Booking window analysis: how far ahead each segment books, and what to do with it

A booking window is the number of days between the day a reservation is made and the day the guest arrives. Read it as a distribution, not an average: the share of room nights booked 90, 60, 30 and 7 days before arrival, split by segment, channel and source market. Each curve supports a different decision.

Two hotels can report the same average booking window and need opposite calendars. One must have next season's rates loaded by March; the other can wait until four weeks before arrival. The average is the reason nobody notices the difference. This guide shows you how to measure the hotel booking window properly: which fields to export, how to build the distribution, how to read it by segment and channel, and which single decision each finding should change.

What is a booking window, and why does the average mislead?#

A booking window is the number of days between the booking date and the arrival date. The average of that number tells you very little, because most hotels book in two or three clumps rather than one.

Fix the convention once: lead time equals arrival date minus booking date, counted in days, and every report in the hotel uses the same rule. A report that mixes nights and days cannot be compared with last year. Write the definition into your reporting notes, because the same word is used loosely across the industry, and a quiet change of definition breaks year-on-year comparison.

The average is dragged upward by a small number of very long bookings, usually groups and contracted allotments. A hotel with an average of 45 days can still take half its room nights inside 14 days. Use the median and the distribution together: the median tells you the typical booking, the buckets tell you where the mass sits.

Keep three ideas separate, because they are often confused. The booking window is a property of one reservation. The booking curve is the cumulative share of room nights already on the books at a given distance from arrival. Pace is how the on-the-books position for a stay date changes from one week to the next. Two hotels with identical averages can need opposite calendars: one is decided early and needs rates ready months out; the other is decided late and needs a close-in pricing plan. For the surrounding vocabulary, see revenue management terms explained, from pickup to displacement.

Which fields do you need before you can measure lead time?#

Five fields per reservation: booking date, arrival date, room nights, segment and channel, plus source market where you record it. Everything else is arithmetic.

Take the data from reservation records, not from a rate sheet or a memory of how the season felt. The analysis is only as good as the booking dates in the system. Use room nights as the unit: counting reservations gives a 14-night group the same weight as a single Saturday, which is how a curve gets misread.

Handle modified reservations explicitly. If a guest moves the arrival date, the current booking date describes a different decision than the original one. Keep the original booking date in a separate field if your system allows it. Treat cancellations as their own series: a cancelled booking with a 120-day window still influenced the rates you loaded and the rooms you held, so removing it from the file hides part of the story.

Export a dated snapshot every month. Reservation records are edited in place, and a curve rebuilt next year from that year's data will not match the one you acted on. Finally, keep personal data out of the analysis file. Segment, channel and market are enough; you are measuring timing, not profiling guests.

How do you read the booking curve as a distribution?#

Sort every room night into lead-time buckets, then add the buckets from the longest to the shortest to get a cumulative curve.

Choose bucket edges and keep them: for example 90 or more days, 60 to 89, 30 to 59, 15 to 29, 7 to 14, and 0 to 6 days before arrival. Changing the edges between seasons makes comparison impossible. Read the cumulative version when you plan: by day 60, a given share of the final room nights was already committed, and that share tells you when the calendar was effectively decided.

Compare the same arrival period across years, never the whole year at once. A summer month and a February month have different curves in the same hotel. Look at the shape before the numbers: a front-loaded curve means the hotel is decided early and late discounting mostly gives away margin; a back-loaded curve means the last three weeks carry the month. Watch the final bucket separately, because the share of room nights arriving inside 7 days is the room you have left to price.

Read the curve with a person, not a model. It is a description of what happened, and interpreting it is a judgment about contracts, campaigns and holidays. The table below is an illustrative example for one fictional resort, not a benchmark.

To turn the curve into a weekly operating rhythm, pair it with how to read booking pace day by day with a pickup report.

Why do tour operators, booking sites and direct guests draw different curves?#

Each channel carries a different kind of commitment: contracted allotments are taken months out, booking sites sit in the middle, and direct business splits between early planners and last-minute guests.

Tour operator business often produces the earliest part of the curve, but what you are seeing may be a contract release date rather than guest demand. Check the contract before reading the bucket as a market signal. Booking site business reflects when travelers shop and what they can see: your rate position, your photos and your availability at that moment. A weak early curve on this channel is sometimes a visibility problem, not a demand problem.

Direct business splits. Repeat guests and campaign subscribers book early; local, drive-in and decision-late guests book close to arrival. One blended direct curve hides both. Domestic and foreign business usually differ in timing, but read your own records: no external figure tells you how early guests from any country book, and importing an assumption is how a calendar gets built on nothing.

Segment before you conclude. A blended curve can look comfortably early while one channel has already taken most of its rooms and another has taken almost none. Then ask a counterfactual: what would this curve look like if one channel disappeared next year? That question turns a report into a distribution decision.

How do bayram and early-booking campaigns bend the curve?#

A fixed-date holiday pulls bookings into a narrow window before it; a campaign pulls them forward across the whole calendar. Both change the shape, and only one may change total demand.

Religious holiday dates move each year, so compare the holiday period with the same holiday in an earlier year, not with the same calendar month. Expect a spike when dates are announced and a second wave close to the date. A curve with two humps around a holiday is normal and should not be read as a data error.

For an early-booking campaign, measure two things: whether the curve shifted left, meaning bookings moved earlier, and whether it rose, meaning more room nights in total. A shift alone is a cash-flow and commitment decision, not proof of extra demand. Set release dates and restrictions before the campaign starts, because changing them afterwards makes the campaign result impossible to attribute.

Watch what the campaign does to the cancellation series. Bookings pulled forward at a discount can also be cancelled earlier, and the curve alone will not show it. Record the campaign dates in the same file as the curve, so next season's comparison knows what bent it. For the pricing side of the same question, see when early booking and non-refundable rates pay off.

Which decisions does the curve support?#

The curve tells you when rates must be ready, when contracted rooms should come back, and how hard to hold price in the final three weeks.

Rate loading comes first. If a meaningful share of room nights arrives before day 90, next season's rates must be loaded before that point, not in the month before arrival. Release dates come from the observed curve of that segment: a release date later than the segment's booking mass leaves you holding rooms nobody can buy. Restrictions, meaning minimum stay and closed-to-arrival, belong where the curve is steepest and the hotel is filling, not on every date by habit.

Close-in pricing follows the final bucket. A large final-week share argues for holding rate and stopping discounts earlier; a small one argues for a deliberate close-in offer instead of an empty room. Displacement is the quiet one: a long-window group blocks rooms that late, higher-rated demand might have taken, so check the curve for those dates before accepting the block. And keep the discipline of one finding, one decision: a curve that produces five recommendations produces none, because nobody owns them.

To extend the curve into a forward view, see estimating the next 90 days from what is on the books.

How do you repeat the analysis each season with the same method?#

Fix the definitions, run it on the same dates each month, and compare like with like.

Write the bucket edges, the segment names and the counting unit into a one-page note the team uses. Definitions drift faster than data. Run the analysis monthly on the next 12 months of arrival dates, and once more after the season closes, when the final room nights are known.

Keep the dated exports, because a curve cannot be rebuilt from records that were edited afterwards. Compare year over year for the same arrival month, and note what changed: a campaign, a contract, a release date, a holiday that moved. Hand each run to one decision, and record what was decided. The value of the analysis is in the calendar it changed, not in the chart.

What to do next#

Export the last twelve months of reservations with five fields: booking date, arrival date, room nights, segment, channel. Build the bucket table once, and read the shape before you read the average. Everything else in this guide follows from that file: the segment curves tell you when each channel commits, the calendar effects explain the humps, and the decisions about rate loading, release dates and close-in pricing all rest on the same distribution. Repeat it on the same dates next season, with the same definitions.