Manual steps between systems
Two people reviewed and corrected around 1,500 products by hand every season.
All work
Carhartt WIP, 2026
Automating the product data flow
Every product travels from PLM and ERP into the PIM, then to the shop system and the CMS. I audited that flow end to end, found what could be automated or fixed at the source, and worked with engineering to connect or replace the manual steps.
- Role
- Product lead for product data
- Timeline
- 2026, FW26 and SS27 seasons
- Team
- Content, copywriting, photo studio, production and engineering
- Scope
- PLM and ERP to PIM, Commercetools and CMS
The problem
Product data moved through five systems, and people patched it by hand at every hand-off.
Rules nobody could untangle
313 PIM rules had grown one at a time, with overlapping and conflicting logic.
Wrong or late on the storefront
Products went live before they were ready, categories were guessed from names, and new-arrival dates got lost.
What I did
Mapped the flow end to end
I traced how product data moves from PLM and ERP through the PIM to the shop system and the CMS, and marked every point where someone re-entered or corrected data by hand.
Sorted every step
Each manual step went into one of three groups: automate it, fix it at the source, or keep it manual on purpose. Merchandising choices like featured slots stay with people.
Built the automations with engineering
Category assignment is derived from the PIM and AI rules enrich color and fabric, both in daily use. Outfit links come straight from the photo studio, and release gates keep unfinished products offline.
The call
Results
Automation now covers seven manual steps, built to replace the seasonal hand review.
7manual steps covered by automation
~1.5Kproducts a season targeted, previously reviewed by two people
313PIM rules audited
33,036products checked for category accuracy
What worked
- Auditing before automating, so every automation sits on a clean base.
- Fixing issues at the source instead of patching them further down the flow.
What didn't
- We didn't measure time spent before we started, so savings are estimated rather than tracked.
- Category assignment and AI enrichment are in daily use. Making the remaining automations part of the team's routine is the current focus.
What I'd do differently
I'd roll automation out step by step within the season instead of all at once, so each gain is visible sooner and easier to attribute.