Ecommerce ops · Production
Booking.com's Enrichment System delivers 100,000 recommendation enrichments per second at 99.99% availability
The problem
Booking.com's previous enrichment approach for recommendation data was complex to implement, tightly coupled with other platform components, and limited in reusability — forcing teams to recreate similar enrichments from scratch and slowing self-service innovation.
Workflow diagram · grounded in source
1
Field mask request
trigger
“users can now request enrichments directly via field masks”
2
Field mask detection
routing
“it checks for the presence of ENRICHED_IMAGES field mask. If found, it executes the corresponding enrichment in the Enrich stage”
3
Data fetch per enrichment
integration
“fetching data from a personalized ranker city search system”
4
Enriched response delivery
output
“The enriched data is then attached to the response”
Reported outcome
The new Enrichment System executes up to 100,000 enrichments per second with 99.99% availability, and simplifies enrichment development so that implementing a new enrichment requires only defining a field mask rather than navigating complex code changes.
Reported metrics
Enrichments per second100,000
System availability99.99%
Reported stack
GraphQL
Source
https://medium.com/booking-com-development/unlocking-the-power-of-customization-how-our-enrichment-system-transforms-recommendation-data-e71832fc4ef4
Read source ↗Frequently asked questions
What did this team achieve with this AI workflow?
The new Enrichment System executes up to 100,000 enrichments per second with 99.99% availability, and simplifies enrichment development so that implementing a new enrichment requires only defining a field mask rather…
What tools did this team use?
GraphQL.
What results were reported?
Enrichments per second: 100,000; System availability: 99.99% (source-reported, not independently verified).
How is this ecommerce ops AI workflow structured?
Field mask request → Field mask detection → Data fetch per enrichment → Enriched response delivery.