How Propeter's AI Revenue Engine Works
Overview: What the Revenue Engine Does
Propeter's AI Revenue Engine is a fully autonomous pricing system that continuously monitors market conditions, forecasts demand, and publishes optimised rates to every connected channel — without requiring a revenue manager to log in each day. It combines six specialised AI agents operating in a coordinated pipeline, backed by a deterministic 13-stage Rate Engine that calculates the final price for every room type and booking horizon.
Summary: The Revenue Engine ingests data from your PMS, OTA market feeds, local event calendars, competitor rate shops, and weather signals. It runs a demand forecast, models price elasticity, and produces a RevPAR-optimal rate for each room type × date × lead time combination. Rates are recalculated every 4 hours and pushed to all channels automatically.
The core design principle is that every pricing decision must be explainable. Unlike black-box ML models, every rate Propeter publishes comes with a human-readable rationale — showing which signals triggered the adjustment and how large each factor's contribution was. This lets your revenue managers stay in control while the AI handles the computational heavy lifting.
Key outcomes delivered by the Revenue Engine
- Automated rate updates every 4 hours across all connected channels
- Demand-responsive pricing that reacts to booking pace, local events, and competitor moves
- Protected floors and ceilings per room type — you always stay within your guardrails
- Full audit log of every rate change with reasoning attached
- RevPAR uplift averaging 18–24% in the first 90 days post-onboarding
The 6-Agent AI Pipeline
Propeter's Revenue Engine is built as a multi-agent system. Each agent is responsible for a specific stage of the pricing decision process. Agents run in sequence within each 4-hour update cycle, and each agent's output feeds into the next. The pipeline is orchestrated by a central Strategy Agent that produces the final rate recommendation.
Pipeline order: Data Ingestion → Market Intelligence → Demand Forecast → Price Elasticity → RevPAR Optimisation → Strategy Agent (Rate Output)
Data Ingestion Agent
Collects and normalises data from your PMS (reservations, cancellations, room-nights remaining), OTA dashboards, and internal bookings. It validates data integrity, flags anomalies, and prepares a clean dataset for downstream agents. Runs on a 4-hour polling cycle with real-time webhook ingestion for same-day cancellations and new bookings.
Market Intelligence Agent
Monitors your competitive set — scraping published rates from Booking.com, Expedia, and direct booking engines for each compset hotel. Also ingests local event data (concerts, conferences, sports), school holidays, public holidays, and weather forecasts. Produces a Market Pressure Index (MPI) score for each future date.
Demand Forecast Agent
Uses a combination of time-series decomposition and gradient-boosted regression trained on 18 months of your property's historical data plus regional market benchmarks. Produces a probabilistic occupancy forecast (P10/P50/P90) for each room type and date, 365 days forward. Booking pace deviation from the expected pace curve triggers immediate re-forecasting.
Price Elasticity Agent
Models the demand-price relationship for your specific property — how much occupancy changes for each £/€/$ increment in rate. This model is trained on your own booking data and updated monthly. It prevents the system from raising rates so aggressively that occupancy suffers, and identifies "inelastic" periods where demand is insensitive to price increases.
RevPAR Optimisation Agent
Takes the demand forecast and elasticity model and solves for the rate that maximises Revenue Per Available Room (RevPAR = Occupancy × ADR). Uses constrained optimisation with your guardrails (floor/ceiling rates) as hard constraints. Outputs a target rate vector across all room types and dates in the pricing horizon.
Strategy Agent
The final decision layer. Applies your hotel's rate strategy rules (rate plan hierarchy, OTA parity requirements, minimum stay restrictions, stop-sell thresholds) to the optimised rate vector. Generates the final publishable rates for each room type, rate plan, and channel. Writes the decision log with a plain-English explanation of each change.
The 13-Stage Rate Engine
Once the AI agents produce a target rate, the 13-Stage Rate Engine applies a series of deterministic calculations and business rules to arrive at the final rate pushed to each channel. This stage is fully auditable — you can inspect each stage's input and output in the Rate Audit Log within the dashboard.
Base Rate Retrieval
Fetches the property-configured base rate for the room type and season band. The base rate serves as the anchor from which all multipliers and adjustments are applied. It is set manually by the revenue manager during initial configuration and reviewed quarterly.
Season Multiplier
Applies a pre-defined season multiplier (e.g., 1.0× shoulder, 1.35× peak summer, 0.80× off-peak winter) based on the arrival date's season band. Multipliers are set in the Rate Configuration screen and can be adjusted at any time, taking effect on the next update cycle.
AI Demand Score Adjustment
Applies the RevPAR Optimisation Agent's recommended rate delta as a percentage adjustment to the base × season rate. A demand score of +15% means the AI is recommending a 15% uplift relative to the base. This is the primary AI-driven variable in the pipeline.
Competitor Position Adjustment
Compares your current rate to the Market Intelligence Agent's compset average. If your rate falls below your configured positioning tier (e.g., "always within 5% of compset median"), a small upward or downward nudge is applied. If the compset average is unavailable, this stage is skipped.
Booking Pace Override
If actual booking pace for the date is more than 20% ahead of the expected pace curve, an additional uplift multiplier is applied. Conversely, if pace is lagging significantly, a discount nudge is applied to stimulate early bookings. This creates a dynamic early-bird vs last-minute pricing curve.
Lead Time Discount Curve
Applies the configured lead time discount curve — typically reducing rates for bookings made far in advance (90+ days) and increasing them as the arrival date approaches and available inventory shrinks. This curve is customisable per room type in the Rate Configuration panel.
Event Premium
If the Market Intelligence Agent detected a high-impact local event (score ≥ 70/100) on the arrival date, an event premium multiplier is applied. Premium tiers are: Low (1.05×), Medium (1.15×), High (1.30×), Exceptional (1.50×). You can override or suppress event premiums per event in the Events Manager.
Guardrail Enforcement
Clamps the calculated rate to the floor and ceiling guardrails set for the room type in the Rate Configuration screen. This is a hard constraint — the AI can never publish a rate below your floor or above your ceiling. If the optimal rate exceeds the ceiling, the ceiling rate is published instead.
Rate Plan Derivation
Derives all child rate plans from the BAR (Best Available Rate): Non-refundable (typically −8% to −12%), Advance Purchase (−5%), Corporate (negotiated %), Package rates (BAR + inclusions value). Rate plan offsets are configured during setup and can be adjusted in the Rate Plans screen.
OTA Parity Check
Verifies that rates pushed to OTA channels comply with your configured parity rules. Rate parity mode options are: Strict (identical rates on all channels), Loose (OTAs may be up to 3% higher than direct), and Differential (custom per-channel rules). If a parity violation is detected, the system applies the corrective delta before publishing.
Minimum Stay Restriction Logic
Evaluates minimum length-of-stay (MinLOS) restrictions for the date. If occupancy on a surrounding high-demand night would benefit from a 2-night minimum requirement, the Stage 11 engine applies and flags the MinLOS restriction for the channel push. This is configurable per room type and date range.
Currency Conversion & Rounding
If your property operates in multiple currencies or serves channels in different markets, the rate is converted using the daily exchange rate and rounded to the nearest psychological price point (e.g., £149 rather than £147.83). Rounding rules are configurable per channel in the Channel Settings screen.
Tax & Fee Application
Applies applicable taxes (VAT, city tax, occupancy tax) and channel fees to produce the gross rate shown to the guest on each channel. Tax profiles are configured per property and updated automatically when tax rule changes are detected in supported jurisdictions. The net rate (what you receive) is logged separately for reporting.
How Rates Update
By default, Propeter recalculates and publishes rates on a 4-hour cycle: 02:00, 06:00, 10:00, 14:00, 18:00, and 22:00 UTC. This cadence means your rates are always fresh and reflective of the most recent market data without creating instability from too-frequent changes.
What triggers an out-of-cycle update?
In addition to the scheduled cycle, the following events trigger an immediate rate recalculation and push:
- Sudden occupancy change: 10+ room-nights cancelled or booked within a single hour on a future date
- Competitor rate drop: A compset hotel drops its rate by more than 15% on a tracked date
- High-impact event detected: A new event is added to the Events Manager with a score ≥ 80
- Manual override: A revenue manager publishes a manual rate or changes a guardrail setting
- Stop-sell threshold breached: Available inventory falls below the configured stop-sell level
Guardrails and human override
Guardrails are the most important safety mechanism in the system. Every room type has a configured floor rate (the minimum you will ever charge, protecting against distressed pricing) and a ceiling rate (the maximum, protecting against brand damage from perceived price gouging). The AI cannot override these values.
Best practice: Set your floor rate at your break-even point plus a small buffer (typically cost per occupied room + 20%). Set your ceiling at 3× your BAR during peak season. Guardrails that are too tight will prevent the AI from capturing revenue opportunity during high-demand periods.
Revenue managers can also publish a manual override for any date and room type. Manual overrides are respected for 48 hours, after which the AI resumes automatic management unless the override is renewed. All overrides are logged with the reason provided by the manager.
Frequently Asked Questions
Common questions from customers who have just onboarded or are evaluating Propeter's Revenue Engine.
How long does it take for the AI to learn my property?
Can I prevent the AI from changing rates during a specific period?
What happens if my PMS goes offline during an update cycle?
Does the AI take into account my direct booking website rates?
How does Propeter handle group bookings and contracted rates?
What data does Propeter use to build the competitor compset?
Was this article helpful?
Enhance customer engagement with our intelligent chatbot solutions. Seamlessly automate conversations and elevate user experiences with cutting-edge AI technology.
Products
Copyright © 2026 propeter | Powered by Propeter