Adgent

The booking converts. The stay is what pays.

Your conversion fires at booking; your revenue exists at arrival. In between sits a cancellation rate that differs by channel.

Your AI employee keeps the operating context: Adgent carries the revenue-management constraints you confirm, such as compression nights and channel cancellation patterns, into later reads. A direct PMS connection is planned, not live.

Pricing See the 79-day gap

The platform reports

a booking, on the day of the clickDay 0
  1. The booking, as a conversioncounted
  2. Booking window (not in the ad account)≈ 40 days
  3. Cancellation lands (not in the ad account)39 days before arrival
  4. Cancel rate, OTA · direct (not in the ad account)21.8% · 10.6%
Days from the click to the cancellation≈ 79

The cancellation lands outside every attribution window.

And since February 2025, no bidding model prices it either.

Day 0: the booking is counted as a conversionAbout day 79: the cancellation lands

One square per day, from the click to the cancellation.
≈ 79 daysfrom click to cancellation
21.8% vs 10.6%cancel rate, OTA against direct, on the same reported conversion
20 Feb 2025Google ended commission-per-stay bidding

Cloudbeds, March 2026, 90 million bookings · Hospitality Net, April 2025.

Agents that read the account every day, not at month-end.

Read every account daily

Meta and Google, with GA4 and Search Console alongside, read-only.

Decisions

Read the creative

Which concept is still scaling, which is wearing out, and where viewers leave a video.

Creative analysis

Check the signal before building

On one hotel account it checked the pixel first: 4 purchases in 30 days, too few to learn from.

Campaign build

Creative DNA

Example Meta ad for Saltbay Resort, a fictional hotel: “This summer, no clocks.”
Hook
Aspirational escape
Angle
Time freedom
Offer
Early booking, 35% off
Persona
Couples planning summer
Concept: scaling

Where viewers leaveone hotel account

  1. Start62,362
  2. 25%17%
  3. 50%7%
  4. 100%2%

Hold curve: top video, one hotel group's account, 28 days to 26 Aug 2026.

Whether they actually arrived.

The gap here is a calendar, not a bug.

Cancellations, no-shows and rate changes live in your booking engine, which Adgent doesn't read. Bring them as a Drive file and the agent reads it in chat.

Running these accounts for clients? How Adgent works across a book of client accounts

The report keeps restating itself for months.

Three things a revenue manager already knows that the ad account doesn't.

A documented case, wrong in both directions.

One hotel group's Google Ads reported $139,500 in revenue at 6.98× ROAS over a 45-day window. Fourteen of those bookings, worth $21,000, were later cancelled and stayed in the report, while 43 real bookings were never tracked at all: 68.4% coverage. The figure was too high and too low at once, which is why it looked plausible.

Source: Stape, hotel group case study, June 2026.

Click date against arrival date.

Dashboards add spend up by the day of the click; you decide by the night of the stay. On a compression night the spend is waste, because those rooms were selling anyway. On a need period thirty to sixty days out it can be the best money in the account. Revenue management reads pickup against the same time last year in the pace report; ad spend has no equivalent.

Brand defence is the quiet leak.

Paying to win back guests who were already coming to you is the most profitable-looking spend in the account and the hardest to defend on incrementality. Branded search and retargeting often report the highest return because they intercept demand you already owned.

The questions revenue managers ask first.

Our booking window is weeks long. Does a daily read help at all?
Yes, because it isn't trying to close the booking window. The outcome arrives about 79 days after the click, a 40-day booking window plus a cancellation landing 39 days before arrival, so no attribution window will hold it. What changes is that a drift is caught in week one instead of in a month-end figure that keeps restating itself. Decisions
Plenty of our bookings land by phone or through an OTA. What then?
Then the ad account is scoring the wrong event, and Adgent says so rather than modelling around it. Stayed revenue and cancellations by channel live in your booking engine or property system, which Adgent doesn't read. Bring them as a Drive file and the agent reads it in chat. Ground truth
Can it tell a compression night from a need period?
Not from ad data alone: occupancy and pickup live in revenue management, which Adgent doesn't read. You can tell it. Facts you confirm in chat are kept as account memory and loaded into later sessions, so “these dates sell themselves, stop buying them” is something it carries rather than something you repeat. You can override it at any time. Account memory
The platform's revenue and my analytics disagree. Which do I believe?
Neither one is stayed revenue. Adgent reads GA4 and Search Console beside the ad platforms as context, read-only; it doesn't reconcile them or pick a winner. Before a number can move money, the Trust Gate checks the window behind it, and a window it can't trust raises a finding, never an action. Trust Gate

See which channels book, and which ones cancel.

Adgent keeps booking and cancellation context beside the ad account, so the next report ranks channels by what stayed booked.

Pricing