, on our numbers. Cleaning takes 17.5% of revenue across 7.6 changeovers, so one changeover is 2.3% of the month, and the month is 25.1 nights: 25.1 x 0.023 = 0.58 nights.
It is written in nights rather than yen because a ratio travels and a price does not. Cleaning prices and nightly rates differ by property, but "one changeover costs a little under six tenths of an average night" is something an owner can recompute in a minute.
That 0.58 is the yardstick for everything below. Removing one clean is worth exactly that much, no more.
How deep can a long-stay discount go before it stops paying?
About 8.8% on our numbers. Merging two 3.3-night bookings into one 6.6-night booking removes one changeover and frees 0.58 nights, while the discount applies to all 6.6 nights, so a d% discount gives away 0.066d nights. Break-even is 0.58 / 0.066 = 8.8%.
| Discount on a 6.6-night stay | Given away (in nights) | Against the 0.58 nights saved |
|---|---|---|
| 5% | 0.33 nights | +0.25 nights |
| 8.8% | 0.58 nights | break-even |
| 10% | 0.66 nights | -0.08 nights |
| 15% | 0.99 nights | -0.41 nights |
| 20% | 1.32 nights | -0.74 nights |
| 30% | 1.98 nights | -1.40 nights |
In which months does that calculation stop holding?
In the months that do not fill. The 8.8% break-even rests entirely on the assumption that turning down a 6.6-night booking still leaves you two 3.3-night bookings.
An 83.6% average mostly satisfies that assumption, but individual months do not. Our weakest month across the twelve was Sapporo in November 2025 at 68.3%, followed by April 2026 at 71.1% (how far occupancy swings month to month). At 68.3%, about 9.5 nights per room are sitting empty.
In a month like that, a long-stay discount is not buying back a saved changeover, it is buying the nights themselves. When the alternative is an empty room rather than a second booking, 8.8% is the wrong test entirely. The reverse is just as true: a deep long-stay discount in a strong month is selling bookings you already had at a lower rate.
The same discount is right in one month and wrong in another, so we do not carry one as a year-round setting.
Where do long bookings damage occupancy?
At the minimum-stay setting, and in the fragments left between bookings. Raising the minimum to chase longer stays makes those fragments unsellable.
Occupancy of 83.6% means 16.4% of nights are empty, about 4.9 nights per room in a 30-day month. They do not sit as one block: they scatter across 7.6 joins, averaging 0.64 nights per join. The typical gap is therefore one or two nights, and a three-night minimum makes it unsellable at any price (the arithmetic of a late discount).
The second cost is shape. A seven-night booking landing mid-month leaves a three-night and a four-night window either side. Those will sell, but the month can no longer be two five-night stays. A long booking lowers cost by spending calendar flexibility in advance.
What we do with 3.3 nights, and what we will not promise
We do not run a target for average stay length. What we watch is the combination of changeover count, rate and occupancy; 3.3 nights is an output of that, not an input to it.
In practice this means moving the minimum stay by month. In months that fill we leave short bookings alone; in weak months we welcome long ones. Long-stay discounts are set month by month once we can see how the calendar is filling, and the 8.8% is a yardstick for that judgement, not a rule to hold all year (the full operating cost breakdown).
We do not tell an owner what their average stay will be. 3.3 nights is what our own six rooms in Tokyo and Sapporo recorded over twelve months, and it moves with layout, capacity, distance from the station and guest mix. What we commit to is measuring the property's own average stay and its own cost-per-changeover in nights during the first months, and setting minimum stay and long-stay pricing from those figures.