Pricing that scales with properties, not bookings.

Five families of signal: where the date sits in your season and week, live occupancy from your PMS, booking pace against the same lead time last year, your competitor set’s rate band, and the events that move demand in your market — both local events and the school holidays of the markets your guests travel from. Fifteen explicit steps, each one auditable, and every recommendation carries the breakdown showing which signals pushed the rate up and which pulled it down.

Yes to both, on a daily cycle rather than a live tick. Competitor rates are scraped on a repeating schedule through the day; occupancy and rooms on the books sync from your PMS on their own cadence; the engine re-prices the forward horizon every hour. Hotel demand does not move fast enough for anything more frequent to be a real advantage — the honest constraint is scrape frequency and cost, not our refresh loop.

Not as a headline percentage, and we will not invent one. Historical competitor and pace feeds were never archived, so any backtested uplift figure would be fiction dressed as evidence — we ran that backtest, found it measured a data-starved engine rather than the real one, and left the number out of our own deck. What we can show you: a shadow run against the expert pricing we replaced that landed within €2 on 93% of forward dates, our own property’s ADR rising every year the engine has priced it, and a rate-change log in your account recording every price we moved so you can run the comparison on your own numbers.

Three things. It already runs a real hotel in production rather than a demo tenant — the same code path you would be on. Every price is explainable in plain English, so you can defend it to an owner without a data science detour. And the pricing itself is deterministic, not a language model guessing: the AI explains the number, it does not produce it. In two years of production our operator overrode 7 recommendations out of more than 600, which is the number we would rather be judged on than a marketing claim.

Mews connects through a six-step self-service wizard — you paste an access token and the system discovers your services, rates and room types. ElektraWeb is native but onboarded white-glove, because their API requires vendor-side authorisation we arrange for you. Any other channel manager — SiteMinder, Cloudbeds, something bespoke — can receive rates through a signed outbound webhook. If your PMS is not on that list, Watch works with no connection at all, which is exactly why the tier exists.

By blending what is already on the books with the same date last year, weighted by how far out you are: far from arrival the forecast leans on last year’s final position, and as the booking window closes it shifts onto your actual pickup. The blend follows a roughly six-week booking lead, which matters for late-booking properties — a flat linear ramp systematically under-forecasts their peak. We price a forward horizon of well over a year; the forecast is most useful inside ninety days and we show it as a gap rather than a guess where the history behind it is too thin.

Your data lives in EU-hosted Postgres, encrypted in transit and at rest, with row-level security isolating every property at the database layer rather than only in application code. PMS credentials go into an encrypted secret store, never into a config table. We are a UK company operating under UK and EU GDPR, we will sign a DPA, and our sub-processors are listed publicly. We do not hold SOC 2 — we are not going to display a badge we have not earned, and anyone showing you one should be asked for the report.

You choose your comp set — the properties you actually lose business to, not a radius the software picked. There is a map picker during setup if you want to browse what is nearby, and you can add or change properties later. For each competitor we track published rates across your forward dates, and we detect when one has sold out, which is the demand signal a plain price list misses entirely.

Published rates are collected on a repeating daily schedule and stored as a history, so you see a trend rather than one afternoon’s snapshot. Every figure on screen carries the timestamp of the scrape behind it. Two deliberate details: we prefer the actual bookable rate over a struck-through headline, and we exclude your own listing from your market median — rate-shopping feeds will hand you your own OTA page, which quietly drags "the market" towards whatever you already charge.

On Pro, a Mews property completes the wizard and gets its first full set of recommendations generated the same day — nothing is pushed to your PMS until you switch pushing on, so there is no risky first night. ElektraWeb takes longer because the vendor authorisation is arranged for you. On Watch there is no PMS step at all: we build your comp set with you on the call. In every case the sequence is deliberately shadow-first — look at the numbers next to your own decisions for a couple of weeks before letting anything write.