Demand Forecasting
In hotel revenue management, demand forecasting is the practice of predicting how many rooms will sell on a future date, and at what rate, so that pricing and staffing decisions can be made before the date arrives rather than explained after it.
At an independent property it is usually simpler arithmetic than the word suggests, and the hard part is not the model but the habit of writing the number down.
See how ampliphi reads demand for independent hotels
Get a free revenue audit · Book a demo
What is demand forecasting?
A statement of expected outcome for a specific future date, made early enough to act on.
It differs from the metrics feeding it in one important way. On the books is a count of what exists now. Booking pace is that count against a comparison period. A forecast goes further and commits to a number: this date will finish at 35 rooms.
That commitment is what makes it useful and what makes it uncomfortable. A forecast can be wrong in a way a count cannot, and the discomfort is why so many small properties stop at pace and never write down an expected outcome.
The basic arithmetic has three parts. What is already booked, plus what you expect to arrive between now and the date, minus what you expect to lose to cancellations and no-shows. Everything sophisticated in forecasting is a better estimate of the second and third terms.
Forecasts also come at different horizons and serve different purposes. A 90 day forecast informs staffing and marketing. A 30 day forecast informs pricing. A 7 day forecast informs housekeeping rotas and food ordering. The same date gets forecast repeatedly, and each version should be better than the last.
Resources: Booking pace · Pickup · Booking window
How demand forecasting works in practice
You build it per date, from your own history, and you keep the ones you made so you can check them.
Worked example. A 46-room inn forecasts a Saturday 35 days out. It has 24 rooms on the books. Its own history says a comparable Saturday picks up about 16 more rooms in the final 35 days. That gives a gross expectation of 40. Its cancellation and no-show history at this lead time runs near 12 percent, so it subtracts about 5. The forecast is 35 rooms, or 76 percent occupancy, and the date is neither a sellout to be priced up hard nor a problem to be discounted.
The three inputs in that example are all things a property can compute from a year of its own data. None of them requires a model, and all of them require records that most independents do not keep.
That is the real obstacle. Forecasting needs a booking curve, which means knowing how a typical date fills over time, and a booking curve is built by recording on the books by date every week for a year. Properties that skip the recording step find themselves forecasting from memory, which reliably anchors to the most recent similar date rather than to the same date last year.
The second discipline is keeping your forecasts. A forecast you did not write down teaches you nothing, because you will remember having predicted whatever happened. Written forecasts compared against outcomes are how you find out that you are systematically optimistic about midweek and pessimistic about holiday weekends, which is worth more than any refinement of the method.
Resources: Need date · Dynamic pricing
Why demand forecasting matters for independent hotels
Because without it, every pricing decision is made in reaction to something that has already happened.
The pattern at a property with no forecast is familiar. A date is priced once, months ahead, from last year's rate and a feeling. Nobody looks at it again until it is close, and by then the only lever left is a discount. Dates that were going to sell out get sold at last year's price and dates that were going to struggle get discounted in the final week, when the remaining guests are the least price-sensitive of the year.
A forecast breaks that pattern by producing an expectation that can be beaten or missed. That is the whole mechanism. Once you have written down that a Saturday should finish at 35, a Saturday sitting at 40 with three weeks to go is visibly a pricing opportunity rather than a pleasant surprise.
It also matters outside pricing. Staffing, food ordering, and deciding whether to take a group at a reduced rate all depend on knowing what the rest of the date is likely to do, and an owner-operator makes those calls constantly.
Resources: Revenue management for independent hotels
How to forecast demand at your property
- Record on the books by date, weekly, on the same day. Without this you have no booking curve and no forecast worth making.
- Build the curve by date type. Weekend and midweek dates fill differently, so a single average curve fits neither.
- Work out your attrition by lead time. Bookings made three months out cancel far more often than bookings made three days out.
- Forecast in three parts. On the books, plus expected remaining pickup, minus expected attrition.
- Write the forecast down with the date you made it. Comparing later is the only way the method improves.
- Adjust for what you actually know. A known event, a competitor closing or a road closure beats any historical curve.
What demand forecasting will not tell you
It cannot see anything that has not happened before. A new event, a competitor opening, a bridge closure or a weather pattern outside your history is invisible to a model built on history, which is why local knowledge stays part of the process at independent properties.
It says nothing about rate on its own. A forecast of 35 rooms is compatible with a very good month and a very bad one, and a forecast that does not carry an expected ADR alongside it is half the picture.
And it degrades sharply at small room counts. At 46 rooms, a single group booking of eight moves the forecast by 17 percentage points, so the honest output is a range rather than a number. Properties that report a forecast to one decimal place at this size are reporting false precision.
How ampliphi approaches demand
This is worth stating plainly, because the category invites the wrong assumption. Ampliphi does not publish an occupancy forecast. It produces a rate suggestion, and the everyday suggestion is demand-based, built on booking pace and occupancy.
The difference matters. A forecast states an expected outcome for a date. The rate suggestion answers a narrower question, which is what to charge given how this date is filling against how it has filled before. Most of the value an independent gets from forecasting is in that pricing decision, and the suggestion addresses it without requiring anyone to maintain a curve by hand.
What it removes is the recording step that stops most small properties forecasting at all, since the pace history accumulates from your PMS rather than from a spreadsheet somebody has to keep. The suggestion covers your base rate and the differential between room types. Competitive insight is a separate view and event data is a separate module again. You approve every rate before it publishes, and ampliphi runs on top of the PMS you already use.
Key takeaways: demand forecasting
- A prediction of how a future date will finish, made early enough to change a decision.
- The arithmetic is three parts: on the books, plus expected pickup, minus expected attrition.
- It needs a booking curve, and a booking curve needs on-the-books recorded weekly by date for a year.
- Write forecasts down with the date made. Unrecorded forecasts teach you nothing.
- At small room counts the honest output is a range, because one group booking moves it by double digits.
- History cannot see a first-time event, which is why local knowledge stays in the process.
Frequently asked questions about demand forecasting
Can a small hotel forecast demand without software?
Yes, and plenty do it well with a spreadsheet and twenty minutes a week.
What you need is a weekly export of on-the-books by date, pulled on the same day, kept for a year. That gives you a booking curve. Add your cancellation rate by lead time, which comes from the same data, and you have the two estimates a forecast requires.
The limitation is not accuracy, it is upkeep. The manual version competes with everything else in the week and is the first thing dropped when something breaks. If you are not forecasting at all, a spreadsheet started this week is worth more than software bought next quarter.
What is the difference between a forecast and booking pace?
Pace is a comparison. A forecast is a commitment.
Booking pace tells you that a date is seven rooms ahead of where it stood last year at the same point. That is a fact about the present and it contains no prediction.
A forecast takes that and says what the date will finish at. It requires two additional estimates that pace does not: how much more will arrive, and how much of the current total will fall away. Pace is easier, more reliable and enough for most weekly pricing decisions. A forecast is what you need when the question is staffing, ordering or whether to accept a group.
How accurate should a hotel demand forecast be?
Accurate enough to change a decision correctly, which is a lower bar than it sounds.
Larger properties often work to within a few percentage points at short horizons. At 25 to 60 rooms that precision is not available, because a single group booking is several points of occupancy on its own. A forecast that gets the direction right and the rough level right is doing its job.
Measure your own error rather than chasing a benchmark. Compare each written forecast to the outcome, look for consistent bias rather than individual misses, and correct the bias. Systematic optimism about midweek dates is a fixable problem. Being three rooms out on a given Tuesday is not.
How far ahead should I forecast?
Three horizons for three purposes, anchored to your booking window.
Around 90 days for staffing and marketing decisions, where you need a rough shape rather than a precise number. Around 30 days for pricing, which for most independents is where a rate change still has time to work. Around 7 days for operational decisions such as housekeeping rotas and food ordering, where accuracy is highest and the decision is immediate.
A property with an 11 day median window should not expect a 90 day forecast to be meaningful for pricing. It is useful for planning and nothing more.
How do I account for cancellations in a forecast?
By calculating your own attrition rate by lead time and subtracting it, rather than treating on the books as final.
Work out from your history what share of reservations cancel or no-show between a given number of days out and arrival. You will almost certainly find it varies sharply, with bookings made months ahead falling away far more often than bookings made days ahead, and OTA bookings behaving differently from direct ones.
Then apply the right rate to the right part of the forecast. This is also the arithmetic underneath overbooking, which is simply the decision to sell the rooms you expect to get back.
Related terms
Booking pace
A future date's current total against the same date at the same point in a prior period. The main input to a forecast, and the one most independents can act on without going further.
On the books
The count of rooms already reserved for a future date. The starting term in the forecast arithmetic and the figure you must record weekly to build a curve.
Pickup
Rooms added over a recent window. Forecasting the remaining pickup for a date is the second term, and usually the hardest one to estimate.
Booking window
The gap between booking and arrival. It sets how much of a date's final total should already be visible, which is what makes a forecast at a given horizon credible or not.
Need date
A future date forecast to underperform. The forecast is what identifies it, and identifying it early is the only reason the forecast was worth making.