PriceCortex™ — How AI Prices Hotel Rooms

What Shipped Recently

The Engine Keeps Getting Smarter

Dynamic pricing is not a one-shot product. Every session we ship improvements — new signals, better matching, sharper explanations. Here's what's new in the engine.

SESSION 28

PMS History Loader

Brand-adapter pattern. Choice xlsx → pms_occupancy_history. Trailing-year occ/RevPAR vs LY per property.

SESSION 28

Event-Aligned LY Matching

52-event calendar (2024-2027). Movable holidays match same-event LY, not same-date LY. Memorial Day vs Memorial Day.

v2 ALGO

Occupancy Pressure Signal

Replacing static heat with dynamic pressure: CY occ trajectory vs LY-event-aligned + rolling 4-DOW. Drives Step 4 risk capacity dynamically.

HALO™

AI Revenue Manager Chat

Conversational interface for owners. Asks about RM concepts, drills into pricing, narrates pressure context. Live at /halo.

FT-TRANSFORMER

Neural Net Rollout

Architecture in active rollout. Rule-based ensemble carries cold-start while collecting 60-90 days of decisions+outcomes for training.

Algorithm Activity

The Engine Is Running Right Now

PriceCortex™ runs twice a day on a schedule. Every run is a fresh pricing decision per (property × stay-date × room-category), with full reasoning logged.

1

Pricing Runs to Date

runs logged in pricing_engine_runs

564

Recommendations Made

every (prop × date × category)

21

Properties in Latest Run

most recent pricing pass

2026-06-01 23:50 UTC

Last Run Timestamp

UTC

4,587

PMS History Rows

trailing-year occ/RevPAR

—

Event Calendar Entries

for LY-event-aligned matching

The Problem

Hotels Price Blind.
We Give Them Eyes.

Traditional revenue management compares your hotel's average rate to the market average. That's like comparing apples to fruit salad. We match room-to-room: your King Suite against their King Suite. At the competitor level, by priority rank, updated twice daily.

No other RMS in the US does room-category-level competitive intelligence for midscale and economy hotels.

PriceCortex™ Pricing Engine

Seven-Step Room-Category
Dynamic Pricing

Every room type gets its own competitively-derived price. Not a property-wide "base rate." Each room category priced independently against the same category at ranked competitors. Steps 1-6 are live in production. Step 7 (occupancy pressure) is in active rollout.

1

Room Category Matching Core Innovation

For each room type you sell (King, Queen, Suite, Standard), the AI finds the same room category across your competitors. Not a multiplier — actual like-for-like competitive rate intelligence. "What is Holiday Inn charging for THEIR King Room tonight?"

YOUR King Room → matched to → Comp A King ($139), Comp B King ($149), Comp C King ($159)

2

Priority Rank Waterfall

Competitors are ranked by market relevance. Rank 1 = closest true competitors who fight for the same guest. The AI prices off Rank 1 first. If Rank 1 is sold out, it waterfalls to Rank 2, then Rank 3. Always finds a market signal.

Rank 1: Carriage House King ($119) → SOLD OUT → fall to Rank 2: Holiday Inn King ($139)

3

Competitive Base Price Per Room Type

The median price of matching room categories across available ranked competitors = the AI-derived base price for that specific room type. This happens independently for every room category sold.

King base = median($139, $149) = $144 | Queen base = median($109, $119) = $114

4

Risk Capacity Owner-Driven

How aggressive does the owner want to go? Conservative = stay within ±5% of competitive base. Moderate = ±10%. Aggressive = ±20%+. Target occupancy numbers can also drive this: "I need 80% occupancy" → the AI calculates the required aggression level automatically.

Owner: MODERATE (±10%) on $144 base → AI range: $130 – $158

5

Market Heat Adjustment Secondary Signal

Cold market (lots of availability) → discount adjustment within risk bounds. Hot market (competitors selling out) → premium adjustment. Because competitive prices already reflect market heat to a large degree, this is a nuance signal rather than a primary price driver.

2/3 Rank 1 sold out → WARM → +3% premium within risk bounds

6

Floor / Ceiling Enforcement

Hard limits the AI never breaches. When a computed price hits either boundary, a breach event is logged with full reasoning and the owner is proactively notified: "Your ceiling was hit 12 times on weekends — consider adjusting."

Computed: $148 | Floor: $89 ✓ | Ceiling: $179 ✓ | No breach → PUBLISH $148

7

Occupancy Pressure Signal In Active Rollout · v2

The next-generation primary signal. PriceCortex now ingests trailing-year PMS history per property and computes a dynamic occupancy pressure score: current-year occupancy trajectory vs LY-event-aligned and rolling 4-DOW averages. Pressure replaces static risk capacity in Step 4 — pricing automatically tightens when occupancy climbs faster than expected, and loosens when it falls behind. Event calendar handles movable holidays (Memorial Day matches Memorial Day, not May-25-2025).

CY occ trajectory +5.2pts vs LY-event-aligned · 4-DOW avg +3.1pts → PRESSURE: HOT → risk capacity auto-shifts toward aggressive

→ Phase 1 (PMS history loader): SHIPPED · → Phase 2 (Forward OTB): in build · → Phase 3 (pressure thresholds): next

In Development

The Roadmap

What we're building next. Public so customers and partners can see exactly where the platform is going.

SESSION 29

Current OTB Loader

Forward-looking on-book occupancy from Raw Occupancy sheet → pms_forward_otb_snapshot. Feeds pressure signal.

SESSION 29

Pressure Bucket Thresholds

5-bucket classification (cold/cool/neutral/warm/hot). Thresholds calibrated against PMS history backtests.

SESSION 29+

Property-Level Pressure Panel

RM view + owner portal surface pressure badge per (property × stay_date) with match-strategy confidence.

FUTURE

Category Pressure (screenshot OCR)

Per-room-type occupancy pressure via team screenshot uploads → vision parsing → category_pressure_signals.

Live Pricing Example

Watch PriceCortex Price
a King Room on Friday

Clarion Inn Strasburg — King Room — Friday night. Real data from the pipeline.

pricecortex_engine v2.2 — US_17579_001 — KING ROOM — FRI

// STEP 1: Room Category Match — KING ROOM
RANK 1 Carriage House → King Room = $119
RANK 1 Strasburg Village Inn → King Room = SOLD OUT
RANK 2 Holiday Inn Express → King Room = $159

// STEP 2-3: Competitive Base Price (Rank 1 available Kings)
Available Rank 1 King rates: [$119]
AI base price (Rank 1 median): $119

// STEP 4: Risk Capacity (owner: MODERATE ±10%)
Allowed range: $107 – $131

// STEP 5: Market Heat (1/2 Rank 1 sold out)
Heat: WARM → +3% premium

// STEP 6: Floor / Ceiling Check
Floor: $89 ✓ Ceiling: $179 ✓ No breach

// STEP 7: Occupancy Pressure (v2 rollout)
CY occ trajectory: +4pts vs LY-event-aligned (Memorial Day weekend match)
Pressure read: WARMING → risk capacity nudged toward aggressive

▶ PRICECORTEX RECOMMENDED PRICE: $125 KING ROOM — FRIDAY

// Halo™ owner narration:
"Your King Room Friday rate is $125. Strasburg Village Inn
sold out their King — market is warming. Carriage House
King is at $119. You're positioned above at moderate risk.
Memorial Day weekend last year ran 12pts hotter than this
weekday usually does — pressure signal agrees. Ceiling
headroom: $54."

AI Reasoning Layer

AI That Explains, Communicates
& Learns From Every Decision

Powered by Claude (Anthropic). Every PriceCortex decision comes with a plain-English explanation. Halo™ — the AI Revenue Manager — turns those explanations into conversations owners can ask back to.

◉

Halo™ — AI Revenue Manager

Conversational AI sub-brand of PriceCortex. Owners ask: "Why is my Friday King at $125?" or "Should I drop my floor next weekend?" Halo answers with grounded data from your property — pressure signal, comp rates, history.

🧠

AI Decision Explainer

Every price recommendation stored with full reasoning chain: which competitors matched, which room categories compared, heat signal, pressure read, risk capacity. Full transparency on every decision.

💬

AI Owner Communicator

Automated WhatsApp & email briefs per owner, generated by AI. "Your King Room is $125 because your top competitor's King sold out." Natural language. Builds trust.

📥

AI Constraint Parser

Owner replies "Don't go above $130 this weekend" → AI extracts the constraint and applies it as a temporary ceiling automatically. Natural language becomes pricing rules.

📊

AI Outcome Tracker

Every decision logged with actual results. Did $125 achieve expected occupancy? Did the override beat the AI? Continuous learning loop that improves accuracy daily.

FT-Transformer Neural Network

AI That Predicts Revenue
Impact Before You Move Price

The rule ensemble says "market says $125." The FT-Transformer predicts "at $125 you'll get 82% occupancy and $102.50 RevPAR — but at $135 you'd get 74% and $99.90 RevPAR. $125 maximizes your target." Rule ensemble carries cold-start; neural net takes weight as data accrues.

FT-Transformer Architecture — Feature-Tokenizer TransformerIN ACTIVE ROLLOUT

Input Features

comp_price + heat + pressure + dow + occ + season + rank

Feature Tokenizer

64-dim embedding per feature

Self-Attention

discovers which features matter per property

Property Embedding

64-dim property DNA — unique personality

Prediction Head

occ% + ADR + RevPAR per price point

SHAP Explainer

feature attribution → plain English

🔮

Revenue Impact Prediction

Before a price change goes live, the AI predicts: what will this do to occupancy, ADR, and RevPAR? Each price point mapped to expected returns. Maximize the owner's chosen target metric.

📈

Price Elasticity Learning

Some properties push $10 above market with minimal occupancy loss. Others lose 15% bookings on a $10 increase. The FT-Transformer learns each property's elasticity curve from historical outcomes.

🏨

Property DNA Embedding

Each property gets a learned 64-dimensional embedding — its unique DNA. Clarion Inn's response to competitor sell-outs differs from Comfort Inn's. The model captures this automatically.

⚡

Ensemble Intelligence

New properties: 100% rule engine. After 60-90 days of decisions+outcomes: NN reaches majority weight. Override rate below 10% for 30 consecutive days = eligible for auto-push. System intelligence grows daily.

Human-in-the-Loop

Revenue Managers
Stay in Command

AI recommends. Humans decide. Every override trains the model to be smarter.

✅

Accept

AI price goes live. As the model learns your market, acceptance rate climbs toward 90%+. Less work for RMs, better outcomes for owners.

✏️

Override

Manager sets a different price with a reason: local event, owner directive, rate too aggressive. This human feedback is the training signal that makes the AI smarter.

📊

Outcome Tracking

30 days later: did the AI beat the override? Override rate dropping = model is learning. Auto-push eligibility at <10% override rate for 30 consecutive days.

Built for Scale

Infrastructure That Grows
With Your Portfolio

500+

Property Capacity

2×

Daily Pricing Runs

10M+

Training Rows / Year

<$2

Per Property / Month

0

Device Dependencies

24/7

Cloud Execution

Competitive Intelligence That Prices Every Room Right