In Split the order of decisions matters more than any of them individually. Set the week first: the long Marjan series gives 17 days above 25°C in September and one in October. Then the area, then the room. Reversing that order is how good bookings go wrong.
Three decisions, and only one order that works
A Split booking contains three choices — which week, which part of town, which property — and they are not independent. The week determines how much choice you have in the other two, so it has to be settled first. Most people settle it last, because the property is the thing with photographs.
The week also has the largest effect. Split airport handled roughly 578,000 passengers in June 2025 and 62,250 in November 2025, and the city’s room stock is fixed. Moving your dates by three weeks changes what is available to you far more than moving your budget does.
What the long record says about September
Split has an unusually long instrumental record at the Marjan station, running from 1948 to 2024. It is a different dataset from the 1971–2000 national normals and the two should never be blended in one table — but read on its own it is the best guide there is to how a Split month behaves.
| Month | Mean °C | Sunshine hours | Days ≥25°C | Rain days |
|---|---|---|---|---|
| Jan | 8.0 | 131.5 | 0 | 11 |
| Feb | 8.5 | 146.0 | 0 | 10 |
| Mar | 10.8 | 188.6 | 0 | 10 |
| Apr | 14.4 | 219.7 | 0 | 10 |
| May | 19.2 | 271.4 | 9 | 9 |
| Jun | 23.4 | 309.0 | 23 | 8 |
| Jul | 26.2 | 353.3 | 30 | 5 |
| Aug | 25.9 | 328.0 | 29 | 5 |
| Sep | 21.6 | 247.1 | 17 | 8 |
| Oct | 17.2 | 198.9 | 1 | 9 |
| Nov | 12.7 | 130.5 | 0 | 12 |
| Dec | 9.4 | 120.8 | 0 | 12 |
September returns 17 warm days and 247.1 hours of sunshine; October returns one warm day and 198.9 hours. That is the cliff, and it falls inside a fortnight. The 1971–2000 normals put the same September figure at 15.4 — a different period producing a slightly different answer, which is exactly why a page should tell you which series it is quoting.
Split Marjan monthly sunshine, hours. July 353.3; November 130.5. Source: DHMZ (meteo.hr), 1948–2024.
Notice how asymmetric that curve is. The rise from March to June is gradual; the fall from September to November is abrupt. A week moved earlier in the spring costs you little. The same week moved later in the autumn costs you a great deal, and it is also the direction in which the ferry network thins.
Then the area, then the room
Once the week is fixed the second decision becomes tractable. In high summer, availability near the ferry terminal is the constraint and everything else bends round it. In the shoulder the constraint disappears and the sensible choice is whatever is warm, quiet and close to somewhere that is still open.
The property comes third and it should. By the time the week and the band are fixed, the list is short and the remaining variables are the ordinary ones: staffed reception or not, air conditioning that someone else maintains, and whether you can get a bag up the stairs.
Why the order is not arbitrary
It would be reasonable to ask why the week has to come first rather than simply being one of three equal choices. The answer is that the three decisions constrain each other asymmetrically. The week determines what is available in every band of the city; the band determines which properties are in play; the property determines nothing about the other two.
Information flows one way through that chain, so working against it wastes the work. Anyone who has picked a property first and then discovered it is only free in the week they did not want has run the sequence backwards and paid for it.
There is a second reason specific to Split. The city’s summer is defined by a fixed room stock meeting a demand curve that moves by a factor of nine between June and November. Under that kind of pressure, the week is not one variable among several. It is the variable, and everything else is a rounding error against it.
Booking the first week of October on September’s reputation
The Croatian shoulder has a good reputation and it is deserved — in September. The long Marjan series records 17 days above 25°C in September and one in October, and 115.5 mm of rain arrives in November.
The other half of the cliff is the timetable. Thirteen of the twenty-three services calling at Split are commercial lines with no public service obligation, and they are scheduled to summer demand. An October booking that assumed September’s weather usually also assumed September’s boats.
It is worth being explicit about what this page is refusing to do. It does not name properties and it does not quote rates, because neither can be verified for the week you are travelling and both go stale within a season. What it does instead is hand you the two series and the register, which do not go stale, and let you apply them to your own dates.
Fix the week against the series above, fix the band against how you are arriving, and the property question becomes small. That is the whole method, and it is the opposite of how a best-of list asks you to work.
Also on Hotels.com
What do you need to work out next?
Order settled. These are the next three steps.
Where to stay
Which band of the city?
Split’s areas divide by walking minutes to the ferry terminal.
Getting around
How do the flights follow the season?
578,000 passengers in June 2025 against 62,250 in November.
When to go
Is September genuinely better than August?
Seventeen warm days against twenty-nine, and far fewer people.
Planning
What do I do with the days?
Six island groups have scheduled services from the same quay.
Climate data: DHMZ (meteo.hr), Split Marjan station, 1948–2024. Passenger figures: Državni zavod za statistiku (DZS), monthly airport traffic releases. Ferry data: Agencija za obalni linijski pomorski promet (ZOLPP), 2026 state coastal line register and per-line timetables.