Stuart Totterdell
Technical Director
In a mid-market plant or distribution centre, the item master rarely looks strategic. It looks like a maintenance chore - codes, descriptions, units of measure, BOM lines, suppliers, lead times, and costing fields someone filled in during go-live and never really revisited since.
Treat it like admin and it will run the business anyway, just badly. Wrong unit-of-measure conversions inflate or deflate stock. Duplicate SKUs split demand so planning never sees the true picture. Soft or incomplete BOMs turn material requirements into a negotiation rather than a calculation, and lead times copied from a supplier quote three years ago drive purchase orders that arrive late or early for no reason anyone can actually defend. Standard costs drift further from reality with every passing month, until margin meetings start to feel like archaeology. This is not "data not ready for AI" in the abstract sense - it is ERP operational master data, the records that tell the system what a thing is, how it is made or bought, how it is stocked, and what it should genuinely cost. If those records lie, every downstream process lies right along with them, and does so with a completely straight face.
The silent tax on every shift
Watch a week of exceptions closely and the tax becomes obvious. Goods-in queries a pack size that does not match the purchase unit of measure. Production issues a kit that does not match the BOM revision engineering believes is current. Finance argues with operations about scrap because the item attributes never captured the yield the line actually gets in practice. Customer service promises a date the ATP calculation cannot genuinely support, because safety stock and lead times have become folklore typed into fields years ago and never checked since.
Each incident looks small and local on its own, but together they form a permanent drag on service, cash, and trust. People compensate with tribal knowledge and spreadsheets, which only starves the masters of correction further, and the ERP remains the system of record in name only while the operation actually runs on exceptions underneath it.
Why this blocks change programmes
Every serious ERP change - an upgrade, a reimplementation, a bolt-on, or a migration - assumes you can move or remap items, BOMs, and related masters cleanly. If those masters are duplicated, inconsistently coded, or missing mandatory attributes, the programme quietly becomes a data archaeology project wearing a software hat. The same wall stops AI pilots that touch planning, inventory, or procurement. Models trained on contaminated item histories do not give you insight - they give you confident nonsense dressed up as analysis.
Cleaning masters is not a side workstream you schedule after the exciting demo. It is the gate everything else has to pass through, which is exactly why data and systems integration work in ERP environments so often starts with item, customer, and supplier truth rather than another dashboard. Connect dirty masters together and all you accelerate is the wrong answers.
Ownership is the missing role
Ask who owns the item master end to end - creation standards, change control, unit-of-measure policy, BOM integrity, obsolescence, and supplier-item linkage - and in many mid-market businesses the honest answer is "a bit of everyone." Engineering creates codes, purchasing edits lead times, warehouse invents aliases, finance touches costs, and IT gets asked to "fix the data" without ever being given the authority to refuse a bad request. Nobody has the mandate to say no, and nobody is actually measured on master quality.
Without ownership, cleanup projects tend to fail the same way every time: a heroic spreadsheet exercise, a bulk load, three months of visible improvement, and then entropy quietly returns because the operating model itself never changed. Master data is not a one-off scrub. It is an operational control, and it needs to be treated like one.
Cleanup before the next big bet
If you are approaching an ERP renewal, an AI inventory pilot, a WMS overlay, or a multi-site rollout, put master quality on the critical path early rather than discovering it halfway through. Define what "good" actually means for a live item - unique identity, valid unit of measure and conversion, a costing method that matches reality, BOM revision discipline, a clear lead time source and review date, and status codes that genuinely drive behaviour - and then measure the estate honestly against that bar. Duplicate and dormant items get retired or merged, critical attributes get real owners and change rules, and high-volume SKUs get cleaned first so the tax drops where it actually hurts the most.
This is unglamorous work, but it is also the highest-leverage ops work many businesses can do without buying a new platform at all. You improve planning, stock accuracy, and costing on the system you already run, and you stop poisoning the next programme before it has even started.
What good looks like in practice
A business that genuinely controls its masters can answer basic questions without convening a war room: what do we actually stock under this demand, what does this product actually consume, what should this item cost at standard and why did last month diverge from it, and which supplier lead time are we trusting, and when was it last verified.
Those answers come from governed ERP masters combined with disciplined business automation around create-and-change workflows - approvals, mandatory fields, exception queues - not from another weekend of Excel heroics. Your item master is already running the business. The only real choice left is whether it runs it with discipline, or continues as a silent tax on every order, every kit, and every board pack that pretends the numbers underneath it are solid.
