Batch-level traceability for a knitwear exporter supplying EU brands

EU fashion brands were sending product passport questionnaires before the textile rules were final. The group needed to trace a carton back to the yarn lots inside it.

Length
12 weeks, piloted on two product lines

Three questionnaires and no batch records

The client. A knitwear group near Dhaka with four factories and about 6,000 workers, supplying three EU fashion brands, with yarn and fabric from 14 suppliers.

Each brand sends its own readiness questionnaire. Fibre composition comes from supplier declarations, recycled-content certificates sit in email, and every file spells the facility names differently.

Nobody can link a carton of finished goods to its yarn lots. The sustainability team answers each questionnaire by hand in about eleven days, and expired certificates surface only when a brand asks.

How the engineers did it

  1. Keep three levels apart

    The data model holds model, batch and item separately, so it can answer at whichever level the final textile rules require.

  2. One identifier per site

    All 14 suppliers upload declarations and certificates against one shared list of facility identifiers, with expiry dates captured on upload.

  3. Link the batches

    Yarn lot, knitting batch, dye lot, cutting order and carton are joined from existing ERP events and the floor's lot cards.

  4. Register every certificate

    Each certificate has an owner, a scope and an expiry. Thirty days before it lapses, the supplier and the merchandiser each get a task.

  5. Your team answers the brands

    One data model exports into each brand's template with its evidence attached, and the sustainability team sends it.

For your engineers

What was built

  • Facility and supplier master data with one identifier per site
  • Evidence hub with upload, scope, expiry and reminder tasks
  • Batch genealogy from ERP events and lot cards, with a coverage dashboard
  • Certificate registry and expiry alerts
  • Template export for three brand questionnaires
  • Data quality dashboard: missing fields by supplier and style

Stack

Python, PostgreSQL, ERP event feed, Dynodoc, Metabase

Length

12 weeks, piloted on two product lines

What changed

MeasureBeforeAfter
Answer to one brand questionnaireabout 11 days by hand2 days
Pilot cartons traced to their yarn lotsnone9 in 10, across two product lines in one season
Certificate expiriesfound when a brand asksflagged 30 days ahead

What we learned

  • Start before the textile rules are final. The gap is at batch level, and closing it takes a season.
  • Certificate expiry is a silent failure. A registry with reminders is the cheapest fix.
  • One supplier identifier list prevents many questionnaire errors.

Bring the brand questionnaire you answer by hand

The diagnostic traces one product line from carton to yarn lot on paper first, then prices the system that would do it every season.