DynamicAdBook a call

Direct bookings, six source markets, on a margin the OTA takes a quarter of.

A boutique design hotel in Psyrri, Athens, with an editorial press footprint and six site-language demand markets. Commerce OS deployed across all six layers against direct-revenue contribution net of OTA cannibalization, with weekly variant cadence across Meta, Google Search, and Hotel Ads.

Variant cadence

1/month → Continuous matrix

Layers engaged

6 of 6 · continuous

Reporting truth

Platform ROAS → Direct-revenue net of OTA

Inventory parity

Hotel Ads · verification in cycle

Buyer trigger

Slot · The Operator With OTA-Trapped Margin and Multi-Market InboundSignal · Active
The Operator With OTA-Trapped Margin and Multi-Market Inbound

Booking volume held, but the take-home margin did not. Booking.com and Expedia routed 15 to 25 percent of every reservation through commission. Six site-language markets ran on one blended campaign rather than geo-segmented prospecting, which masked which inbound markets actually compounded. The booking widget existed; the direct-vs-OTA contribution gap was not in any operator dashboard. The operation needed direct-booking acquisition, geo-segmented prospecting, and OTA-net reporting on the same ledger.

Signals firing the engagement

Platform-attributed ROAS told a different story than the post-stay contribution read, no reconciliation in flight
Six site-language markets ran as one blended campaign; high-yield markets subsidized low-yield ones invisibly
Hotel Ads inventory parity issues were degrading Google bidding without the operator seeing why

Which layers were engaged, and why.

Six layers continuous, lifecycle and acquisition operated against the same hypothesis ledger and the same source-market cohort definitions. The Cohort + LTV model net of OTA is the reporting truth; platform ROAS is a diagnostic input, not the headline number.

LAYER 06ENGAGED

Dynamic Analytics

Cohort & LTV Model for the direct-revenue contribution payback decision. Survey-validated direct revenue net of OTA replaces platform ROAS as the operator read.

LAYER 05ENGAGED

Dynamic Optimization

Algorithmic Media Operation + Creative Testing Loop + Lifecycle Model. Lifecycle is pre-arrival upsell and post-stay rebook, on the same ledger as paid acquisition.

LAYER 04ENGAGED

Dynamic Ad Engineering

Creative Production for variant velocity across the six site languages. Editorial photography and UGC briefs share one hypothesis ledger.

LAYER 03ENGAGED

Dynamic Strategy

Creative & Brand Model translates the intelligence into the per-market angle hypothesis ledger. Each angle ships against a written hypothesis, not a vibe.

LAYER 02ENGAGED

Dynamic Intelligence

Competitor & Market Model + Audience & ICP Model for the source-market read. Where the design-hotel inbound converges by language and lead time, where the editorial press footprint already earns demand.

LAYER 01ENGAGED

Dynamic Infra

Signal Audit + Conversion Surface for the booking-engine handoff, server-side parity, and Hotel Ads inventory feed. Server-side GA4 + Meta CAPI + post-stay survey ground-truth.

How the layers spoke to each other.

The source-market cohort is the unit of analysis. Pre-arrival and post-stay flow behavior, paid creative variants, Hotel Ads inventory parity, and the bidding signal all read against the same acquisition cohorts. A market that compounds rewrites the brief; a market that decays flips a kill criterion. The operator authorizes every material move, the audit trail records the call.

Peer flow · See the Medistat mechanism →

  1. FLOW · 01

    Source-market behavior briefs the next creative cycle

    Direct-booking conversion by language, lead time, room category, and repeat-booking rate
    Creative Testing Loop hypothesis ledger reads the source-market cohort definitions
    Ad Engineering briefs variants against markets that compound, not markets that booked-then-vanished

    Artifact · Hypothesis ledger, weekly refresh

  2. FLOW · 02

    Hotel Ads inventory parity unlocks the bidding signal

    Conversion Surface closes the booking-engine + Hotel Ads parity gap
    Google Hotel Ads and Search recover the bidding model view of the property
    Optimization reallocation prices source markets against direct-revenue contribution, validated monthly

    Artifact · Inventory parity audit + monthly direct-contribution snapshot

  3. FLOW · 03

    One ledger, two surfaces

    Editorial-grade Meta and Search variants from Creative Production
    Pre-arrival and post-stay flow briefs target the same hypothesis rows
    Both surfaces score on one Creative Testing Loop scoreboard, winning angles compound

    Artifact · Variant + lifecycle ledger

What changed in the operating state.

Process metrics · published

Variant cadence

1/month → Continuous matrix

Layers in coordination

6 of 6, on one ledger

Hotel Ads inventory parity

Closed · verification in cycle

Reporting truth

Platform ROAS → Direct-revenue net of OTA

Directional reads · qualitative

  • Pre-arrival and post-stay flows ship on the same hypothesis ledger as paid creative; segmentation is shared by source market, not parallel
  • Source-market cohort model replaced blended ROAS as the operator primary read
  • Hotel Ads inventory parity restored; Search and Hotel Ads bidding see consistent room-level signal
  • Survey-validated direct revenue net of OTA is the reporting denominator, platform ROAS is a diagnostic input

Quantified lifts · publish at cycle closeADR band, direct-revenue mix, RevPAR uplift, and cohort LTV figures publish on the next cycle close. Process facts above are on the record now.

What ships at close.

A weekly source-market cohort read and a continuous variant ledger, joined on market definitions. Every angle, every flow, every reallocation logged as a prediction-versus-actual row against the same baseline.

Artifact · Apollo Palm HotelContinuous

Cohort + Variant Ledger

  • Signal Audit + Conversion Surface at week 1
  • Strategic Read at week 4
  • Cohort + LTV model live at week 6
  • Standing variant + lifecycle ledger from week 7 forward
  • Monthly direct-contribution snapshot, quarterly compounding review

The standing rhythm.

  • Weekly variant cycle
  • Bi-weekly lifecycle flow review
  • Monthly source-market cohort refresh
  • Quarterly Strategic Read

The operation runs continuously across acquisition and lifecycle on the same ledger. The variant cycle runs weekly, lifecycle flows refresh every two weeks, the source-market cohort and direct-contribution snapshot publish monthly, the Strategic Read refreshes quarterly. The audit trail records every cross-layer move.

Frequently asked questions

How does Apollo Palm Hotel measure direct bookings against OTA cannibalization?
The Cohort + LTV model uses survey-validated direct revenue net of OTA commission (Booking.com and Expedia charge 15 to 25 percent per reservation) as the primary reporting truth. Platform ROAS is a diagnostic input, not the headline number. Source-market cohort definitions are shared across paid acquisition and lifecycle flows on one ledger.
What platforms does Commerce OS run on for Apollo Palm Hotel?
Commerce OS runs Meta, Google Search, and Google Hotel Ads in active coordination. Server-side GA4 and Meta CAPI provide the bidding signal. The Hotel Ads inventory parity gap was closed at week one of the engagement, restoring the bidding model's view of the property across Search and Hotel Ads.
What does the Cohort + Variant Ledger contain for Apollo Palm Hotel?
The Cohort + Variant Ledger carries: Signal Audit and Conversion Surface from week one; Strategic Read from week four; Cohort + LTV model live from week six; a standing variant and lifecycle ledger from week seven onward; monthly direct-contribution snapshots; and a quarterly compounding review. Every angle, flow, and reallocation is logged as a prediction-versus-actual row against the same source-market cohort baseline.

The other engagement on the record.

← All case studies
Lead OS · B2B healthcare · Active engagementMedistat

Biostatistics and clinical-data consultancy. Narrow vertical, considered purchase, low conversion volume. Lead OS deployed across five layers, the operating record reads below.

Read the engagement →

Run this operation on your account.

Engineered cadenceContinuous readOutcome at handoff

Run the same operation on your account
After the projectDynamic Commerce OS.See how Commerce OS runs end-to-end.
Apollo Palm Hotel · Commerce OS Case Study | Dynamic Ad | Dynamic Ad