SM-SS-2-0417. Anyone who reads it knows everything at a glance: sheet metal, stainless steel, 2 millimeters thick, serial 417. Until a design change turns the part into aluminum and a choice lands on the table that nobody wants to make: let the number lie, or renumber and drag along every BOM, drawing and stock card it appears on.
This is the classic dilemma of a part numbering system: meaningful numbering feels clever and sooner or later runs aground, meaningless numbering feels bare and turns out to be surprisingly robust. In this article we put the two side by side, with the pros and cons from everyday machine building and sheet metal work, plus a migration path for anyone stuck in a legacy scheme. And because a number only works while it stays unique: who issues numbers, and how do you prevent duplicates? A toolbox such as Thundercad then helps put the meaning neatly into metadata, where it belongs.
The promise of the meaningful number
The appeal is understandable. A number that encodes material, type and thickness is a miniature drawing in text form: the work planner immediately spots the sheet metal lines in the BOM, the warehouse sorts by the first letters, and when searching you simply type the prefix. In a small company with one product group that works fine for years, which is precisely why so many companies started out this way.
The promise rests on a single unspoken assumption, though: that tomorrow's product range fits into today's boxes. When the company grows, product lines get added, or a design changes material, the scheme starts to pinch in places you can no longer easily reach.
Where meaning starts to pinch
It rarely pinches on day one; it pinches on change. The patterns are predictable. The material changes and the number no longer tells the truth, as with the plate above. A category fills up: two digits for the product group seemed generous, until group 99 came into view. A part lands in two boxes at once: is a laser-cut bracket with a bend line sheet metal or frame work, and who decides? Every doubt spawns a discussion, because a wrong box here does not mean a wrong field in a database, it means a wrong number that creeps into everything.
Run the numbers with an assumption: if engineers create five new items per week, spend ten minutes debating the right category per borderline case, and one in five is a borderline case, you are looking at nearly an hour per month of pure classification debate. Not dramatic, but structural. The real damage is in the exceptions it breeds: numbers that lie, because renumbering was too costly. From that moment on, nobody can trust the number anymore, and with that the entire advantage is gone.
Meaningless numbering: dull, and strong for it
The alternative is a bare serial number: every new item gets the next number from a single sequence, nothing more. All meaning moves to fields built for the job: description, material, category, thickness as separate iProperties on the CAD file and as columns in every list. You search on those fields, not by deciphering numbers. A material change then is one field change; the number simply stays, because it never promised anything about material.
Side by side, the trade-off becomes concrete:
| Criterion | Meaningful | Meaningless |
|---|---|---|
| Recognizable at a glance | Yes, as long as you know the code | No, that is what the description is for |
| Robust against change | No, the number must lie or change | Yes, fields change along |
| Issuing a number | Requires classification and debate | Next in the sequence, seconds |
| Scales with growth | Boxes fill up or overlap | The sequence never runs out |
| Error risk at creation | Wrong box happens | Close to zero |
The price of meaningless numbering is discipline on the metadata: an empty material field stands out more than a missing letter in a code. A fixed data card per document type, the way the iProperty Panel provides one, keeps that input consistent, with pick lists instead of free text. The meaning then sits where everyone can filter, sort and correct it.
If the meaning moves out of the number and into the metadata, that metadata has to be in good shape. With Thundercad you manage iProperties through a configurable data card per document type, so every item gets the same fields filled the same way.
Try 30 days freeWho issues numbers, and how do you prevent duplicates
A numbering scheme is only as good as its issuing process. The rule that prevents the most misery: there is one counter, and it lives in a system, not in someone's head. Two engineers who both grab "the next free number" from a list will, guaranteed, one day stamp the same number on two different items. So route issuing through a single source: your ERP, a central item register, or at minimum one managed file with one owner. Supporting ground rules: never reuse numbers, not even from obsolete items, and let gaps in the sequence exist; whoever starts filling gaps turns issuing back into thinking work.
Duplicates, by the way, more often arise the other way around: not two items with one number, but two numbers for the same physical thing, because searching took longer than creating anew. That problem starts on the CAD side and deserves its own approach; we wrote about it earlier in Preventing duplicate items, starting on the CAD side. And how items stay one single truth between CAD and ERP, including who owns which field, is covered in Article management between CAD and ERP: one item, one truth; we deliberately leave that aside here.
Migrating without renumbering everything
Anyone stuck in a meaningful scheme today does not have to start from a blank slate; that is hardly ever feasible anyway, because released items are tied to stock, purchase orders and drawings. A workable path looks like this:
- Set up the metadata first: fields for material, category and type, with fixed pick lists. Fill them for the items in circulation; the information already sits in the old numbers and can largely be derived from them.
- Pick a cutover point: from now on, new items get a meaningless serial number from a range that clearly falls outside the old format, so confusion is ruled out.
- Leave existing numbers alone. An old number that has started lying gets corrected in the metadata and the description, not by renumbering.
- Only renumber at a natural moment: an item that is getting a new revision or a new design anyway can hop over to the new range if that genuinely solves something.
After the cutover, two formats live side by side, and that is fine: the old format dies out on its own with the items attached to it. Clarity matters more than speed: everyone must know that from now on meaning lives in fields and the number is nothing but a name.
Frequently asked questions
How long should an item number be?
As short as the sequence allows. Six to eight digits lasts most manufacturing companies for decades and stays easy to read out and retype. Do watch the systems along the way: Excel, for one, strips leading zeros off a digit string, so treat numbers as text everywhere or start the range above zero.
Is a light touch of meaning acceptable, say a single prefix letter?
One stable trait that never changes, such as a letter marking the number range itself, is manageable. But any trait that describes a property of the item can one day start to lie. The honest question is always: what do you do when this trait changes? If you have no good answer, keep it out of the number.
Where do I start if our numbers already lie?
Start with the fields, not the numbers: capture material and category as metadata, and the damage of a lying number is contained immediately. After that you can pick a cutover point at your leisure. The data cards in Thundercad turn that input into picking instead of typing: try it 30 days free.