Why Tool Room Inventory Goes Bad: The BOM Revision Gap
Walk through the material store of any tool room — whether it makes stamping dies, injection molds, or die casting tooling — and you will find the same rack. Blocks that were cut for a job that changed. A casting at the wrong height. Catalogue components in their original packaging, for a tool that shipped two years ago. Everybody knows the rack is there. Nobody quite knows how to stop it growing.
It is usually filed under purchasing. Over-ordering. Poor planning. Someone should have checked.
That diagnosis is comfortable and wrong — and it is wrong in a way that guarantees the rack keeps growing — because almost every item on it was bought correctly, on time, at the right price, against an approved document.
The document changed afterwards.
How a good purchase order becomes dead stock
Follow one casting through a normal project.
Day 1. Design releases the BOM. Upper shoe casting, 1380 × 890 × 490. Long lead item — pattern work, pour, cool, rough machine. Purchasing does its job and places the order on day 2, because waiting is how tools get late.
Day 9. Tryout feedback from a similar tool comes back. The designer increases the shoe height to 565 to get more spring travel. Correct engineering decision, made for good reasons — and updated in the NX, Creo, or SOLIDWORKS model that same afternoon.
Day 10. A revised BOM goes out. It is a 200-row spreadsheet attached to an email that says "revised BOM attached, pls note changes."
Day 10, purchasing side. The buyer has that email, seven others, and a delivery meeting. The file looks like the last one. Comparing 200 rows by hand against the previous version is a half-day job, it is not what anyone is measured on, and it is not a task anyone does reliably twice. So the buyer scans it, sees nothing obviously new, and files it.
Day 34. The casting arrives at 490. The design has needed 565 for three and a half weeks.
Now count the damage. The 490 casting cannot be returned — it is a custom pour. It cannot be used — there is no other tool with that footprint. It goes on the rack. The 565 casting has not been ordered, so the full lead time starts again from today, and the tryout date moves. You pay for two castings, you get one tool, and it is late.
Nobody in that sequence made a mistake in their own job. Design made the right change. Purchasing bought on time, at the right price, against the approved BOM. The handover between them had no mechanism.
The missing artefact
Purchasing did not need a new BOM on day 10. They needed the difference.
Not "here is Revision 3" but:
Changed — UPR SHOE CAST: height 490 → 565; stock size 1380 × 890 × 490 → 1380 × 890 × 565
Changed — ADAPTER LIFTER: qty 4 → 3
Removed — PUSHER PIN D322
Removed — BASE PLATE D314 (qty 4)
Unchanged — 104 rows
Read that as a buyer and the actions are immediate and unambiguous:
- Amend the casting PO today, before the pattern is cut.
- Reduce the lifter order by one.
- Cancel two lines outright.
- Ignore 104 rows entirely.
That is a five-minute task instead of a half-day comparison — which is exactly why it actually gets done. The reason this handover fails is not that buyers are careless. It is that the artefact they need has to be produced by hand, and work that has to be produced by hand under delivery pressure does not get produced.
Junk inventory is the cost of a missing diff.
Where else it comes from
The superseded revision is the largest source, but there are three smaller ones with the same root cause — a BOM that is assembled and reconciled by hand.
Duplicates under different names. The same nitrogen cylinder bought twice, six months apart, under two part names, because nothing in the process recognises that they are the same item. Both sit in stores. Neither is found when needed, so a third gets ordered.
A catalogue part bought as a special. Nobody recognised it as a standard part, so it was drawn, quoted and made. It cost several times the catalogue price and took weeks longer. Its unused twin from the last project is already on your shelf.
Phantom parts from a badly split assembly. A purchased sub-assembly — a gas spring unit, a hot runner manifold — gets split into its individual bodies during BOM preparation in NX, Creo, or SOLIDWORKS. Each body becomes a row. Purchasing tries to source components that were never separately available, and orders something approximate to fill the line.
All four sources are variations of the same failure: the BOM is not a controlled record of what the tool needs. It is a spreadsheet snapshot of what somebody believed the tool needed, on the day they typed it.
Why "be more careful" does not work
The standard response is a process instruction: purchasing must verify every revision against the previous one; design must highlight changes in the covering email; someone must maintain a change log.
These fail in the same way every time, and it is worth being precise about why.
- Manual verification competes with delivery pressure, and delivery pressure always wins. A half-day comparison loses to a delivery meeting every single week.
- Highlighting changes by hand is itself error-prone. The changes the designer remembers to mention are not always the changes that matter to purchasing. A quantity change from 8 to 6 on a standard part rarely feels worth mentioning. It is worth ₹40,000.
- Change logs go stale the first week someone is on leave. Any control that depends on a person remembering to update it has a known half-life.
The fix is not more diligence. It is making the difference list a by-product of publishing a revision rather than a separate task someone has to remember to do.
What good looks like
Four properties. Any system that has all of them will stop the rack growing; any system that lacks them will not, regardless of what else it does.
1. One current revision, in one place. Not an attachment. A link. Purchasing, program management and the tool room open the same record, and "latest" means latest, not whoever emailed most recently.
2. Full history, with approval. Every revision retained, with who published it and when, and an explicit approve or reject. When something goes wrong six months later, the question "which revision were we buying against?" has an answer.
3. Automatic comparison between any two revisions. Added, removed, changed — with the exact field, the old value and the new value. Produced by the system on publish, not assembled by a person on request. This is the single control that stops the day-9 casting.
4. Standard parts identified against a shared, growing library. So catalogue items are recognised as catalogue items every time — whether your design team works in NX, Creo, SOLIDWORKS, CATIA, or Solid Edge — duplicates are caught before they become two purchase orders, and purchased assemblies are never split into phantom components.
Notice that three of the four are about change, not about the initial release. That is the correct emphasis. The first BOM is usually fine. It is revisions two, three and four — made under pressure, communicated by email — where the inventory is created.
A test you can run this week
Take one tool you delivered in the last year — stamping die, injection mold, or die casting tool. Pull the material issued against it and the material actually fitted to it. List the difference.
Then, for each item on that list, ask one question: was this bought against a BOM that later changed?
You will not need a spreadsheet to see the pattern. And once you have seen it, the rack in the store stops looking like a purchasing failure and starts looking like what it is — a record of every design change that was never precisely communicated.
Closing the gap
SmartAPQP — available through DDSPLM — publishes each BOM revision to one shared link with full history, approve/reject in place, per-row comments, and a comparison against any earlier revision showing added, removed and changed rows with the exact field, the from-value and the to-value. Standard parts are matched automatically against your own shared library, growing every time an engineer confirms a part across NX, Creo, SOLIDWORKS, CATIA, and Solid Edge assemblies. Stock sizes and part photos are generated from the model and regenerated whenever a part changes — so what purchasing and your vendors see is the current design, not last month's.
Because export templates are defined once, re-issuing after a change is two clicks to every vendor on the same day it is approved. A change that reaches purchasing on day 10 costs an amendment. The same change reaching them on day 34 costs a casting.
Try it against a real project. Send us one assembly from a tool you have already built — stamping die, injection mold, or die casting tool — and we will return a published BOM plus a comparison of two of its revisions. Then check that comparison against what actually ended up on the rack.
← Back to all posts