Should the weather move your nightly rate?
· 5 min read
It depends on the property — and that is a measurement, not a dodge. The first version of this post, in August 2026, said that we collect forecasts every day and still keep them out of the price, because nobody had measured how much weather really moves demand. We measured it. Below is what came out, together with the reason we still say nothing about weather to some hosts.
Why it sounds obvious
Because in a sense it is. A rainy weekend in the mountains means fewer people packing for a hike. Thirty degrees in mid-June means a sudden rush to the coast. Every host knows these stories and every host can name a week the weather turned upside down.
The trouble starts at “how much”. Ask whether weather matters and everyone nods. Ask by what percentage to raise Saturday’s rate if the forecast says sun and 26 degrees, and the room goes quiet — rightly so, because that question is answered by measurement, not conviction.
What we did instead of guessing
We asked the question separately for each property, against its own history. Not “does rain hurt accommodation in general”, but: did nights wetter than usual sell worse than the rest at your place. The answer comes from your calendar and from forecasts recorded on the day guests were deciding — not from the weather that eventually showed up.
That difference is the point, not a technicality. A guest books on what they saw a week earlier. If the forecast was wrong, their decision was still made on it — and we price guests’ decisions, not the weather.
“Wetter than usual” also means wetter than usual for you, not past some fixed number of millimetres. The same number means one thing on the Baltic coast and another in the mountains, and a threshold shared by both would be convenient for us and untrue for either.
The interesting part: we measure the direction too
Industry intuition says precipitation always hurts. In the mountains in winter, precipitation is snow — and snow is a reason to come, not to cancel.
So our measurement allows the counter-intuitive result and treats it as correct rather than as a bug to fix. Had we assumed that wet means worse, we would be cutting rates in ski towns exactly when they should rise — and it would have looked like a thoughtful algorithm.
When we say nothing about weather at all
When your history is too thin to tell weather apart from coincidence. The factor then stays silent — it does not “take a neutral value”, it simply does not appear in the explanation of your price.
The same happens when there is nothing to compare: if every night in the window was similarly dry, there are no two groups. We will tell you what is missing instead of filling the gap with a number that looks clever. The “what switches on next” plan in your panel says so outright.
Why we did not add this factor by feel
Because it is the easiest thing in the world and would look better than anything else we do. “Raise the price 8% because sun is forecast” sounds smarter than half our methodology. It would also be a number pulled out of thin air.
A host who accepts that price has no way to check whether the 8% came from a measurement or from imagination. All they see is a tool saying something specific, and specifics earn trust. That is exactly why an invented specific is worse than none: it buys trust you cannot buy back.
Nor is our correction a separate “sunshine bonus” bolted on the side. Weather reaches the price by the same route as every other recognised demand signal — because we have one answer to “how much of the demand we recognise may be turned into price”, not a different one per cause.
What this post says about us
The first version ended with a promise: when we measure it, we will write about it separately — including if the effect turns out smaller than everyone assumes. This is that moment, and the result is less spectacular than the promise: weather matters at some properties, in a direction that depends on where they are, and only when there is enough to compute it from.
We kept this post at the same address instead of publishing a new one beside it. A text that predicts its own expiry should be checkable where you read it.