GuestBrain
Back to blogRevenue

RevPAR, ADR and occupancy: how to forecast hotel KPIs with AI

Forecasting models applied to the three key hotel KPIs.

Álvaro García Published: 2026-03-05 13 min

What each KPI means

  • Occupancy: rooms sold divided by rooms available.
  • ADR (Average Daily Rate): room revenue divided by rooms sold.
  • RevPAR (Revenue per Available Room): occupancy × ADR. The king KPI.

Why predict all three together

If you forecast 90% occupancy but ADR drops 20%, RevPAR falls. The three are correlated (raising rates usually lowers occupancy), and serious models predict them as a system rather than one by one.

Models that work in 2026

  • Gradient boosting (XGBoost, LightGBM): the workhorse.
  • Time-series models (Prophet, NeuralProphet): capture seasonality.
  • Hybrid models that incorporate events: among the most advanced approaches.
  • LLMs as an explanation layer, not for pure prediction.

Variables feeding the forecast

  1. Booking and cancellation history.
  2. Local events calendar (sports, conferences, concerts).
  3. Historical and forecast weather.
  4. Google search trends.
  5. Current pickup vs the same date last year.
  6. Compset (estimated competitor prices and occupancy).
  7. FX rates for foreign source markets.

How to measure accuracy

  • MAPE (Mean Absolute Percentage Error) by horizon: 7, 30 and 90 days.
  • A realistic error target for each KPI and horizon, set from your own history.
  • A comparison against your current human forecast (the model has to beat it to earn its keep).
  • Stable error across seasons.

How to integrate into operations

A forecast is useless if nobody sees it. Ideally the system delivers it each morning in plain language: "This week will close 4 points below forecast on weak midweek demand. Worth driving demand for Saturday and Sunday."

From forecast to action: marketing's role

A forecast that flags a soft week only adds value if someone acts on it. Price and restriction levers belong to the revenue manager and their RMS. Demand levers belong to marketing: a segmented email to past guests, a Google Ads or Meta Ads campaign, an offer for the direct channel.

GuestBrain works on that second part and does not forecast or change rates. It unifies your guests in a CRM with data from your PMS and booking engine, sends segmented newsletters, optimizes your campaigns toward real cost per booking and measures which bookings each channel drives. Learn more on our AI hotel marketing software page.

Is your forecast flagging soft dates? Request a demo and see how GuestBrain helps drive direct demand with CRM, email and campaigns.

Frequently asked questions

As a rule of thumb, at least 12 months to capture one seasonal cycle. With 24-36 months, trends are clearer and unusual years can be isolated.

Yes, with more uncertainty: market or compset data is typically used as a baseline while the hotel's own history builds up.

No model predicts black swans. What a good system can do is spot the anomaly early (a pickup drop, a spike in cancellations) so the team can re-forecast as soon as possible.

No. GuestBrain doesn't forecast KPIs or recommend rates; that's the job of your RMS or revenue team. GuestBrain is the AI marketing layer that works alongside them: a CRM with one profile per guest, segmented newsletters, Google Ads and Meta Ads campaigns optimized toward real cost per booking, booking attribution, and anomaly and parity alerts.

Key concepts that show up in this article

  • ADR

    Average Daily Rate. The average rental rate of an occupied room. Calculated by dividing room revenue by the number of rooms sold.

    Read more

  • RevPAR

    Revenue per Available Room. Revenue per available room. The key hotel performance indicator, calculated as ADR multiplied by occupancy rate.

    Read more

  • Revenue management

    The discipline of optimizing the price-inventory-channel mix to maximize hotel revenue. Combines historical analysis, demand forecasting and pricing rules.

    Read more

  • Pickup

    Net change in rooms booked for a future date, measured between two time cuts. Key indicator to evaluate booking pace.

    Read more

  • Forecast

    Demand and occupancy prediction for a future horizon. AI improves the accuracy of classic forecasts by adding external signals like events, weather and search trends.

    Read more

  • BAR

    Best Available Rate. The best available rate published by the hotel for a specific date. The reference rate on which the rest of the rate plan is built.

    Read more

  • Compset

    Competitive set. Group of hotels defined as direct competitors against which occupancy, ADR and RevPAR are compared.

    Read more

More articles to go deeper into AI and hospitality