inventory_2Garment industry · ERP
Garment sourcing ERP, built from zero and made 100× lighter as it grew
Costing, proforma invoices, order book and a four-stage approval chain for an apparel sourcing group. When data grew, we cut the main screen from 11 seconds to 0.2.
- Client
- An apparel sourcing group, Bangladesh
- Built
- Web apps (Flutter), API (Go), Postgres
- Our role
- Designed and built from zero; ongoing
- Live
- Two production sites, daily use
The product
- check_circleMerchandisers build cost sheets line by line: fabric, trims, packing and other charges, each with consumption, wastage and price.
- check_circleThe system works out total cost and margin. Industrial engineers, audit and the CEO each approve or send back at their own stage.
- check_circleSuppliers' proforma invoices are matched against the cost sheet, so planned and actual material cost sit side by side.
- check_circleEach role has its own app, and everyone sees the same live data.
The challenge as it grew
- check_circleUsage grew to hundreds of cost sheets and thousands of lines. The engineers' main screen began to take about 11 seconds to load, and one factory's Excel report started coming out short.
What we did
- check_circleWe measured before changing anything. Every visit downloaded 76 MB to show 27 rows, and one screen made 28 extra server calls per visit.
- check_circleEach screen now gets only the data it shows, and each record is sent once instead of many times.
- check_circleThe web server compresses data and caches correctly.
- check_circleExcel export now produces the complete list or a clear error, with a progress bar.
- check_circleWe released in small steps and checked each one live, with no rewrite and no downtime.
Results
| Measure | Before | After |
|---|---|---|
| Engineers' main list | 76 MB · ~11 s | 0.76 MB · ~0.2 s |
| App download size | 5.2 MB | 1.5 MB |
| Extra server calls per visit | 28 | 0 |
| Second factory site, product list | 83 MB | 3.45 MB |
| Excel report | Sometimes short | Complete, or a clear error |
What this means for you
Software that works on day one has to keep working at ten times the data. We measure, fix the real cause, and show you before-and-after numbers.
Talk to us about your projectarrow_forward