Modernizing food data for a global weight management leader
Consolidating years of fragmented food data into a unified, high-performance platform
About
The client is a leading global weight management company whose digital platform enables members to track food and nutrition as part of their wellness programs. The company's food data infrastructure had evolved into a collection of disparate systems: multiple legacy databases, siloed search tooling, and food data tightly coupled with program logic — a structure that had become a meaningful constraint on performance, flexibility, and growth.To address this, the company engaged Lumenalta®to lead a comprehensive food data modernization initiative. Lumenalta®'s team embedded directly with the client's engineering organization, working across database migration, search infrastructure, API modernization, and data quality — serving as an integrated delivery partner from discovery through production release.
Zero
Service interruption on the first production release
100%
Data consistency maintained across the migration
8+
Years of accumulated system complexity replaced with a clean architecture
Challenge
The client's food data infrastructure had grown into a system of formidable complexity over more than eight years. Two separate legacy document-based database systems operated independently — one managing core food data, one managing member-created content — each with its own schema, logic, and maintenance burden. This fragmented architecture created duplicated data across markets, blocked data warehousing, and required specialized database expertise maintained through a third-party support vendor at a cost of approximately $200–250K per year.Performance was a persistent concern. The legacy system averaged 300ms per API response, with CPU usage spikes stemming from the absence of proper connection pooling. The document-based structure also prevented multi-lingual support within a single market and blocked integration with the company's own data science work on food taxonomy — capabilities central to the long-term product roadmap.
Search compounded the challenge. The client relied on legacy third-party search platforms managing 300 indices replicated across two separate accounts, with result quality constrained by outdated taxonomy and limited relevance ranking. And the tightest constraint of all was architectural: the food database was deeply coupled with the company's program logic, making it nearly impossible to evolve either system independently. Any path to global expansion or future product flexibility required untangling years of accumulated interdependencies first.
Solution
Lumenalta®led the modernization in two phases — database consolidation and API migration first, followed by search infrastructure replacement.Database Consolidation & Migration
- Lumenalta®designed and executed a migration from two separate legacy document-based database systems to a single unified PostgreSQL database, bringing all food data types — vendor foods, member foods, recipes, and meals — under a normalized relational schema with clearly separated concerns across food, nutrition, portions, and program data.
- A dual-write approach was used throughout the transition: changes were written to both the legacy and new systems in parallel, ensuring zero data loss and full continuity of service at every stage.
- Real-time change data capture was implemented using Debezium, with Apache Kafka as the message broker — providing continuous synchronization between systems during the full migration window.
- Batch reconciliation processes ran alongside real-time replication to validate completeness, integrity, and timeliness, with the data quality bar set at 100% consistency.
- A feature flag–based phased rollout gave the client granular control over traffic migration, with rollback capability maintained throughout the entire release cycle.
- Lumenalta®completed a V8 to V9 API endpoint migration in parallel, modernizing the interface layer in step with the underlying data store.
Search Infrastructure Modernization
- Lumenalta®scoped and planned the replacement of the client's legacy third-party search infrastructure — spanning 300 indices across two accounts — with OpenSearch 2.9.0.
- The new architecture replicates data from the unified food database directly into OpenSearch indexes, enabling improved taxonomy via the company's Knowledge Graph and delivering better relevance, ranking, and performance than the legacy solution.
- A parity testing framework was developed to validate search result equivalence before any traffic migration, with an incremental percentage-based rollout and continuous performance monitoring planned for the production transition.
- An analysis of the member food search indices found that 90–95% could be decommissioned, significantly reducing the ongoing index management burden.
Results
Architecture & delivery- Phase 1 database migration completed; first production release deployed with passing smoke tests and zero service interruption
- All QA data indices populated and validated; V8 to V9 API migration completed
- Dual-write architecture and continuous batch reconciliation maintained full data consistency throughout the transition
- Feature-flag rollout enabled a controlled, reversible migration at every stage of the release cycle
Operational efficiency
- Migration away from the legacy database platform removes the requirement for a specialized third-party database support vendor — a contract of approximately $200–250K per year — setting the client up for meaningful cost reduction
- Deprecation of legacy third-party search platforms, targeted for completion by mid-2026, eliminates associated recurring licensing costs
- A normalized relational schema unblocks data warehousing, previously impossible under the schemaless legacy structure
Platform capability
- Food data is now decoupled from program logic, enabling the client to evolve its wellness programs independently of food data for the first time
- The unified database architecture introduces foundational support for global food data, removing a key structural barrier to multi-market expansion
- OpenSearch integration creates a direct path to incorporating data science–driven taxonomy improvements, setting the stage for meaningfully better member search experiences
With Phase 1 complete and the search migration underway, the leading global weight management company now operates on a modern, consolidated data foundation. Eight years of accumulated complexity have been replaced with a clean relational architecture built for scale — one that reduces operational overhead, enables data-driven search improvements, and positions the company to expand its food platform across global markets. The foundation is in place. What comes next is engineered to grow.
