Part numbering systems design for PLM software
software has been designed to support a wide variety of part numbering
systems, including manually-assigned intelligent part numbers,
category-based semi-significant part numbers, and
easy-to-manage sequential part numbers.
This topic explains why most industries have adopted simple sequential identifiers
as their part numbering system best practice.
This discussion relies on PDXpert PLM software's ability to auto-generate part numbers and support easy free-form (Google™-like) text searches. We'd be happy to explain how PDXpert PLM software's flexible source relationships help companies transition an existing part numbering system to a simpler design.
Goals for part numbering system design
Let's step back a moment and consider our essential requirements for a part numbering system.
The goal of any part numbering system is to uniquely identify the component approved for a specific application. Accurate, consistent, unambiguous part identification is essential for correct product assembly, testing and maintenance.
We must ensure that a new identifier is assigned whenever a variation in attributes has a material effect on the item's form, fit, or function in the application.
A part number allows us to clearly distinguish one part from other parts when the difference is meaningful, and to ignore irrelevant differences. This is what distinguishes non-interchangeable changes from interchangeable changes. We also need to ensure that parts with non-meaningful variations can be used for the same application.
Part numbering system constraints
Any part numbering system design must also consider some purely human elements.
The part numbering system must accommodate high-volume users — people who work with hundreds or even thousands of part numbers each week. Document control analysts, warehouse clerks, purchasing agents, production supervisors, and field technicians need to accurately and concisely convey part numbers, often over the phone in noisy environments.
Part number length
A practical part numbering system design should account for the limits of short-term memory. The "magic limit" is typically considered to be 71. Many years of academic study, verified by real-world experience, proves that data entry errors increase as the number of characters increase. After a certain length, errors increase at an increasing rate: at 15 characters, the error probability is close to 100%2.
Any scheme longer than seven (a) requires most users to write down, rather than simply remember, the part number for even short-term usage; and (b) increases the likelihood of data-entry errors3.
Part number character set
Another part numbering system design consideration is eye-hand coordination, i.e., "clerical speed and accuracy". Many of our users will prefer using only the characters commonly available on a computer's numeric keypad. This allows single-hand operation, and the layout is easily memorized and generally independent of local language variations (e.g., if you build or service your product in another country, how easy is it to enter your non-numeric characters on their keyboards?).
Part numbers are often used as the filename for importing bills of materials, and exporting design packages. It's therefore excellent practice to avoid using characters that are commonly incompatible with computer filenames, such as * and / symbols. Also avoid underscores that can be obscured within hyperlinks: 27_6542_00.pdf becomes . The safest delimiter character is usually hyphen: 27-6542-00
Restricting our part numbers to numeric characters, with hyphen as an optional delimiter, will provide the fastest data entry for heavy users in purchasing, manufacturing, receiving, service and other places where employees constantly work with a wide variety of part numbers.
There is one other good practice you should consider: never start a part number with a "0" (zero) character. A number with a leading 0 has two unfortunate effects: some people may write 123 for your part 000123, and some computer applications (like Microsoft Excel) discard all leading zeros during import, making a mess of your bills of materials. Although this rule reduces your possible part universe by 10%, it's well worth the cost.
Human factor-constrained part number design
Taking into account human-factors constraints, this very simple rule has been successfully applied in a huge variety of industries, organization sizes and technical environments:
A part numbering system using 7 or fewer numeric characters is the easiest to manage for the majority of our users.
Five to seven numeric characters permit a universe of up to 10 million (without leading zero, 9 million) unique parts, far more than most companies will see over their entire lifespan. Six characters is probably the most common.
If you must have a longer part number, break it into chunks, 3 to 5 characters each, using hyphens.
Authorities for using short, numeric, non-significant part numbers
Many experts strongly recommend the use of short, numeric-only, non-significant item numbers. We have not found any modern authority that recommends using significance in part numbering. Here are some relevant comments from our PLM book list:
Clement, et al.: Manufacturing Data Structures, page 20:
Another important point about item numbers is that they should be as short as possible. Part numbers are keyed, copied and used as verbal identifiers. The shorter the numbers, the more accurate people can be. Obviously, the greater the number of digits in a part number, the greater chance of error. We also recommend that only numeric digits be used.
Garwood: Bills of Materials: Structured for Excellence, page 73 (author's emphasis):
The solution...is to use shorter non-significant part numbers. We have found that part numbers of 5 or 6 digits are the most effective.
Guess: CMII for Business Process Infrastructure, page 81:
Identification numbers are preferably short, not long. The characters that make up the number are preferably numeric, not alpha. Any symbols to be used with the numeric characters are preferably limited to dashes.
Mather: Bills of Materials, page 100:
All tests point to numbers alone as being more easily identifiable with far less chance of error... Tests have also shown that smaller numbers are easier to write and remember accurately. Hence, the ideal part number is all numeric with as few characters as possible.
Monahan: Engineering Documentation Control Practices and Procedures, page 33:
I prefer a non-significant number because there is a longer life and less error... Typically, companies run out of numbers in certain categories of a significant number. Also, a non-significant part number is more cost-effective to use than a significant part number.
Watts: Engineering Documentation Control Handbook, page 49:
The most critical of these issues is that, over time, the significant numbering systems tend to break down. ... As time passes, variations arise which were not foreseen. One digit was set aside where two are now needed. Significant numbers thus tend to lose their significance. They no longer do the classification coding function intended by their inventors.
And, as Francis Criqui of General Motors Corp. has said:
As Director of Engineering Standards at General Motors, I had responsibility for issuing blocks of part numbers for all GM engineering units. I was involved [in] multiple efforts to define a "smart" part number scheme that works. We never found one. .... The easiest and simplest solution is a purely sequential numbering system where the number has no significance other than [as] a tracking ID.
Intelligent part numbering systems: problems and costs
Traditional part numbering systems and document identification schemes originated over 50 years ago. At the time, a basic consideration was that unstructured information was very difficult to find, and it was therefore necessary to overload document identifiers and part numbers with search-related "helper" data.
In the absence of better alternatives, this approach and the resulting practices represented reasonable, though costly and cumbersome, compromises.
Fortunately, times and technology have changed. Let's examine the possibilities...
To encode or not to encode?
There's an almost mesmerizing fascination, even an obsession, with designing the perfect part numbering system. Everyone starts by envisioning how convenient it would be to tell, at a glance, the important characteristics of a part, or the document number that describes the part. Yet, after more than a century of industrial production, the goal remains elusive because each person has a different idea of what part attributes are important — and that person's opinion will change over time! A fluid idea of usefully encoded properties inherently conflicts with the part numbering system's essential role in establishing an unchanging reference to an item.
Good part numbering practice: Our first clue
Let's first examine the two possible extremes: encode all meaningful attributes in the part numbering system, or encode no attributes. It's immediately obvious that a part numbering system that encodes all meaningful attributes of each part would be impossible to manage. The super-significant part number could potentially include: all physical dimensions, materials and properties; environmental limits; contract & regulatory conformances; purchasing and inventory management characteristics; customer/project IDs; licensing or legal constraints; accounting treatment rules; maintenance & servicing intervals; field interchangeability properties; inspection and test procedures; and a host of other attributes. A super-significant part number can easily require 50 or even 100 characters.
Obviously, if we are forced to the extremes, only a non-significant part number is viable.
More practical intelligent numbering schemes benefit fewer users
However, suppose we take a middle path, encoding only certain attributes in the part numbering system, and leaving other attributes to the part description. Which of these attributes shall be favored?
In theory, a significant numbering scheme permits all users to accurately identify all useful characteristics of all parts. In practice, however, some people can learn to decode some useful characteristics of some parts, while making some errors in assignment and interpretation.
Whoever designs the part numbering system will decide the "useful" attributes: engineers will focus on technical properties (size, weight, power, tolerance, material, etc.); purchasing agents may prefer procurement attributes (make/buy, vendor, and commodity codes); production may desire handling codes, tooling and fixture dependencies, shelf-life limits, and plant, line or bin locations; sales and customer support will certainly agree on embedding customer identifiers; and finance will want to encode general ledger accounts, project codes and purchase authority. Quality, marketing, legal, and field service will have their own competing views.
These criteria require us to abandon our human factors constraint of 7 numeric characters.
And the original goal becomes impossible as your company adds employees and suppliers, extends the set of useful characteristics, creates new parts (and part categories), and discovers errors in number assignments.
At the outset, one group designs a somewhat useful part number scheme for themselves (perhaps with only an occasional wrong guess), while everyone else tolerates an error-prone and costly system.
However, there's one tiny bit of consolation: a significant numbering system can't last forever.
Significant numbering schemes are quite fragile
Every intelligent part numbering scheme is temporary: initial design assumptions, practical day-to-day decisions, and the march of technology inevitably make the original categorizations obsolete. Repeated part numbering system tweaks and, finally, redesign are inevitable because:
- Mistakes are often made in encoding the part description into the number, some of which can be quite costly. The engineers who "know the code" use the part number, everyone else reads the part description. If a part's number says "10 amp fuse" while the description is "1.0 amp fuse", a lot of product can get shipped or serviced with the wrong fuse.
- You can never simply ignore an incorrect part assignment. Otherwise, the category number becomes an unreliable indicator of its content. An aluminum casting that's accidentally assigned to the steel castings category requires an engineering change and full dispositioning. And the error becomes more painful if your number has been cast, engraved, etched or printed on the part.
- If there are a series of items (O-rings, screw lengths, resistors) with a common base number, one bad part assignment may block a future assignment, and therefore require that the entire series be renumbered.
- Overlapping or ambiguous categories are functionally equivalent to having no categories, since predictability is lost. For example, is it immediately obvious that an LED is a "display" and not a "diode" ... or an "indicator"? Who's to say, and how do they protect the original definitions over the next few decades?
- Categories become over-used ("Miscellaneous"), ambiguous ("Hardware"), or irrelevant ("Vacuum Tube").
- Designers may be tempted to apply the numbering rules to assign their own fully-significant part numbers as a shortcut, rather than rely on the approved number creation procedure. Highly-predictable placeholder numbers can spontaneously appear during design, yet may never get formally released by engineering or forecasted by production.
- Proprietary information — such as material, value or precision — may be encoded, and is therefore discoverable by third parties.
- Encoded attributes may force unnecessary new part numbers despite on-going interchangeability. For instance, a brass part is revised to use stainless steel; if the base material is reflected in the part number, this revision forces a new number even though the new part is completely interchangeable in the application.
- Merging multiple numbering systems, whether through acquisition of a new product or company, or just when the previous system is deemed obsolete, is often extremely time-consuming to reconcile.
- There's always pressure to make long part numbers even longer: existing categories overflow and new categories are added to differentiate similar items. And most people will become annoyed using many more characters than are really needed.
Avoiding these errors requires additional human and/or automation resources for the initial design, as well as on-going monitoring. The cost of processing engineering changes to re-issue an incorrectly-defined part may easily wipe out any hoped-for advantage of the intelligent part numbering system.
Short, numeric, sequentially-assigned, non-significant part numbers are:
- unambiguous in almost any font style and size
- faster to assign
- never in conflict with the part description or other attributes
- consistent in length & format, and make it easier to identify missing or extra characters
- quicker to enter into computer systems
- less subject to clerical transposition errors
- easier to recall from short-term memory
- more clearly communicated in noisy environments and in written communications
- not artificially constrained to a specific product, project, plant or customer
- always correct when marked directly on parts
- simpler to merge with an acquired or legacy part number system
- difficult to invent "on the fly" outside of normal design release procedures
- unhelpful for deciphering proprietary information
- fully compatible with interchangeable changes to part attributes
Special cases for including significance
Product numbers versus part numbers
It's important to distinguish marketing product (or catalog) numbers from part numbers.
Significant model numbers suggest feature continuity despite non-interchangeable product changes. A BMW 325i car built in 2010 is quite different from the version sold 25 years earlier. It's nonetheless useful for customers, sales and marketing personnel, distribution channel partners and customer purchasing agents to refer to the "same" general product while engineering makes radical product changes over time.
More formally, the product number represents high-level characteristics that can be encoded in the marketing number, and non-interchangeable changes in the engineered product only bump the marketing number's revision.
However, the "significance" risk remains much the same: if there are more than a handful of model numbers, or if a product number is displayed on the product or its packaging, errors in intelligent numbering are easy to make and expensive to correct. An extensive product catalog drives towards non-significant product numbers.
Part category ("commodity code") prefix
Where the business rules are fundamentally coupled to the physical part, or nearly so, they'll sometimes be usefully embedded in the part number. The leading example is when the physical part is handled differently within the inventory or production management process. For example, if you manage a wide variety of part types, say structural steel and electronic components, then these may be managed in different parts of the warehouse. It may be convenient to prefix the part number with a "hint" as to how the part is handled.
So, in a "semi-significant" numbering system, an electrical component may start with "1" and the structural steel may begin with "2", but each part within the category is assigned a sequential, non-significant identifier (e.g., 10001, 10002, ... and 20001, 20002, ...).
Since these part categories represent clear and unchanging attributes, you'll want to keep the numbering groups very large. There's no benefit to creating subgroups that have no distinguishing business rules, or to requiring any thought about whether a part belongs to one category or another.
For instance, in the vast majority of cases an "electronic component" number will be quite sufficient, and there's no need to create separate Resistor, Capacitor and Diode commodity code prefixes unless the physical parts are — and always will be — treated quite differently from one another. These parts types, even with separate and unique custom attributes, can all share the same part number format.
If the business rule (such as part handling) could change over the life of the part, then the business rule shouldn't be reflected in the part number.
A "related item" may be:
A non-interchangeable change to an existing part that requires a new (but not too new) part number.
A member of a part "family" that shares most characteristics with other family members, but varies in only a few properties (resistor value, screw length, O-ring diameter, etc.).
The physical implementation of a related drawing or specification (e.g., document 12345 defines part 12345-001).
As with other types of intelligent numbering, related-part suffixes have problems:
- Part number suffixes are often 2 or 3 added digits (-01, -001) which makes all part numbers longer even if the vast majority of parts never require more than one, or at most a few, variations. For example, HP uses "six-dash-three" (like 655228-001), with the vast majority of parts having only one variation (-001).
- There are occasions when some part families extend beyond the "standard" (e.g., even a -000 suffix can't support high-precision resistors), in which case the range of values must extend into the main number.
- A set of logical incrementing values (e.g., ¼ is -01, ½ is -02, ¾ is -03, 1 is -04, ...) is put into question when unplanned values (e.g., 1.090) must later be included in the series.
- The risk of coding error remains and the entire family may require renumbering if a part gets incorrectly assigned to a value that later cannot be used for the correct part.
- Part and documents have different lifecycles and therefore synchronizing part and document numbers is no longer a useful practice.
- Having a common base number may suggest interchangeability that's not intended. Clearly, there's no question when part 200061 is replaced by 310503. However, if part 20006-2 isn't immediately available, will the servicing technician consider 20006-1 to be "close enough"?
While applying a related-part suffix to a non-significant number is certainly feasible, the practice isn't ideal.
What about part revisions?
Part revisions are not included in the identification of the physical part because, by definition and best practice, all revisions of a single part number are interchangeable.
Consequently, all revisions of a part number can be intermixed in a single inventory location. If this isn't true, then you don't have interchangeable parts, and they require distinct part numbers and inventory bins. For a more detailed discussion, see Do parts have revisions?
Part numbering system best practices for PLM
Product lifecycle management software provides a significant level of automation, and will influence how your part numbering system is defined.
In a computer-based item numbering system, use a short, non-significant number for all document numbers and part numbers. Most attributes that people typically want to encode in their part number system can be either explicitly defined within the PLM system attributes (such as document and part type, make or buy sources, commodity codes, component versus assembly, etc.) or in the item description (part name or document title). Use the product structure to manage the relationships of documents-to-documents (i.e., a "data list"), parts-to-parts ("bill of materials"), and parts-to-documents ("part control data"). Other item identifiers defined within the PLM database can be used to satisfy almost any user's information or query needs.
If you've decided to adopt category prefixes, use a very limited set - preferably fewer than 10 - with categories sufficiently broad that the correct choice is immediately obvious to the most casual user.
If we record all meaningful part attributes within a database, and use the part number simply as an index to that data, then our part numbering system goals can easily be met using only 5 to 7 digits, which can represent 100,000 to 10 million unique part numbers. Relying on a Google-like free-form text search to find part data shortens the part length, speeds data entry, reduces clerical errors, and eliminates the need to periodically redesign an attribute-based part numbering system as usage and technology alter the initial categorization.
Part numbering systems supported by PDXpert PLM software
A good part list tool will automatically assign an appropriate part number, prompt users with how the part should be described, provide information on the part supplier sources, permit user-attached specification and process files, and track the part release history. The best parts list software is part of a complete PLM system that includes bill of materials management, document & file control, and change control.
Like most part numbering software, PDXpert PLM software can follow industry part number best practices by sequentially assigning document and part numbers. If you help your users by separating your part numbers into short groups (e.g., 1234-5678), PDXpert PLM software will ignore the delimiting characters when incrementing: 1234-9999 is followed by 1235-0000.
If your part numbering system uses a "semi-significant" scheme, PDXpert PLM software can assign document and part numbers based on your defined item types. Each part and document template can "subscribe" to a unique or shared item numbering format that permits category prefix (e.g., "HW-"), sequentially-assigned base number, and a fixed suffix (such as "-01"); the next part number from HW-12345-01 will be HW-12346-01. Users can further modify these assigned values as needed.
And, if your company must use an "intelligent" part numbering system that requires human interpretation and assignment, PDXpert PLM software will accept manually-entered document and part numbers for all of your items. The software easily accepts mixed schemes, such as importing legacy significant numbers while auto-assigning shorter non-significant numbers.
In addition to accommodating your specific part numbering system, PDXpert PLM software supports all organizational stakeholders:
- Duplicate part numbers are automatically prevented based on rules that can optionally include part or document class or subclassed type.
- When creating a new part record, users are prompted for consistent item descriptions using "text templates" based on the part type.
- A specifying designer can provide unit costs for part list roll-ups.
- Each part can have its own supplier source parts list, which includes the ability to rank preferred usage.
- Users can add a virtually unlimited number of file attachments that are either "locked down" on item release, or can be modified throughout the entire part life cycle.
- Even non-technical users can easily find parts by relying on familiar "Google-like" free-form text searches that not only return the most relevant parts, but also can return similar items that may be substituted, with the resulting part list ranked by relevance.
- Complete part histories are tracked by releasing and canceling parts revision records using change forms.
- G. A. Miller, Psychological Review (1956) "The magical number seven, plus or minus two: some limits on our capacity for processing information."
- Dave Garwood, Bills of Materials: Structured for Excellence (1995), page 70.
- Ehap H. Sabri, et al: Purchase Order Management Best Practices (2006), page 114: "Typical error rates for manual data entry are about 1 error for every 300 keystrokes."
Contact us if you'd like to discuss how the general concepts summarized in this note may be applied to your situation. We'd be happy to address other PLM software good practices - ask us!
PLM Good Practices