Project Name
35-Minute Inventory Sync Cut to 3 Minutes With Server-Built SQLite Delivery
A SaaS-based inventory counting platform operating across more than 20 countries with over 250,000 app downloads and more than 30,000 annual counts enables retail organisations to conduct full store counts, cycle counts, and ad hoc counts using their own in-store devices. Offline scanning capability is a core product promise. As the platform’s retail client base scaled and individual catalogues grew to millions of SKUs, the time required to prepare the on-device barcode database before each count became a critical performance liability. Applying its AI-First approach, Ksolves redesigned the sync strategy around a server-built SQLite .db file delivered as a compressed archive – completing the same sync in 2 to 3 minutes, eliminating all on-device write load, and making the pre-count delay a solved problem.
- 30 to 35 Minute Database Build Before Every Count: Downloading raw barcode data and inserting millions of records into Realm took 30 to 35 minutes per sync - blocking every count session before a single scan could be made.
- Millions of On-Device Realm Write Operations: Individual Realm insertions on mobile hardware saturated CPU and storage I/O with no architectural path to acceleration short of a fundamental redesign.
- Device Overheating and Battery Drain: Sustained write loads across the 30-minute window caused overheating and significant battery drain before counting started, reducing device readiness across the fleet.
- Partial Corruption on Interrupted Syncs: A crash or connectivity loss mid-insertion left Realm in an incomplete state requiring a full re-sync from scratch, compounding the original delay.
- No Scalability Path as Catalogues Grew: Every additional SKU added incrementally to the sync window - making the problem structurally worse with every client growth milestone.
- Platform Reputation Risk: A 35-minute pre-count delay was inconsistent with the platform's market promise of fast, on-demand inventory counting and created churn risk among clients evaluating alternatives.
Ksolves shifted all database construction to the server. A fully indexed SQLite .db binary is built server-side and delivered as a compressed archive. The device downloads, unzips, places the file, and opens a connection - zero insertions, zero write overhead. Build once on the server, deliver once to the device.
- Server-Side SQLite Database Build Pipeline: Backend process assembles the complete barcode catalogue into a fully indexed SQLite .db binary on every catalogue update - schema, indexing, and relationships applied server-side, file arrives ready for immediate querying.
- ZIP-Compressed Distribution via Cloud Object Storage: .db file compressed into ZIP archive and pushed to cloud object storage for CDN-backed delivery across 20+ countries - single atomic download replaces the full multi-step extraction pipeline.
- Intelligent Update Strategy: App evaluates each update and chooses the faster path - full .db file replacement for large catalogue changes, or targeted on-device insertions for smaller ones - balancing speed and performance by update size.
- Path-Based Device Placement: .db file written to a pre-agreed filesystem path. Existing SQLite connection configuration points to that path - no changes to counting logic, query code, or UX flows required.
- On-Device Write Load Eliminated: Zero record insertion during full sync. CPU and I/O usage drops to a file download and copy. Devices stay cool, battery consumption minimal, sync runs in the background.
Technology Stack
| Category | Technology |
|---|---|
| Database | SQLite |
| Infrastructure | ZIP Compression |
| Native Mobile | Android (Kotlin) & iOS (Swift) |
- Sync Time Cut From 35 Minutes to Under 3: Pre-built SQLite .db delivery completes in 2 to 3 minutes - over 90% reduction from the 30 to 35-minute Realm insertion process that blocked every count session.
- Millions of On-Device Writes Eliminated: Zero write operations during full sync. Database arrives fully built and immediately queryable. No CPU or I/O saturation.
- Count Sessions Start Near-Instantly: Sync completes in the background so counting begins the moment staff are ready - platform's on-demand positioning restored.
- Sync Corruption Eliminated: Atomic .db replacement means a failed download is a clean retry - no partial database state, no 30-minute rebuild required.
- Sync Performance Unaffected by Catalogue Growth: Device workload is always a single file download regardless of record count - the scalability constraint permanently removed.
“The sync that used to hold our clients’ teams up for the better part of an hour now runs quietly in the background in a couple of minutes. Store staff arrive to a system that’s already ready, and we haven’t seen a corrupted or incomplete sync since the new architecture went live.”
– Head of Product / CTO.
A SaaS inventory counting platform whose store teams waited 30 to 35 minutes before every count due to millions of on-device Realm insertions, suffering overheating, battery drain, and corruption on interrupted syncs, was transformed through Ksolves web and mobile development services. Server-built SQLite .db delivery now completes the same sync in 2 to 3 minutes with zero on-device write load. Over 90% sync time reduction. Corruption eliminated. Count sessions near-instant. Sync performance scalable regardless of catalogue size.
Is Your Mobile Platform Spending 30 Minutes Preparing Data When It Could Be Ready in 3?
