Data lived in a 12-year-old PIM with no API and six teams keeping parallel spreadsheets. We modelled 41 data classes, ran a reversible migration in parallel for five weeks, then cut over on a Tuesday morning.
The problem
A building-materials distributor ran its catalogue from a PIM installed twelve years earlier. It had no API, so six teams — e-commerce, print, marketplaces, sales, support and procurement — each kept their own spreadsheet export and edited it by hand. A price change took three days to reach every channel, and the channels rarely agreed.
The approach
We spent the first two weeks on the data model alone: 41 classes, an inheritance plan for product families and variants, and a classification store for the 2,000-odd technical attributes that vary by category. The model was signed off by every team that owned data before we wrote any import code.
The migration ran in parallel for five weeks. A nightly job compared every SKU in the old and new systems and reported the differences. We cut over once the diff had been empty for seven consecutive days — on a Tuesday morning, so the whole team was in.
The result
A change made in Pimcore now reaches the storefront, the marketplaces and the print pipeline in about four minutes. The six spreadsheets are gone.
Client name, figures and images are placeholders.