Case study

The Plaza Hotel Edirne — our own hotel, priced by Compass since 2024.

An independent boutique hotel in Edirne, Türkiye: expatriate and Balkan transit demand, domestic leisure, and strong event-driven peaks. It is the property Compass was built on — not a pilot we recruited, the hotel we run. We have run structured rate testing and market intelligence on the property since 2022; the engine has priced it live since 2024. And since June 2026 the system prices it fully autonomously: a 480-day rate horizon computed and pushed to the PMS every day, inside operator-set guardrails, with every change in an append-only audit log.

of expert manual pricing on 93% of days (full-year shadow run)

Within €2

of engine recommendations accepted by the operator

~98.9%

weekly manual pricing effort (owner estimate)

~10 h → ~1 h

Four years of rate discipline

ADR has risen every single year the engine has priced the property — €62 → €78 → €90 → €100 → €102, +13.1% across the two most recent years and +65% over four. Revenue-weighted, from the hotel's own PMS records.

The honest read on 2025: occupancy and RevPAR dipped because the hotel expanded its sellable inventory by roughly 12% — more rooms mechanically lower the percentages even when demand holds. ADR still rose to a record €102 across the larger inventory: the engine absorbed the new capacity without discounting it away. More rooms, higher rate, revenue defended.

The proof: it matched a human expert for a year

Before Compass touched a live rate, it ran in shadow for a full year of dates — generating recommendations in parallel with the property's expert manual pricing, fed live occupancy, booking pace and competitor data. It landed within €2 on 93% of days (mean delta €1.21), and on the largest-gap dates it was the more correct price: it caught live festival compression and a 93%-occupancy day the manual baseline had missed.

Eight days after that verdict, we switched the hotel's previous pricing process off and made Compass canonical — the exact shadow-then-cutover path we offer every prospect, rehearsed on ourselves first.

What we won't claim

• No "+X% revenue" promise. A defensible uplift number would need archived demand feeds for the counterfactual year, and they don't exist — so we prove parity with expert pricing and show the live demand feeds instead.
• The Plaza has been closed for renovation since 9 June 2026. The system keeps pricing the forward book autonomously through the closure, and the dashboard honestly shows the empty book as an empty book — we don't quote 2026 trading data anywhere.
• The 2025 revenue story is "defended through a capacity expansion", not a hockey-stick. We think honest numbers are the point of a case study.

Sources: the property's own PMS records (revenue-weighted, pulled 2026-07-16), the 2026-06-06 formal shadow-parity verdict, and owner-confirmed operational estimates. Methodology available on request.