We wrapped a hotel group's ageing PMS in a modern API layer with caching and live sync, so new guest features no longer need risky changes to the core system. It runs at 99.99% availability, with 45ms API responses.
Industry: HospitalityService: SaaS & SoftwareUpdated
Overview
The challenge
The group's growth was held back by its ageing PMS, which had no modern API. Adding simple guest features, such as digital room keys or live pricing, took 6-month development cycles and risky changes made directly to the database.
Requests were handled one after another, so a single slow query could stall the whole booking flow. Each new digital partner needed its own hard-coded connector and XML mapping.
Traffic from mobile and web apps often overloaded the older database and caused outages across a whole property.
Slow API responses made guests abandon direct bookings, and room inventory fell out of step with online travel agencies (OTAs).
Technical debt blocked personalised guest services, which weakened the brand's luxury positioning.
The solution
Using the strangler fig pattern, a facade now receives every digital request. It answers read-heavy queries from a fast cache and protects the PMS's own write path.
A sync bridge streams changes from the older database into the cloud platform in under 200ms.
New cloud microservices handle guest identity and room state, kept apart from the PMS's inventory logic.
Before and after
Before
One request at a time
Requests queued in line, so one slow query could halt the entire booking flow.
After
Fast and non-blocking
Event streams and cached reads return p95 responses in 45ms.
Before
Bigger hardware
Scaling meant expensive hardware upgrades and planned downtime.
After
More services on demand
Services scale out straight away during flash-sale traffic without touching the older core.
Before
Manual, 6 months
Each new digital partner needed a custom, hard-coded connector and XML mapping.
After
Standard APIs
Standard REST and GraphQL APIs let the team connect new services in weeks rather than months.
Key features
Room inventory and rate data are cached at the network edge, cutting round-trip times for global booking engines by 85%.
Business impact
45ms responses
A dedicated security layer removes personal information and tokenises card details before any request reaches the unencrypted older core.
Business impact
Simpler PCI DSS audits
Automatic failover stops a slow PMS from dragging down the guest-facing mobile apps.
Business impact
99.99% uptime
Results
85% Response Time Reduction
Serving reads from the cache, instead of the older PMS, removed the 800ms wait in the guest journey.
Before: 800msAfter: 45ms
99.99% Availability
Circuit breakers and automatic scaling meant zero outages during holiday traffic surges.
Before: 96.4%After: 99.99%
5x Faster Release
A standard integration layer lets the digital team bring new partners live 5x faster.
Before: 6 MonthsAfter: 4 Weeks
Faster delivery
Reusing these components saved about 8 weeks of setup and helped the platform go live in 12 weeks.
Gave the team a tested way to read from and write to the older PMS without changing its database.
Provided the routing, OAuth2 security and caching rules behind the new facade.
Tracked cache hits and response times at the network edge across the 200+ properties.
Kept the cost of scaling services during flash sales and holiday peaks in check.
Trust and control
Payment data is tokenised at the edge, so no unencrypted card data reaches the older core, in line with PCI DSS.
Secure paymentsThe strangler fig approach allowed a step-by-step move with no disruption to day-to-day operations at any property.
Zero downtimeAutomatic cloud scaling keeps the API layer responsive whatever the transaction volume.
Scales with demandOn completion, the client owns 100% of the IP in the connectors, facades and deployment systems.
100% ownedIts 200+ properties ran on a 15-year-old on-premise PMS with no modern API. Adding features such as digital room keys or live pricing took 6-month development cycles and risky changes to the database. Traffic from mobile and web apps often overloaded the system and caused property-wide outages.
We put a strangler fig facade in front of the PMS. It answers read-heavy requests from a Redis cache and protects the older write path, while a sync bridge streams database changes to the cloud in under 200ms. New cloud services handle guest identity and room state, apart from the PMS inventory logic.
API responses dropped from 800ms to 45ms, and availability rose from 96.4% to 99.99%, with no outages during holiday peaks. New digital partners now go live in 4 weeks instead of 6 months. The platform launched in 12 weeks, and card data is tokenised before it reaches the older core.
If your PMS makes every new guest feature a six-month project, we can help you put a fast API layer in front of it without replacing it.