Skip to main content
Back to Blog
Performance

Speed Caching: Reducing Supplier Load by 80%

Travelgate's Speed is a dynamic caching layer that sits between your application and supplier systems. Correctly configured, it is the single highest-impact performance optimisation available in the Hotel-X stack — and it requires almost no code change to activate.

Stunning Club Engineering March 2026 8 min read
Fast data flow representing Speed caching

80%

Reduction in live supplier API calls

3x

Faster average Search response time

TTL

Configurable per booking window

0

Cache invalidation calls required

What is Speed?

Every Search request in Hotel-X calls supplier systems in real time to fetch live availability and pricing. At low volumes this is fine. At scale — a B2C platform running thousands of concurrent searches, or a chatbot serving a high-traffic marketing campaign — the cumulative supplier load becomes a bottleneck: slow responses, rate limiting, and degraded user experience.

Speed intercepts Search requests at the Travelgate layer. If a matching cached response exists within the configured TTL, it is returned immediately without contacting the supplier. The supplier sees only the cache-miss requests — typically 20% of total volume — while your users see the full result set at cache speed.

Speed is activated in HotelSettingsInput — no code change is required beyond setting the mode and TTL. It is transparent to the booking flow: the optionRefId returned from a Speed-cached Search is valid and can be passed to Quote exactly as a live Search result would be.


Standard vs Fast Mode

Standard Mode

mode: STANDARD

On a cache miss, Speed calls the supplier in real time and caches the result for subsequent requests. The first request for a given search combination is always live; subsequent matching requests within the TTL window are cached. Recommended for most deployments.

Fast Mode

mode: FAST

On a cache miss, Speed returns an empty result rather than calling the supplier. A background job populates the cache for that search combination. The next request for the same combination returns from cache. Maximum supplier protection, minimum latency — at the cost of occasional empty results on first request.

For B2C consumer-facing applications, Standard mode is the right default — users expect results on first search. Fast mode is appropriate for pre-warming scenarios, such as destination landing pages where you control the search parameters and can pre-populate the cache before users arrive.


TTL Configuration

TTL (time-to-live) determines how long a cached Search result is considered valid before Speed calls the supplier again. Setting it correctly is the most important Speed configuration decision.

The right TTL depends on your booking window distribution:

Same-day / next 72 hours

5–15 minutes

Inventory moves fast at short notice. Long TTL risks showing sold-out options.

1–4 weeks ahead

30–60 minutes

Moderate volatility. Balance freshness against supplier load.

1–6 months ahead

2–6 hours

Low volatility for availability; prices more stable. Longer TTL is safe.

6+ months ahead

12–24 hours

Advance inventory is highly stable. Maximise cache hit rate.

TTL is configured in HotelSettingsInput as a number of minutes. If you serve multiple booking window profiles, use separate client codes per profile so each can have an independently tuned TTL.


When to Enable Speed

  • B2C platforms with concurrent user search volume — any shared destination search benefits from cache sharing across users
  • AI chatbots serving high-traffic campaigns where the same destination/date combinations are searched repeatedly
  • Destination landing pages pre-loaded with curated hotel results
  • Corporate booking tools with predictable date windows for upcoming quarter travel
  • Any deployment where supplier rate limiting is a risk factor

When Not to Use Speed

Speed is a Search-level optimisation only. It must not be applied to Quote or Book — those steps must always call the supplier live to ensure price accuracy and reservation confirmation.

  • Quote — must always be a live supplier call. Revalidation is the entire point of Quote.
  • Book — must always be a live supplier call. A cached Book response is a booking that was never confirmed.
  • Very low-volume integrations where the overhead of cache management exceeds the benefit
  • Suppliers with extremely frequent inventory changes (e.g. auction-based pricing) where any TTL produces misleading results

Monitoring Cache Effectiveness

Travelgate's Stats dashboard shows Search request volume broken down by cache hits and cache misses per supplier access. The target hit rate for a well-tuned Standard mode deployment is 70–85%.

A hit rate below 50% usually means the TTL is too short for your booking window distribution, or your search parameters are too varied (e.g. a wide radius query with many unique combinations).

A hit rate above 90% with high DeltaPrice rejection at Quote suggests the TTL is too long for the supplier's pricing volatility. Reduce TTL for that access until the DeltaPrice rejection rate drops to below 3%.

"The relationship between Speed TTL and DeltaPrice rejection rate is the most important performance dial in the Hotel-X stack. Get it right and your users see fast, accurate results. Get it wrong and they see fast, inaccurate results — which is worse than slow accurate ones."


Activation

Speed is activated by setting the speed field in HotelSettingsInput in your Search request. No separate provisioning is required — Speed is available to all Hotel-X Buyers.

settings: {
  context: "YOUR_CONTEXT",
  client: "YOUR_CLIENT_CODE",
  auditTransactions: false,
  testMode: false,
  speed: {
    mode: STANDARD,           # or FAST
    ttl: 30,                  # minutes
    forceRefresh: false       # set true to bypass cache for this request
  }
}

Set forceRefresh: true on individual requests when you need guaranteed live data — for example, immediately before a Quote to validate that Search options are still current. This bypasses the cache for that single request without changing the global TTL setting.

Search with Speed enabled

Tanya's Hotel-X integration uses Speed in Standard mode — fast results, live revalidation at Quote.

Try with Tanya
Speed Caching: Reducing Supplier Load by 80% | Stunning Club — Stunning Club