Read the booking window as a distribution
To examine early and last-minute demand together, consider when a reservation was first made and when the guest will arrive. One average can hide a mixture of same-day and far-ahead bookings. Calculate the days between original booking and arrival, group them into explicit bands, and separately track snapshots for the same future arrival date. This separates “how early do guests book?” from “how many rooms did this date gain recently?” The following framework supports a decision table, not a promised price change or revenue increase.
Two clocks answer different questions
Cloudbeds’ public data dictionary defines booking window as days between booking and check-in. It is not length of stay. Its Rate Evolution help presents pickup through daily rooms booked and rate categories as arrival approaches. Those sources explain terminology, without recommending a product or claiming a Stayax capability. Here we additionally define net pickup as the change between two snapshots for one future arrival date. Put your exact definition in the report heading. Do not replace the original booking timestamp with the latest edit timestamp.
Fix the sample before comparing
State whether a row represents a reservation or a room. A multiroom group and an individual booking answer different questions when weighted equally. Begin with single-room records and a consistent status set. An arrival-date cohort is different from all reservations created this week, which may concern many future dates. Use the property’s local date boundary and check missing, imported or backdated timestamps. Our example bands are 0–7, 8–30 and 31+ days: working groups, not universal quality standards. Apply the same bands again when comparing earlier reports.
Fictional example: mean 14.8, median 6 days
These are invented values, not property or guest records. Ten comparable single-room reservations have booking windows of 0, 1, 2, 3, 5, 7, 14, 21, 35 and 60 days. They total 148 days; divided by ten, the mean is 14.8. The middle values are 5 and 7, giving a median of 6. Six bookings, or 60%, fall in 0–7 days; two, or 20%, in 8–30; and two, or 20%, in 31+. Saying “guests book about two weeks ahead” hides the majority in the final week. Median alone hides the long-window records. Show mean, median, band counts and sample size together. Ten records cannot establish future demand.
Separate example: net pickup for one date
In this example there are no continuing stays from other arrival dates or sales blocks; every record is for one room and one night. Consider a different invented scenario. For arrivals on 20 November, a 20-room hotel has 8 rooms booked in the 6 November snapshot and 12 in the 13 November snapshot. Net pickup is 12 − 8 = 4 rooms. If five new single-room bookings were added and one room cancelled, report gross additions 5, cancellations 1 and net change 4 separately. Date moves or room-count edits also need explanation. Capacity is fixed here; eight remaining rooms do not forecast future sales. Comparing different arrival dates or room categories would invalidate the snapshot comparison.
Turn the table into questions
For each arrival date, record days remaining, rooms currently booked, recent net pickup, comparable past observations at the same point and unresolved questions. Matching weekdays alone is insufficient when season or announced events differ. Neither a long window nor a short one is automatically good or bad. If bookings appear low, first check whether sale conditions and the offered room are correct. If pickup accelerates, distinguish one group from several independent bookings. Give each finding a review question and next check time. This organises evidence before a decision; it is not an automatic pricing rule. More booked rooms do not establish higher revenue or profit without their price and included services. Review remaining inventory, the rate scope offered for that date and total reservation value separately before deciding to keep or change a price.
Start with a small weekly routine
Choose a limited set of future dates and capture them on the same weekday and time. Next week, check changed definitions, unexplained movements and excluded records with missing dates. Use an aggregate version without personal data. Do not substitute these fictional figures for actual bookings. The first benefit is turning “demand arrives late” into a defined question. A few snapshots cannot prove a sector trend or guaranteed revenue impact. Extend comparisons only with consistent scope, using the linked analysis archive for calculation discipline. Record reasons for keeping a decision unchanged as well as reasons for acting.