Does Booking Three Weeks Earlier Actually Save Money? We Priced 4 Cities Across 10 Nights and the Median Moved 0%

Within the same month, booking 7 days ahead and booking 28 days ahead produced the same median price in Bangkok, Chiang Mai and Phuket — a 0 percent difference. Tokyo came out 2 percent cheaper closer to the date.

What actually moves the price is not how far ahead you book, it is which month you go. A Tuesday night in early December in Phuket costs 2.05 times a Tuesday night in early October, while Bangkok moved by 2 percent across the same window.

We did not ask anyone's opinion and we are not citing someone else's article. Every figure on this page comes from our own accommodation price feed, pulled on one day under one set of conditions.

• Pulled once on 29 September 2026 — 132 API calls in total, zero errors
• Identical conditions every time: 2 adults, 1 room, 1 night
• The nights measured are 10 consecutive Tuesdays, 6 October to 8 December 2026. All Tuesdays, to hold the day-of-week variable still — Friday and Saturday nights are always dearer than midweek, and mixing them makes the result unreadable
• Each city uses its 120 most-reviewed properties, and that same set is priced on every night — not a fresh search each time whose results are then compared
• We compare each hotel against itself as a ratio, then take the median of those ratios — we never subtract the median of one result set from the median of another
• The night of 24 November is excluded from the price calculation because it is Loy Krathong night in Chiang Mai, where so little inventory is left that averaging it in is meaningless. It gets its own section below

We compared the night of 6 October, which is 7 days ahead, against the night of 27 October, which is 28 days ahead, hotel by hotel. 416 properties matched across the four cities.

The median ratio came out at 1.00 in Bangkok, Chiang Mai and Phuket — no movement at all. Tokyo came out at 0.98, meaning 2 percent cheaper closer to the date.

The spread is more interesting than the median. 176 of the 416 properties, 42 percent, did not move by a single baht across four weeks. Another 137, or 33 percent, were cheaper closer to the date, and only 103, or 25 percent, were dearer.

In Phuket and Tokyo, nearly twice as many properties got cheaper closer to the date as got more expensive.

Stretch the window across all 10 nights, index the 6 October night to 1.00, and the picture changes immediately.

Phuket climbs from 1.00 in October to 1.64–1.67 in November and 2.05 in early December. Put another way: the same Tuesday night in early December costs twice what it costs in early October, while booking three weeks earlier returned 0 percent.

Chiang Mai rises more gently, from 1.00 to 1.13–1.24. Bangkok barely has a season inside this window at all, topping out at 1.02.

Tokyo runs the other way. Its index drops to 0.85 and 0.83 in early December, which is 15 to 17 percent cheaper than early October.

24 November 2026 is Loy Krathong and Yi Peng night in Chiang Mai. In that city's set of 120 properties, other nights still had roughly 111 to 118 available. That night had 6.

Filter to properties rated 8.5 or higher with 500 or more reviews — 48 of them in this set — and 24 November leaves exactly one.

We did not trust that figure from a single search, so we ran it again with a second method that does not use the original list at all. The feed still returned 134 properties for that night, similar to other nights, but a different set: only 4 of them were among our 120 most-reviewed, against 55 and 52 on 17 November and 1 December.

This is direct evidence for something we have said all along: a festival night does not push every room up in price, it removes the good rooms until only long-tail options are left.

Tokyo is the only one of the four cities where inventory genuinely thins out closer to the date. The night 7 days out had 93 properties bookable from the set; the night 35 days out still had 114.

Yet the rooms that remain are not dearer — the median ratio sits at 0.98. What booking late costs you in Tokyo is the number of options, not the amount you pay.

The other three cities show no such effect. Bangkok, Chiang Mai and Phuket keep roughly 111 to 119 properties bookable on almost every night, with Chiang Mai's festival night as the one exception.

If you only have time for five things, do these five.

• Pick the month before you pick the booking day — the month moves the price several times more than lead time does
• Midweek nights are always cheaper than Friday and Saturday, so compare the same weekday across several weeks
• If you are travelling on a festival night, book as early as you possibly can, because what disappears is the good accommodation, not the price
• If your plans are not settled, book a free-cancellation rate first and re-check later
• Always re-check the price near the travel date — a third of the properties in this set were cheaper closer to the day

This dataset has real limits and we are not cutting them out, because without them you would use it wrongly.

• This is a snapshot of the prices visible on one day, 29 September 2026, for various future nights — not a price for one night tracked all the way to check-in
• Because of that, the night 7 days out and the night 70 days out are different nights. Any gap you see mixes the effect of lead time with the effect of season
• The way we separated the two in this round is to look only at the four October nights on their own — same season, same weekday. What is left there is mostly the lead-time effect
• The price used is the per-night rate of the entry-level room the feed returns for that property, not every room type
• The set is each city's 120 most-reviewed properties, so it leans towards large, popular, affordable places — it is not the average of the whole city
• A real answer to how far ahead is cheapest needs several rounds of measurement. This is round 1

We intend to run the same set again with the same method, and update this page each round — saying plainly which round changed the answer and which did not.

In the meantime our booking-window page breaks the guidance down by city, and says openly that its figures are still general industry guidance rather than a result from our own data. This page is the first round of real data that will gradually replace it.