Forecasting basics — what forecasting means

In lessons 16-18 we learned how to read demand moving in real time — pickup, pace, same-point. Now comes the next step: what can you expect in the future? How do we translate the current pickup curve into a concrete revenue projection for the next 30, 60, 90 days?

The forecast is one of an RM’s most important outputs — not so we can “predict the future” (that won’t happen), but so we can run the whole hotel on forward-looking knowledge. The morning front-desk headcount, this week’s spa-product order, the marketing campaign planned two months out — all of them rest on the forecast.

The goal of this lesson is to understand what a hotel forecast is, who it’s for, what types it has, and how we measure its accuracy.

What a hotel forecast is

A hotel forecast isn’t a single number — it’s a multi-dimensional projection that, for every future date (or period), contains at least:

• Occupancy — expected room-nights
• ADR — expected average room rate
• RevPAR — expected RevPAR (the product of the two)
• Segment breakdown — expected occupancy and ADR by segment
• Expected cancellation level — the expected cancellation rate of booked room-nights
• Expected walk-in / pickup — the new bookings still to arrive

The classic, beginner-level forecast states only occupancy and ADR. A mature forecast does all of this at segment level too — because in lesson 9 (Seasonality) we saw that the aggregate number hides a mix effect that’s invisible without the segment breakdown.

Hotel Peaqplus City’s November 25 (Saturday) forecast, as of today (7 days out):

Metric	Expected value	Uncertainty range
Occupancy	82%	78-87%
Average rate (ADR)	EUR 118	EUR 112-125
RevPAR	EUR 97	EUR 87-109
Cancellation (expected)	4 rooms	2-6
Walk-in (expected new bookings)	8 rooms	5-12

The uncertainty range is crucial here. A mature forecast is not a point but an interval. A single number will never be exact — but a 78-87% range helps the hotel manager plan.

Who the forecast is for — 5 stakeholders

The forecast is not just an RM-internal tool. Five different stakeholders use it, each for a different purpose:

1. HR / operations — staffing the front desk and housekeeping

If the hotel knows 25 days out that November 25 will be 80% (vs. an average 50% Saturday), then:

• The front-desk roster can be tuned (more people for the check-in peak).
• Housekeeping scheduling can be organized ahead (cleaning isn’t left to the last minute).
• F&B shifts can be planned (more waiters for Friday-Saturday evening).

A bad forecast = operational over- or under-staffing — both expensive for the hotel (guest experience vs. labour cost).

2. F&B / procurement — breakfast and dinner planning

The breakfast order (coffee, pastry, fruit) goes to suppliers 5-7 days ahead. If the forecast says November 25 will be 80% (~120 guests at breakfast) and procurement matches it, there’s no over-order and no shortfall. A 30% forecast error = 15-20% waste or shortage in the restaurant.

3. Marketing — campaign timing

Covered in depth in lesson 31 (Marketing and RM in concert). The marketing team decides off the forecast:

• Where is the forecast low? → launch a promo campaign here.
• Where is the forecast high? → don’t spend marketing budget, it’ll fill on its own.

The forecast is the basis of marketing-budget allocation.

4. Bank / financier — cash-flow planning

A bank or investor wants a monthly cash-flow projection — expected revenue, expected cost, expected net cash. Its basis is the monthly RevPAR forecast × room count × 30 days. If the hotel forecasts with a ±15% error, the bank prices an uncertainty premium into the loan rate. An accurate forecast = cheaper financing.

5. Owner / GM — strategic decisions

Owner- and GM-level decisions need a monthly and quarterly forecast:

• What will the November-December season bring?
• Does it meet the budget target?
• Do we need to launch a new sales campaign?

These are all forecast-based decisions.

Top-down vs. bottom-up forecast

Two fundamental forecast-building methods exist:

Top-down

Builds from the top: a monthly target set for the whole hotel is broken down to the daily, then segment, level.

For example: “We plan November for 80% monthly occupancy. That’s 30 days × 80 rooms × 0.8 = 1,920 room-nights. We distribute that to the daily level by the seasonality pattern.”

Pro: fast, easy to understand. Con: it ignores daily-level pattern differences — it would deliberately over-smooth an event-peak day, for instance.

Bottom-up

Builds from the bottom: we forecast at the daily level for every day, off the pace curve, same-point, segment mix and events. The monthly forecast is the sum of the daily forecasts.

For example: “November 1 — 62%, November 2 — 65%, November 3 — 58% (Sunday), November 4 (Monday) — 70%…”. If you forecast every day, the monthly total comes out automatically.

Pro: more accurate, because it accounts for the daily pattern differences. Con: time- and tool-intensive — building a daily forecast by hand for 30+ days takes hours.

The modern approach — combined

A mature RM organization uses both:

• Bottom-up is the detailed, daily forecast for the concrete decisions (rate revision, restrictions, marketing campaign).
• Top-down is the monthly sanity-check — it verifies that the bottom-up, summed up, stays strategically sensible.

The Peaqplus Forecast module generates a daily-level (bottom-up) forecast and also computes a monthly aggregate. The discrepancy between the two (if the bottom-up doesn’t make monthly sense) is a warning sign for the model.

Daily Forecast vs. Monthly Outlook

The forecast works on two main time horizons:

Daily Forecast

• Horizon: the next 0-30 days.
• Refresh: daily, every morning.
• Users: RM, sales team, F&B manager, HR planner.
• Use: rate revision, restrictions, group decisions, operational scheduling.
• Accuracy expectation: ±5% on an average day, ±10% on an exceptional (event-peak) day.

The Daily Forecast sits at the centre of the RM’s daily routine (lesson 12).

Monthly Outlook

• Horizon: the next 1-6 months.
• Refresh: weekly or biweekly.
• Users: GM, owner, finance, bank.
• Use: budget revision, cash-flow planning, strategic decisions.
• Accuracy expectation: ±10% on an average month, ±15% on a seasonal peak.

The two forecasts are built with different methodologies and stand at different uncertainty levels. The Monthly Outlook is the Daily Forecast aggregate + the further-out (1+ month) estimates.

Forecast accuracy measurement

A forecast is only worthwhile if we measure its accuracy. The classic metrics:

MAE (Mean Absolute Error)

MAE = Σ |Expected − Actual| / N

The mean absolute deviation. Hotel Peaqplus City’s November Daily Forecast vs. the actual result:

Date	Expected occupancy	Actual	Absolute deviation
Nov 1	62%	65%	3 pp
Nov 2	65%	62%	3 pp
Nov 3	58%	55%	3 pp
…	…	…	…
MAE total			4.2 pp average

The 4.2 pp MAE = the forecast deviates on average by 4.2 percentage points from actual occupancy. A good forecast moves in the 3-5 pp range at the daily level.

MAPE (Mean Absolute Percentage Error)

MAPE = Σ ( |Expected − Actual| / Actual ) × 100 / N

Shows the deviation in percentage form, which is comparable across hotels of different sizes. A good hotel forecast’s MAPE is 5-10% at the daily level.

Bias

The measure of systematic over- or under-estimation. If the forecast leans +2 pp on average, there’s a systematic over-estimate — something in the model always shows the forecast more optimistic than reality. A bias away from 0 = a calibration error to be corrected.

The Peaqplus Forecast module

The Peaqplus Forecast module generates a daily-level (Daily Forecast) and monthly (Monthly Outlook) forecast for the next 90 days. Its key capabilities:

Hybrid logic

Covered in detail in lesson 38 (Smart Forecast). The essence: it doesn’t use a single model but 3 layers:

• Layer 1: a statistical pace model (your own historical pace curve).
• Layer 2: pickup-trend extrapolation (the method from lessons 16-17).
• Layer 3: manual corrections (events, holiday shift, calendar adjustment).

The three layers combine, weighted, into a concrete daily forecast.

Segment-level forecast

Not just total, but by segment too. The transient business, transient leisure (OTA), transient leisure (direct), corporate and group segments each get their own forecast, and the aggregate total is the sum of the parts.

Confidence range

Every forecast number is also available as a range (expected value ± 1 standard deviation). In bank communication we use this range.

Continuous calibration

The model learns daily from the actual data. If on a given day the forecast deviates dramatically from actual, the model refines the weighting next time.

MAPE tracking and bias watching

The surface shows the module’s rolling MAPE (last 30 / 90 days / 1 year) and its bias level. The RM sees how far to trust the forecast — and if the MAPE rises, it signals the model needs calibrating.

In lessons 20 (Simple forecast methods) and 38 (Smart Forecast — hybrid forecasting) we go deeper into how such a model is built. Here we showed only the frame.

Key takeaways

• A hotel forecast isn’t a single number, but a multi-dimensional projection — occupancy, ADR, RevPAR, segment breakdown, cancellation, walk-in.
• A mature forecast gives an uncertainty range, not a point — a number hit exactly is just luck.
• 5 stakeholders use it: HR/operations, F&B/procurement, marketing, bank/financier, owner/GM — each for a different purpose.
• Top-down vs. bottom-up — two methods, worth using both. Peaqplus generates the bottom-up and gives a monthly aggregate.
• Daily Forecast (0-30 days) vs. Monthly Outlook (1-6 months) — different horizon, different accuracy expectation.
• MAE, MAPE, bias — the forecast-accuracy metrics. A good forecast’s MAPE is 5-10%, MAE 3-5 pp.