Occupancy is a rear-view mirror. By the time it tells you demand shifted, the booking window for those nights has already moved on — the guests who were going to book have booked, somewhere. Forecasting moves the information to where it is useful: ahead of the decision.
The booking window is the whole game
Every night you sell has a window during which it can be sold, and that window has structure. Some markets book months out; others fill in the final two weeks. The same market shifts by season, and holiday weekends behave differently from ordinary ones. Pricing decisions made outside the window are wasted, and decisions made late in it are made from weakness. Knowing where you are in the window is the precondition for pricing well inside it.
Pace: the leading indicator that matters most
The workhorse of demand forecasting is pace. Pace means 7-day pace: this year’s 7-day pickup minus last year’s 7-day pickup for the same stay dates at the same number of days before arrival. It is described as positive or negative. Positive 7-day pace means demand is arriving earlier than it did last year; negative 7-day pace means the nights are at risk. Occupancy tells you what happened. Pace tells you what is happening, while there is still time to respond.
Two neighbors of pace are easy to confuse with it. On the books vs. last year is the year-over-year gap in position, not pace. Market pace compares your 7-day build this year with your comp set’s 7-day build this year, both from the same year; it is not year over year. The comp side aggregates all the comps’ data for what you have selected: a listing, a group or the portfolio. 7-day pace, market pace, market occupancy and last-year occupancy all sit on the Insights page, and market pace is on the Occupancy KPI card.
What feeds a real forecast
A useful forecast blends more than your own booking curve:
- Market pickup — when the listings around you start filling, your remaining nights change value.
- Seasonality — the repeating annual shape of demand, separated from one-off noise.
- Day-of-week structure — Tuesdays and Saturdays are different products with different curves.
- Events — a citywide event compresses supply and steepens the price guests will accept.
- Lead-time distribution — how far out your market actually books, which sets your decision deadlines.
From forecast to action
A forecast earns its keep in the response it enables. Positive 7-day pace is a signal to hold or raise: early demand is telling you the rate may be too low. Negative 7-day pace calls for targeted moves: adjust the soft nights, revisit minimum-stay rules that block the bookings actually being searched for, fix the gap, rather than blanket-discounting the month. Market pace tells you whose problem a soft stretch is: if your comp set is building at the same 7-day rate as you, the whole set is soft; if it is building faster, look at your price, stay rules and listing. The difference between a two-week head start and a two-day panic is usually the difference between a small correction and a margin-destroying one.
Forecasts must be graded
A forecast you never check is an opinion. Comparing forecast to actual — and tracking that error over time — is what separates a demand model from a guess, and it is the foundation for evaluating whether any pricing system is working. See it early, price it right, and grade the forecast afterward: that loop is where the compounding happens.
How seasonality is set
Seasonality is the repeating annual shape of demand, and at Quibble it is not hand-picked. Bookable Search™ counts only the searches that ended in a purchase: your listing appeared in the search and the shopper booked something very similar. The bookings predicted from those searches, across your comp set, set the seasons. Revenue managers do not decide how many seasons a property needs.
It also answers an old data-selection problem. Using all of a property’s history can miss new trends; using only recent months can overreact to a disruption such as a pandemic. A forecast built on purchase searches across a comp set leans on the market’s current behavior rather than on one listing’s short or noisy history.
Thin data: forecasting across the comp set
A single rental rarely has enough bookings to forecast on its own. A 300-room hotel can absorb a 10% forecast error across its rooms; one rental is either booked or vacant. So Quibble predicts total demand across a small comp set and splits it by choice probability: expected bookings = set demand × your choice probability. About 30 comps are attached to each listing, and for each night the 10 best matches that are still available go into the model.
You never see the road not taken
When a property sits vacant at $500 a night with a 3-night minimum, you cannot know what would have happened at $450 with a 2-night minimum. Forecasting rental nights means estimating those unobserved scenarios statistically: the probability a night books at each price. That probability, not the price, is where the science sits.
Go deeper: Introducing Quibble Bookable Search™, Reading the comp set, How optimization models work, The case against last-minute discounting.