ERP Data Migration: Why Dirty Item Masters Kill Projects

Operbit · July 28, 2026

Where ERP Projects Actually Die

Ask anyone who has watched an implementation fail and you'll hear about software, vendors, or budgets. Look closer and you'll usually find the same corpse: master data. Items with three names for one product. Customers duplicated across a decade of sales reps. Units of measure that mean different things in different spreadsheets. The system was configured beautifully, around data that couldn't carry it.

What "Dirty" Costs, Specifically

• Duplicated customers split purchase history, so credit limits, pricing tiers, and sales reports are all quietly wrong.
• Inconsistent SKUs make inventory counts unreconcilable: the system says 40, the shelf says 12, and the other 28 live under a second item code.
• Wrong units of measure are the expensive one: buy in cases, sell in eaches, record it inconsistently, and every margin number downstream is fiction.
• Missing costs mean the ERP can't value inventory or compute margin, the two reports you bought it for.

The Uncomfortable Rule

Cleaning happens before migration or it doesn't happen. Once dirty data is inside the new system, it acquires transactions, and every transaction is another root holding the weed in place. Companies that say "we'll clean it up in Odoo" are describing a second project, one they will not fund.

How to Prepare, Practically

• Appoint one owner per master. One person decides what an item is called, one decides what a customer is called. Committees produce duplicates.
• Dedupe with rules. Legal name wins, one tax ID one customer, one physical product one SKU.
• Kill the dead weight. Items with no sales in 24 months and customers with no orders in 36 don't migrate. Archive them in the old system.
• Standardize units before costs. Cost per what? is the question that unravels most item files.

The Payoff

Clean masters migrate in days and make everything after cheaper: testing finds real issues instead of data noise, training uses recognizable records, and the first month-end reconciles. It's the difference between the 90-day plan holding and slipping, and it's a line you control completely, before spending a dollar on software. Data-shape is one of the four inputs in our cost calculator for exactly this reason.