There is a specific failure mode in hospitality performance media that is almost universal, rarely detected, and expensive in a way that compounds. It is not a failure of media buying. It is a failure of measurement that makes competent media buying impossible.

The mechanism is simple. Advertising platforms report conversions based on a tag that fires on the booking confirmation page. That tag fires when a booking is made. It does not fire again when the booking is cancelled, when the guest fails to arrive, or when the same booking is claimed by a second platform that also touched the journey. Optimisation algorithms then allocate budget toward whatever produces the most of these events.

The size of the gap

In a Mumbai boutique engagement we reconciled eighteen months of platform-reported conversions against property management system arrival records. Reported bookings exceeded actual arrivals by 41%. The composition of that gap is instructive.

Component Bookings % of reported
Platform-reported conversions (Google + Meta) 4,182 100.0%
Less: cross-platform duplication -611 -14.6%
Less: cancellations -438 -10.5%
Less: no-shows -141 -3.4%
Less: test and internal bookings -38 -0.9%
Reconciled PMS arrivals from paid media 2,954 70.6%
Overstatement 1,228 41.6%

Reconciliation of platform-reported conversions to PMS arrivals, 18-month period, 60-key boutique property

Cross-platform duplication was the largest single component. Google and Meta each claimed the same booking under their respective attribution models, and no deduplication had ever been implemented. Neither platform was behaving incorrectly — each was reporting what it observed, under an attribution window it disclosed. The error was in summing two overlapping figures and treating the result as a count of bookings.

Why the error is self-reinforcing

An inflated conversion count does not distribute evenly across campaigns. Campaigns operating in the retargeting and brand-defence layers sit closest to the conversion event and therefore accumulate the most duplicated credit. Prospecting campaigns, which create demand that converts later through another touchpoint, accumulate the least.

The optimisation algorithm reads this as evidence that retargeting outperforms prospecting, and shifts budget accordingly. Prospecting is starved, the retargetable audience shrinks, retargeting performance declines, and the account enters a contraction spiral that presents to the operator as market softness. Nothing about the market has changed.

When we re-ranked campaigns on PMS-reconciled arrivals rather than platform-reported conversions, the identity of the top-performing campaign changed in four of the five accounts we examined. In two of those, the campaign that had been receiving the largest budget share ranked fourth on reconciled data.

Building a signal that reconciles

The fix is neither exotic nor especially expensive, but it must be treated as an ongoing operating discipline rather than a technical implementation that completes.

  1. Deduplicate at the event layer. Implement server-side conversion tracking with a shared, stable event identifier — typically the booking reference — passed to every platform. Platforms will then recognise and discard duplicate claims on the same event.
  2. Deduct cancellations and no-shows. Send a negative or adjusted conversion event when a booking is cancelled, and reconcile no-shows on a monthly cycle from the PMS. A booking that did not arrive is not a conversion.
  3. Reconcile to arrivals, not confirmations. The authoritative record is the PMS arrival, because that is the event the property banks. Every other figure is an estimate of it.
  4. Report a variance figure every month. Track the gap between platform-reported and PMS-reconciled bookings as a standing metric. A widening variance is an early warning that tracking has broken.

What changes once the signal is trustworthy

In the Mumbai engagement, the six months following reconciliation saw media spend fall 11% in absolute terms while PMS-confirmed bookings from paid channels rose 68%. No new channel was introduced and no new creative approach was adopted in that window. The entire delta came from allocating the same budget against an accurate signal instead of an inflated one.

That is an uncomfortable finding for anyone who has spent years optimising against the wrong number, and it is worth stating the implication directly: in an account with corrupted measurement, effort spent on media craft has a low ceiling. Bidding strategy, creative rotation and audience refinement all operate on top of the conversion signal. If the signal is wrong, sophistication in the layers above it produces confident movement in the wrong direction.

Diagnostic questions for your own account

  • Does the sum of platform-reported bookings exceed PMS arrivals for the same period? By how much?
  • Are cancellations deducted from reported conversions, or only counted at booking?
  • Is a shared event ID passed to every platform, or does each fire independently?
  • When did someone last compare the media report to the property's own arrival record, line by line?
Written by
Manav Grewal Partner, Data & Attribution

This analysis draws on engagements led by the author. Findings are anonymised at client request; underlying figures are taken from client property management, channel manager and advertising platform records.