How long should the test run?

Long enough to cover several full weeks of weekdays and weekends and the booking window for the dates you test. Around two months is a sensible minimum for most city properties; avoid a window that starts just before a big holiday.

What if I only have a few units?

With fewer than about six comparable units, a split is noisy. Compare each unit with the median of similar properties in your market for the same months last year instead, and read the result with care.

Can I use occupancy or average price as the measure?

Not on its own. Either can rise while revenue falls. Use revenue per available night, ideally after commission, and look at occupancy and price only to explain the result.

Test a pricing tool on half your units, not against last year

Answer

Split similar units into two groups: one priced by the new tool or manager, one left as it is. Agree the measure, the read date and what counts as a win before you start, and compare the change in each group, not this year against last year alone.

Why this year against last year is not enough

Comparing this year with last year mixes two things: what the new pricing did and what the market did. If the whole market rose by as much as you did, your gain proves nothing. If it fell, a small loss may be a good result. A control group answers the question the comparison cannot: what would have happened without the change?

In October 2026 an operator we spoke to proposed the cleanest version of this we have been offered. He had a block of near-identical apartments in one development, run by the same team for years. His idea: half of them priced by us, half left as they were, same building, same season, same guests. It is the design we recommend for any pilot where the units allow it.

How to check yours

Pick units that are alike: same building or area, same size, same quality, similar results last year.
Split them into two groups and check that both groups earned about the same per available night in the same months last year. If not, re-split.
Agree one measure: revenue per available night, ideally after commission. Count only nights that were open for sale, so an owner's blocked weeks do not distort it.
Agree the read date and what counts as a win before you start, in writing. For example: the test group's change beats the control group's change by a set number of points.
Change nothing else for either group during the test: same listings, same photos, same channels.
On the read date, compare the change in each group against its own last year. The difference between the two changes is the effect of the new pricing.

The number

Use this table on the read date. Fill it with your own figures; the verdict is the last row.

	Test group	Control group
Revenue per available night, same weeks last year	A1	B1
Revenue per available night, test weeks	A2	B2
Change	(A2 - A1) / A1	(B2 - B1) / B1
Effect of the new pricing	Test change minus control change

For how a result reads when the whole market moves, see the Berlin example in When the market falls, do you hold occupancy or rate?

When not to do this

If your units are all different, a split will compare unlike with unlike. Benchmark each unit against similar properties in the market instead.
If you have fewer than about six comparable units, results swing on a handful of bookings. Run longer or use the market benchmark.
Do not run the test across a change you cannot control for, such as a renovation, a new channel or a big event in only one group's dates.

Frequently asked questions

How long should the test run?: Long enough to cover several full weeks of weekdays and weekends and the booking window for the dates you test. Around two months is a sensible minimum for most city properties; avoid a window that starts just before a big holiday.
What if I only have a few units?: With fewer than about six comparable units, a split is noisy. Compare each unit with the median of similar properties in your market for the same months last year instead, and read the result with care.
Can I use occupancy or average price as the measure?: Not on its own. Either can rise while revenue falls. Use revenue per available night, ideally after commission, and look at occupancy and price only to explain the result.