What is the recommended Online Travel Agency (OTA) sales and operations tech stack in 2027?

Direct Answer

An Online Travel Agency (OTA) in 2027 runs on a stack built around proprietary marketplace infrastructure, GDS connectivity, and ruthless conversion optimization at billions of search queries per day. The marquee apps are proprietary in-house marketplace platforms (Booking and Expedia both run their own at massive scale) wired to Sabre, Amadeus, and Travelport for GDS air and hotel inventory, SiteMinder and Cloudbeds for property-side channel management, Adyen for global payments, Forter for fraud, Salesforce for B2B partner and supply CRM, and Snowflake plus Databricks for the ML pricing and ranking layer.

Everything else — Iterable for lifecycle marketing, Zendesk plus Twilio Flex for support, SA360 and Skai for paid search — bolts onto that core.

Why the Online Travel Agency Stack Works Differently

An OTA is not a generic e-commerce or SaaS business, and four mechanics force a specialized stack rather than the Shopify-plus-Klaviyo kit a normal merchant would use.

1. Two-sided marketplace with perishable inventory. Hotel rooms and airline seats expire daily, so the system has to clear inventory in real time across hundreds of millions of search results. Pricing, availability, and rate-parity logic update thousands of times per property per day. Generic catalog software cannot model rate plans, length-of-stay restrictions, occupancy-based pricing, or last-room-availability rules; an OTA needs a purpose-built booking engine.

1. GDS and channel-manager connectivity is non-negotiable. Air inventory comes through Sabre, Amadeus, or Travelport under deep commercial agreements. Hotel inventory comes either through GDS aggregators or directly from property management systems via SiteMinder, Cloudbeds, Profitroom, or D-Edge. A new OTA without these pipes has nothing to sell. The integration work is years long and the contracts are commercial events, not procurement decisions.

1. Fraud, chargebacks, and payment complexity scale with GMV. Travel is the highest-fraud vertical in card-not-present commerce — stolen cards used at the last minute for non-refundable bookings the criminal will never use. OTAs run Adyen or Stripe for global acquiring across 130+ currencies, layered with Forter, Riskified, or Stripe Radar for real-time decisioning under 100 ms. A point lost on approval rate is millions of dollars; a point gained on fraud is the same.

1. Conversion is bought, measured, and optimized at industrial scale. Booking Holdings spent roughly $7 billion on performance marketing in 2025, and Expedia is in the same zip code. That spend runs through Google Ads, SA360, Skai, and proprietary bid-management ML reading from Snowflake and Databricks. The ratio of marketing spend to gross bookings is the most-watched number on the earnings call, and the stack is built to move it.

The Core Stack, Layer by Layer

This is the recommended set of products by functional layer for an OTA operating at meaningful scale. Layers that genuinely do not apply are skipped; nothing is padded.

Proprietary Marketplace & Booking Engine — In-house platform (Mirakl or Travelport Trip Services for sub-scale OTAs). Booking, Expedia, Airbnb, Trip.com, MakeMyTrip, and Hopper all run proprietary booking engines because no SaaS handles their query volume, ranking ML, and rate logic.

New entrants below true global scale can start on Travelport Trip Services, Sabre SynXis, or marketplace toolkits like Mirakl while they build, but every breakout OTA ends up rewriting the core. Budget: enterprise engineering org of 500-5,000 engineers, not a SaaS line item.

GDS Connectivity — Sabre, Amadeus, and Travelport (one or all three). Air content, car rentals, and a significant chunk of hotel inventory move through the big three GDSs. Connectivity is a multi-year commercial and technical project; pricing is per-segment or per-transaction with substantial volume commitments.

Expect six- and seven-figure annual minimums and revenue-share economics that get renegotiated every few years (the Amadeus-Sabre commercial fights are public).

Property & Rate Connectivity — SiteMinder, Cloudbeds, Profitroom, D-Edge. On the hotel side, OTAs connect to hundreds of thousands of independent properties through channel managers and PMS providers rather than the GDS. SiteMinder reaches 450+ distribution channels and 350+ PMSs; Cloudbeds runs both a PMS and a channel manager; Profitroom is strong in EMEA.

Connectivity is typically reciprocal — the OTA does not pay the channel manager; the hotel does, at roughly $50-$300/property/month.

Payments & Acquiring — Adyen (Stripe as the alternative). Adyen is the dominant OTA acquirer because of its single global platform, native multi-currency, and integrated risk module. Booking, Vrbo, and many large travel brands run Adyen. Stripe is the choice for newer entrants and for OTAs that already standardized on Stripe's ecosystem.

Pricing is interchange-plus with volume discounts; large OTAs negotiate well below 2%.

Fraud & Risk Decisioning — Forter (Riskified and Stripe Radar as alternatives). Forter protects top-five OTAs with real-time approve/decline decisions in under 100 ms and chargeback guarantees on approved orders. Riskified plays the same role at competing OTAs. Stripe Radar is the integrated choice when Stripe is the acquirer.

Pricing is a percentage of approved GMV, typically 0.3-0.8%, with the math justified by approval-rate lift.

Data Platform & Pricing ML — Snowflake plus Databricks (BigQuery as the alternative). The pricing, ranking, and personalization models that decide which hotel appears on row one for a given query run on a lakehouse. Snowflake handles the warehouse and analyst access; Databricks runs the model training and feature store.

Pricing is consumption-based; large OTAs spend tens of millions per year across both. Google BigQuery is the alternative at OTAs already on GCP.

CRM & Supply Partnerships — Salesforce Sales Cloud (HubSpot for B2B-only sub-scale). Hotel chain relationships, airline contracts, advertising-partner deals, and B2B affiliate programs are managed in Salesforce with custom objects for properties, rate agreements, and commission structures.

Pricing is roughly $165-$330/user/month (Enterprise/Unlimited). Most OTAs build heavy customization on top.

Lifecycle Marketing & Email — Iterable plus Braze. Cross-channel messaging — search abandonment, booking confirmation, pre-trip prep, in-trip support, post-trip review prompts — runs on Iterable or Braze, both of which scale to billions of messages per month. Pricing is per-MAU and lands in the high six to low seven figures annually at OTA scale.

Salesforce Marketing Cloud is the legacy choice; most modern OTAs replaced it.

Loyalty Programs — Proprietary (Genius, One Key, Airbnb Guest Favorites). Booking's Genius, Expedia Group's One Key, and Airbnb's Guest Favorites are all in-house. Off-the-shelf loyalty platforms cannot model OTA loyalty math (tiered discounts, rebate currency, partner reciprocity); every meaningful OTA builds its own.

Customer Support — Zendesk plus Intercom Fin plus Twilio Flex. Zendesk handles ticketing and case management; Intercom Fin runs AI-first deflection on the highest-volume questions; Twilio Flex powers contact-center voice and SMS at OTA scale. Pricing is per-agent and per-resolution; large OTAs spend mid eight figures across the three.

Ad Tech & Paid Search Automation — SA360, Skai, plus proprietary bid ML. OTAs are the largest single buyers of Google Hotel Ads and travel search inventory. Google SA360 and Skai are the campaign-management layers; the actual bid logic at Booking, Expedia, Trip.com, and Trivago is proprietary ML reading attribution from the data platform.

Skai pricing is enterprise; SA360 is bundled with Google Marketing Platform.

HR & Workforce — Workday. Workday is the default at OTAs above ~2,000 employees because it handles global payroll, equity, and the complex contractor relationships common in travel. Pricing is enterprise-only, typically $100-$300/employee/year all-in.

Layers deliberately skipped: there is no off-the-shelf "OTA-in-a-box" because the marketplace itself is the moat; there is no separate inventory system because the booking engine is the inventory system; there is no general-purpose CDP because the data platform plays that role.

Real Operators & What They Run

Public footprints, engineering blogs, and industry reporting point to the following stacks at named operators.

• Booking Holdings (Booking.com, Priceline, Agoda, KAYAK) — proprietary booking engine and ranking ML, Adyen as the primary acquirer, Forter for fraud, heavy Snowflake and proprietary data infrastructure, Iterable-class lifecycle marketing, Salesforce for supply partnerships, and an in-house bid-management layer above Google Ads.
• Expedia Group (Expedia, Hotels.com, Vrbo) — proprietary marketplace, One Key loyalty, Adyen and Stripe across brands, Databricks as a public reference customer for the data platform, and a heavy SA360/Skai paid-search posture.
• Airbnb — fully proprietary marketplace and trust/safety stack, Stripe as the primary acquirer historically, Braze and in-house messaging, Salesforce for hosts and B2B, and a famously deep Apache Airflow plus Presto/Trino data platform.
• Trip.com Group (Trip.com, Ctrip, Skyscanner, Qunar) — proprietary booking engine across multiple brands, deep GDS plus direct connectivity in APAC, large in-house ML and a Snowflake/Databricks-equivalent posture, Adyen and Alipay/WeChat Pay on the payments side.
• MakeMyTrip — proprietary engine for the Indian market, deep direct-connect with Indian hotel chains, Razorpay and Adyen for payments, Salesforce for B2B and corporate travel, and an in-house data platform.
• Despegar — Latin America-focused proprietary marketplace, Adyen plus regional acquirers, heavy local-payment-method coverage (Pix, Mercado Pago), and a Snowflake-class data platform.
• Trivago — metasearch rather than booking, so the stack centers on a proprietary bidding marketplace, Google-scale ad tech integrations, and a much smaller payments footprint than Booking or Expedia.
• Hopper — mobile-first, proprietary price-prediction ML, Stripe plus Adyen, Databricks as the data platform, and a heavy fintech overlay (Hopper Cloud, Price Freeze) on top of the booking layer.

Integration Architecture

The stack only works when supply (GDS plus channel managers), demand (the search and booking funnel), payments, fraud, and the ML data platform exchange data in milliseconds. The proprietary booking engine is the system of record for the marketplace; GDS and channel-manager feeds keep inventory fresh; Adyen and Forter resolve the transaction; Snowflake and Databricks score every search and every order; Iterable and Braze run the lifecycle that brings the user back.

An iPaaS layer (typically Workato at this scale, or proprietary internal frameworks) handles the connections that do not warrant custom code.

The most important integration is the inventory loop between GDS, channel managers, and the booking engine — a stale rate or oversold room destroys trust and triggers chargebacks. The second-most important is the payments-to-fraud loop, where every basis point of approval-rate lift compounds into nine-figure annual GMV.

The third is the ML feedback loop, where every search, click, and booking trains the next ranking model and the next paid-search bid.

Failure Modes

Four stack mistakes show up repeatedly when OTAs stall or lose share. (1) Buying a SaaS booking engine and never replacing it — sub-scale OTAs start on Travelport Trip Services or a marketplace toolkit and never invest in a proprietary engine, capping ranking quality and conversion forever; every breakout brand rewrites the core within five years.

(2) Underinvesting in fraud decisioning — treating fraud as a back-office function rather than a revenue lever costs 1-3 points of approval rate on legitimate travel transactions, which at OTA scale is hundreds of millions of dollars per year; Forter or Riskified pays for itself in the first quarter.

(3) Letting marketing data and pricing data live in different warehouses — when SA360 attribution and the pricing ML read from different tables, paid-search bids stop reflecting true unit economics and CAC drifts up unnoticed for quarters. (4) Outsourcing supply CRM to generic Salesforce without travel objects — hotel chain agreements, parity clauses, commission overrides, and rate-plan exceptions do not fit standard opportunity records, and the supply org ends up running on spreadsheets parallel to a half-built Salesforce.

Budget & Sizing

Monthly software cost scales with GMV and headcount. These ranges cover the recommended stack, not enterprise edge cases.

• Emerging OTA (under $50M GMV, 25-100 staff, regional focus). Hosted booking engine (Travelport Trip Services or Sabre SynXis), one GDS, SiteMinder for hotel connectivity, Stripe plus Stripe Radar, HubSpot for B2B CRM, Iterable entry tier, Zendesk, Snowflake starter, Workday or Rippling. Expect roughly $60,000-$200,000/month in software, dominated by GDS minimums and payments fees.

• Regional leader ($500M-$5B GMV, 500-3,000 staff). Hybrid proprietary plus licensed booking engine, two or three GDSs, SiteMinder plus Cloudbeds plus direct connects, Adyen as the primary acquirer, Forter for fraud, Salesforce Enterprise, Iterable plus Braze, Zendesk plus Twilio Flex, Snowflake plus Databricks, SA360 plus Skai, Workday. Expect roughly $1.5M-$8M/month in software plus payments and fraud fees scaling with GMV.

• Global OTA ($20B+ GMV, 10,000+ staff, multi-brand). Fully proprietary marketplace across multiple brands, all three GDSs, every major channel manager and direct-connect path, Adyen plus Alipay/WeChat Pay/regional acquirers, Forter plus in-house risk models, Salesforce Unlimited at scale, Iterable plus Braze plus proprietary CDP, Zendesk plus Twilio Flex plus in-house support automation, Snowflake plus Databricks plus in-house lakehouse, proprietary bid ML over SA360/Skai, Workday. Expect $40M+/month in software and infrastructure, plus billions per year in payments and performance marketing.

30/60/90 Day Implementation Plan

A staged rollout protects bookability, since the marketplace cannot go dark and supply contracts cannot be re-signed in a quarter.

Days 1-30: Lock the supply foundation. Sign or renew the GDS contract (Sabre, Amadeus, or Travelport) and stand up SiteMinder or Cloudbeds connectivity to the first tranche of hotel properties. Stand up Adyen for payments and Forter for fraud in sandbox; pass the first end-to-end test transaction.

Bring Salesforce online for supply partnerships and load the existing hotel-chain and airline contracts. Pick Snowflake as the data warehouse and begin landing booking, search, and clickstream events.

Days 31-60: Light up demand. Ship the first version of the booking funnel against the GDS and channel-manager feeds. Turn on Iterable for booking-confirmation and pre-trip messaging.

Stand up Zendesk for customer support and Intercom Fin for AI deflection on the top 20 question types. Connect SA360 and Skai for paid-search automation and wire the attribution pipeline back to Snowflake. Hire or contract the Databricks team that will own pricing and ranking ML.

Days 61-90: Optimize and instrument. Move Adyen and Forter to production with real fraud thresholds and measure approval-rate lift weekly. Ship the first ranking ML model from Databricks into the live search funnel and A/B test against the baseline.

Stand up the executive dashboard covering GMV, take rate, conversion rate, CAC, approval rate, and fraud rate. Add the second GDS or channel manager if supply gaps remain. Exit the quarter with a single operator view leadership trusts.

FAQ

Should we build our own booking engine or buy one? Buy at launch, build by the time you cross roughly $500M GMV. Every breakout OTA — Booking, Expedia, Airbnb, Trip.com, Hopper — eventually rewrites the marketplace core because ranking, pricing, and rate logic are the product and no SaaS keeps pace.

Adyen or Stripe for payments? Adyen wins at OTA scale for one-platform global acquiring and integrated risk; Booking, Vrbo, and many large brands run it. Stripe is the right pick for newer entrants and for OTAs already standardized on the Stripe ecosystem, especially in the US.

Do we really need both SiteMinder and Cloudbeds? Often yes. SiteMinder dominates channel management at independent and small-chain hotels worldwide; Cloudbeds owns a different slice with its PMS plus channel-manager combo. Large OTAs connect through every meaningful channel manager to maximize property coverage.

What fraud rate is realistic for an OTA in 2027? Best-in-class OTAs run net fraud well under 0.2% of GMV with Forter or Riskified on top of Adyen or Stripe Radar, while keeping approval rates above 95% on legitimate traffic. Anything worse is leaving money on the table.

Why Snowflake plus Databricks instead of just one? Snowflake is the warehouse for analyst access, BI, and SQL workloads; Databricks runs the model training, feature store, and Spark jobs the pricing and ranking team need. Trying to force one to do both job families is a known anti-pattern at OTA scale.

Is in-house bid management worth it over pure SA360 or Skai? At Booking and Expedia spend levels, yes — every basis point of bid efficiency is worth more than the engineering team. Below roughly $250M/year in paid-search spend, SA360 plus Skai is enough; above that, proprietary bid ML reading from the data platform is the standard.

Sources

• Skift — Amadeus Q1 2026 earnings and competitive dynamics with Sabre and Travelport (2026)
• SiteMinder — Channel manager coverage, OTA integrations, and PMS connectivity overview (2026)
• Cloudbeds — 2026 OTA guide and channel-manager product documentation (2026)
• Forter — Travel fraud platform overview and OTA case studies, integration with Adyen and Stripe (2026)
• Adyen — Risk management and OTA payments documentation (2026)
• Booking Holdings — 2025 10-K and Q1 2026 earnings on performance marketing spend and supply growth (2026)
• Expedia Group — 2025 10-K and One Key loyalty program disclosures (2026)
• Airbnb — Engineering blog on data platform (Airflow, Presto/Trino) and trust/safety (2025-2026)
• Databricks — Expedia Group and Hopper customer references for travel ML workloads (2026)
• Salesforce — Sales Cloud Enterprise and Unlimited pricing for travel and supply CRM use cases (2027)