I'm 60% full for Saturday night. Is that good news or bad news? The honest answer is: "you don't know" — because 60% is a level, and a decision needs a direction.
The most common mistake in revenue management is reading occupancy as a number on its own. "I'm 60% full for next Saturday" sounds like information, but it is not decision-grade information. That same 60% can be excellent for one date and catastrophic for another. What makes the difference is comparing that 60% with where you normally are at the same number of days out.
This article covers the three concepts that turn a static occupancy number into an actionable signal: booking window, pace, and pickup — and why their strongest anchor, STLY (same-time-last-year), is both indispensable and dangerous.
1. Three concepts: level, position, momentum
Let's define the terms first, because most confusion comes from mixing them up.
- On-the-books (OTB): The reservations accumulated so far for a future date. This is a snapshot — a level. It's the raw number everyone watches.
- Booking window (lead time): The number of days between when a reservation is made and the check-in date. Measured per reservation, reported as an average.
- Pace: How fast reservations are building for a date compared with a benchmark — most often the same point in time last year. It answers: "is demand building faster or slower than it historically did at this lead time?" Pace is OTB in context; it turns a level into a trajectory.
- Pickup: The net change in reservations for a specific stay date over an interval. Daily pickup = today's OTB − yesterday's OTB. This measures momentum, not fullness.
The critical sentence: OTB answers "how full am I?"; pace answers "am I ahead of or behind the curve I'd normally be on at this lead time?" Because pricing and inventory decisions can only be made while the room is still for sale, only the second question is decision-grade.
2. Occupancy is not a number, it's a curve
The booking curve is the cumulative profile of how a date fills as check-in approaches. And it is not a metaphor — it's a measurable, near-universal shape.
Shintani and Umeno's 2023 study in Scientific Reports showed that average booking curves across perishable-inventory industries follow an exponential law: expected reservations behave as E[X(t)] ≈ A·exp(−βt) in the days remaining (t). Here A represents the magnitude of demand, and β represents exactly booking pace: a large β describes a date dominated by advance bookings, a small β one dominated by last-minute demand. Of the 24 property/period combinations examined, 18 (75%) fit this shape strongly.
The practical implication is striking: 60% OTB is a single point on a curve whose shape (β) and height (A) are what actually determine the outcome. Interpreting that point without seeing the curve is like judging a film from a single frame.
3. STLY — the best anchor, but fragile
The default benchmark for pace is STLY: the OTB position of the equivalent date one year ago, at the same number of days out. STLY controls for seasonality — you compare last August to this August, not to winter. But STLY does not control for demand shocks, moving holidays, day-of-week alignment, or shifts in the event calendar.
A concrete, verifiable example: Easter fell on March 31 in 2024 and April 20 in 2025 — the holiday moved from one month to the next. STR/CoStar data showed this mechanically depressed April 2025's year-over-year comparison (not because demand fell, but because the calendar shifted) while inflating the March comparisons. This is the textbook case for "STLY lies when the calendar moves." Reading a raw year-over-year delta without alignment is noise.
This fragility became obvious after COVID. Webb and colleagues (2020) showed that when the booking window itself shifts, some forecasting methods become unstable, whereas booking-curve-based methods stay more consistent. In other words, when the environment changes, blindly trusting "whatever happened last year" is the riskiest option.
4. The booking window shortened — read pace on today's curve
Knowing the range your booking window sits in is essential, because reading pace against the wrong curve misleads you.
The global average booking window is not a single fixed number: SiteMinder data puts the average at roughly 30–36 days depending on source and period, within a 20–60 day band, with a post-pandemic low reported at 22.68 days. The same dataset shows an average cancellation rate of about 19%.
But the average is only half the story. The last-minute tail grew markedly: per Amadeus (with UN Tourism), 51% of reservations in the Americas in early 2024 were made within a week of travel. D-EDGE measured that searches made within 28 days of stay rose from 32% of queries in Q1 2023 to 46% in Q4 2025.
The nuance here matters — and is usually missed: D-EDGE's 2026 distribution report found that although the last-minute tail grew, the average window actually lengthened from 2024 to 2025, and varies widely by channel (direct bookings carry a longer window and lower cancellation). Season is decisive too: in SiteMinder's summer data, Northern-Hemisphere summer stays are booked about 140 days in advance. So "windows are shortening" is true for the last-minute end, debatable for the mean, and outright wrong for resort/high-season.
The practical takeaway: read pace on today's (shorter, steeper) curve, not the 2019 one. If an urban hotel's demand structurally lands late, don't panic-react to a read that looks "slow" 30 days out.
5. Three things that distort OTB
To make pace reliable, you must be sure the two OTB figures you're comparing actually measure the same thing.
- Cancellations. OTB is a net figure: gross bookings minus a cancellation tail you can't see yet. With global cancellations near 19% and channel-dependent, today's OTB overstates the demand that will actually materialize.
- Channel-mix drift. A property whose OTA share has moved 10 points in 12 months can't use last-year-same-date OTB as a clean comparison — when the channel mix changes, so does the shape of the curve. STLY then compares two different curves.
- Unconstrained demand. If you sold out early on a high-compression date, that day's OTB is capped below true demand. Without accounting for turned-away demand (denial) and sell-out timing, you anchor next year's pace to an artificially low base.
Practical conclusion: no single signal is enough — blend them
- Never price off an occupancy level alone. Pull the pace/STLY position at the same days-out first. 60% is strong on a date where you're normally at 40%; it's an alarm on a date where you're normally at 80%.
- Align the comparison: match the day of week, shift moving holidays (Easter, Ramadan, Eid), correct for channel-mix drift — then trust STLY.
- Watch pickup (momentum) as much as pace (position). A date behind STLY but with accelerating daily pickup is delayed demand, not lost demand — a completely different pricing action.
- Blend the signals. Booking curve + live pace/pickup + aligned STLY + forward market/event data. The academic record supports this: Heo and colleagues' (2023) extended pickup method cut the mean absolute percentage error (MAPE) from 23.9% to 4.6% at a 28-day horizon — and concluded that historical data still works even in uncertain times.
Summary: Occupancy tells you how full you are; pace tells you whether you're ahead of or behind the curve you'd normally be on. Only the second is decision-grade, because pricing decisions must be made while the room is still for sale. 60% isn't an answer, it's a question: "On this date, at this lead time, where would I normally be?" A good revenue manager looks not at a number but at a trajectory — and reads that trajectory only after aligning it for day of week, moving holidays, and channel mix.
Sources
- Shintani, M. & Umeno, K. (2023). Average booking curves draw exponential functions. Scientific Reports, 13, 15773. (Exponential booking-curve law; β = booking pace)
- Webb, T., Schwartz, Z., Xiang, Z. & Singal, M. (2020). Revenue management forecasting: The resiliency of advanced booking methods given dynamic booking windows. International Journal of Hospitality Management.
- Heo, C. Y., Viverit, L. & Pereira, L. N. (2023). Does historical data still matter for demand forecasting in uncertain and turbulent times? An extension of the additive pickup time series method for SME hotels. Journal of Revenue and Pricing Management. (MAPE 23.9% → 4.6% at 28 days)
- SiteMinder — Hotel Booking Trends and Booking Lead Time (global average window ~30–36 days, 20–60 day band, post-pandemic low 22.68 days; cancellations ~19%)
- Amadeus + UN Tourism (January 2024, Americas) — 51% of reservations made within a week of travel
- D-EDGE — 2026 Hotel Distribution Report (within-28-day searches 32% → 46%; average window lengthened 2024→2025; direct = longer window / lower cancellation)
- STR / CoStar — Easter 2024 (Mar 31) vs 2025 (Apr 20) calendar shift and the April 2025 year-over-year distortion
- Cornell Center for Hospitality Research — work on aggregating forecasting methods (Anderson & Xie; Thompson)