Skip to content
+358 50 464 7513 [email protected] Rovaniemi · Remote-first
  • EN
  • FI
  • EN
  • FI
Book a Free Analysis
Revenue Management & Pricing

Demand Forecasting for Hotels: Methods and Data Sources

Demand forecasting for hotels explained: what feeds a forecast, how the pickup method works, constrained vs unconstrained demand and how to check accuracy.

Kraków spires silhouetted against an orange sunset, with light bars

Picture a city-centre hotel in Kraków setting its rates for a week in October the same way as every other week that quarter: last year’s booking pace plus a small rise for inflation. Last year’s numbers don’t show that a large conference is coming to the city. Attendees book through the convention centre’s own housing block, so the demand never appears on the hotel’s booking curve until people start calling the front desk. By the time the pattern is clear, the hotel has sold out at rates set for an ordinary week, while competitors’ rates on Booking.com climb well above its own. The rate-setting process is careful. The demand estimate underneath it is already out of date.

A sound pricing strategy can still get a date badly wrong if the demand estimate rests on the wrong inputs. This article covers demand forecasting for hotels: the data that feeds a forecast, the most common method, the difference between constrained and unconstrained demand, where forecasts break and how to check accuracy over time.

What Feeds a Demand Forecast

No single input is enough on its own.

  • Historical pace and pickup. How a date is booking today compared with similar dates at the same point in earlier years, for example 30, 60 and 90 days before arrival. This is the backbone of most forecasts and comes from the property management system (PMS) or channel manager.
  • Booking window trends. The booking window is how far ahead of arrival guests book. A pace comparison only holds if the window is stable. If guests are booking closer to arrival than last year, a date will look weaker than it really is.
  • Local events calendars. Conferences, festivals and sports fixtures that pull demand into specific dates. This input goes stale most easily, because someone has to check venue calendars by hand.
  • Competitor rates. Rates from comparable hotels, usually collected from Booking.com, Expedia and the hotels’ own websites, give a rough sense of whether a shift in demand affects the whole market or just one property. A displayed rate shows what a competitor asks, which may differ from what it sells.
  • Wider travel trends. Regional arrival figures, flight capacity and search interest in the destination. These move slowly but shape forecasts several months out.

Revenue management systems (RMS) such as IDeaS, Duetto and Atomize (owned by Mews since 2024) combine pace and pickup data automatically and can add competitor rate feeds. Ask how each handles local event data. Our guide to revenue management software covers what these systems do more broadly.

The Pickup Method

Pickup is the number of new bookings a date gains between two points in time. The pickup method turns that into a forecast. Take the rooms already on the books for a future date. Then add the average number of rooms that similar past dates picked up between the same point and arrival day.

Say a Saturday 30 days out has 60 rooms booked, and comparable Saturdays picked up 25 more in their last 30 days: the forecast is 85. This is the additive version. A multiplicative version applies the average growth rate instead.

In a study of hotel data published in the International Journal of Forecasting, Larry Weatherford and Sheryl Kimes found pickup methods among the most reliable, alongside exponential smoothing and moving averages (Weatherford and Kimes, 2003). Its weak point: it assumes the future booking curve will look like the past one.

Unconstrained vs. Constrained Demand

Constrained demand is what a hotel sold on a given date, limited by the rooms it had. Unconstrained demand is an estimate of what it would have sold with no limit on rooms. It shows how much demand the market really had for that date.

The gap between the two stays hidden unless a hotel looks for it. Once a date sells out, it stops collecting new bookings. A forecast built only on past sales therefore can’t see the demand above the cap. In a case like Kraków, a past sellout looks in the data like a strong but ordinary week, with no sign of the guests who were turned away.

A common way to estimate unconstrained demand is to log turned-away requests on dates that close. These come in two kinds. Denials are requests the hotel refused because nothing was available. Regrets are guests who chose not to book, usually because of price. Only denials point to demand above capacity, and many booking systems can’t record the difference reliably, so some RMS vendors, IDeaS among them, prefer statistical unconstraining methods. Either way, a sold-out date in last year’s data needs a second look before it becomes next year’s baseline.

Common Forecasting Mistakes

These patterns show up often in forecasts that miss.

  • Relying too heavily on last year. Same-date comparisons break quietly when an event moves or a new hotel opens.
  • Ignoring a shifting booking window. If guests now book later or earlier than last year, comparing pace 60 days out pushes the forecast in one direction.
  • Missing local events. Event dates move, and nothing in a PMS flags it.
  • Treating group and individual bookings as one pool. A date that looks soft may already hold a group block.
  • Skipping the manual check. An RMS only works from the data it receives. A person who knows the local risks catches what the model can’t.

How to Measure Forecast Accuracy Over Time

Forecast accuracy is the gap between what was forecast for a date at a set point before arrival and what happened once the date was over. Comparing forecasts made at 90, 60, 30 and 7 days out with the final result shows whether the forecast improves as more data comes in.

A common way to express this is the mean absolute percentage error (MAPE). For each date, take the difference between forecast and actual rooms sold, ignore whether it was too high or too low, and divide by the actual figure. MAPE is the average of those percentages across all the dates you check. A lower MAPE means a more accurate forecast. Also track bias, whether the forecast runs high or low on average, because MAPE hides the direction of the error.

Split the error by cause. A forecast that misses on event dates but holds on ordinary weeks points to a calendar problem. One that misses mainly on dates that sell out early points back to unconstrained demand. A monthly review of these errors is a sensible habit.

To check whether better forecasts are paying off in profit as well as revenue, track results against gross operating profit per available room (GOPPAR) and revenue per available room (RevPAR). For the person who owns the forecast, see the hotel revenue manager role.

What Resaco Recommends

Resaco is a hospitality marketing agency and doesn’t build or sell forecasting software, so we have no platform to recommend. IDeaS, Duetto, Atomize and similar tools automate the pace and pickup calculations, and the right choice depends on the property. Forecasts often break before any tool is involved: an events calendar nobody owns, or a habit of treating sold rooms as the full demand picture.

Your own website also produces early demand signals: searches in the booking engine for specific dates, direct enquiries and visits to package or event pages. These often appear before the PMS shows pace, and they tell you what draws guests to your area. Bringing that data into forecasting is the part Resaco works on with hotels, and it gives an earlier warning that a date is about to run hot, which is exactly what the Kraków hotel lacked. Direct bookings also avoid the commission that Booking.com charges.

If you’d like to know what your website data could add to your forecast, book a free analysis.

Frequently Asked Questions

What’s the difference between demand forecasting and revenue management?

Forecasting is the input: an estimate of what will happen on a date if nothing changes. Revenue management is the set of decisions made in response, including pricing, segmentation and channel strategy. Good decisions on a wrong forecast still produce poor results.

How far ahead should a hotel start forecasting a specific date?

As soon as bookings start to arrive for it, refining the forecast weekly as the date gets closer. Flag event dates as soon as they are confirmed.

Can a small independent hotel forecast demand without dedicated software?

Yes, with more manual work. A spreadsheet that tracks pace against the same days last year, a maintained events calendar and a weekly look at a few competitors’ rates on Booking.com cover the basics.

Related Reading

Written by

Daniel Laurean

Growth Consultant EMEA. Daniel supports Resaco’s clients in international growth and EMEA expansion, aligning sales and marketing to create predictable, scalable results. As an ultra runner and podcast host, he brings discipline and long-term thinking to every build.

Enjoyed the read?

Let’s see how this applies to your business.

Book a free 45-minute session with a senior strategist. We’ll look at your specific situation and tell you honestly what we’d tackle first — no pitch deck required.