The end of part numbers

A new year is a time for renewal, opportunity, and new beginnings. For engineers, however, it's a chance to argue about part numbers some more. I love the smell of part numbers in the morning. Whether your allegiance lies with the Generic Numbering Coalition (GNC) or the Confederacy of Intelligent Numbers (CIN), valid arguments worth defending exist on both sides. We've covered both factions and spaces in-between. The pursuit of part number perfection, however, may lie in mutually assured destruction. The part number is a lie. For one day, perhaps sooner than you think, part numbers will be no more.


Mixed signals

One clear sign of the partocolypse is that outright victory eludes each of our stalwart, diametrically-opposed numbering religions. The prevalence of engineering data management tools such as PDM and PLM secured wins in the transition to generic numbering, capitalizing on the implementation of such tools as a perfectly justifiable breakpoint to expunge the old ways. After all, it makes perfect sense that capabilities like classification should trump any humanly contrived numbering schema, destined at one time or another to become wholly obsolete. Yet intelligent numbers thrive; they remain both abundant and widespread. What gives?

The reality of part numbers

  • Old numbers die hard: For one, part number legacies are indeed numerous. Old part numbers aren't going to simply dissolve into the ether. And yet many more new parts contain some semblance of intelligent part numbers, if only because users of those part numbers have sometimes fought and won against their database-minded IT oppressors.
  • Intelligence and metadata aren't mutually exclusive: Second, from strictly a usability standpoint, generic numbers offer nothing compelling. The metadata and/or classification used to parse and search generically numbered parts can be implemented on any part, regardless of its numbering structure. A common complaint leveraged on intelligent numbers is you don't know what you're doing unless you're in on the coding scheme. As if you have no other choice but to memorize or interpret the number manually. In a database driven environment built on intelligent numbers, you don't have to understand the number. In fact, given the metadata, the system could even build the number for you. Knowing the schema provides a shortcut, surely, but otherwise you search for it just like you would have to always do with a generic number. So from a usability perspective there is no compelling case here.
  • No system exists in isolation: Then there are the endless company mergers, acquisitions, spin-offs, buyouts, reorganizations, etc., where two or more numbering systems collide. The end result is a common cross-pollination of part numbering schemes ranging from the meticulously contrived to the totally random and everything in between. So universal adoption of pure, organic, grass-fed, free-range, vegan generic numbering is – at best – inconsistent. The realities of part numbers and their dependencies on historical data are far too complicated.

But the truth has eluded us. For the perfect part number scheme is not a part number scheme at all. That's not existential nonsense, but rather the reality of computer science taken through the lens of usability.

UID: not for you, not for me

Part numbers are meant to be unique. Ideally they are system generated. Consequently they're used as an indexing mechanism by the database technology in engineering tools. That very fact makes part numbers Unique Identifiers (UIDs). And that's bad, because UIDs are for machines, not people. All the technology you use employs such UID constructs, but you never are expected to interact with them.

People should not deal with indexes. But we deal with part numbers. That's why we're not getting anywhere - we're taking a touch point that's been hijacked as an index, and turning it into a proper index that makes a lousy touch point. You can't win this way. Which is why the part number debate goes in endless circles.

If a part number is to truly become an index they might as well be universally unique identifiers (UUIDs). UUIDs take things a step further by creating a standard for "practically unique" global identifiers, technically suitable for interchange across disparate systems and databases. Incidentally, many engineering database systems don't have part number fields that can support the 32 hexadecimal characters needed for a 128bit UUID (I'm looking at you, SAP).

By the way, just to blow your mind, UUID's have structure and, thus, can in some ways be considered an intelligent part number schema. Ah, the irony is strong in this one.

A numberless future

Does Tony Stark ask Jarvis for specific part numbers? How many phone numbers do you remember? Do you know your IP address without looking?

A more important question: do you really have to remember anymore?

An even more important question: do you care to remember?

There's a lesson here that applies to the future of part numbers. Systems of the future interacting across a federation of technology will be exchanging your part number as an index, but you will be just dealing with the meta data and the pretty pictures. The index will be as long as it needs to be truly unique and will carried forth as necessary in the background on every object for configuration control, but you'll just worry about the part, the design. The stuff you actually care about.

The problem is that type of future is still a few years out. But in the meantime, moving to generic numbers in today's systems doesn't bring you any closer to that reality. So expect a few more years of this whole brouhaha, before the age of part numbers comes crashing down. Until then, feel free to declare your allegiance in the comments below and we'll argue some more.

A Part Number Anthology part-number-anthology-small

Part numbering. For most engineers, this two-word phrase is all it takes to conjure up especially strong feelings about what it means to be “right”, and what it means to be very, very “wrong.” We've assembled a handful of our part number greatest hits in this eBook anthology.


  • Carlsbad CAD

    the lack of comments displays the boring subject at hand, haha. part numbers convey information, and help manage part inventories, the name may change but the info is still needed

  • Carlsbad CAD

    also part numbers along with batch info, revision status and PLM are required by the FDA & FAA in the USA for safety reasons

  • I actually have quite a few phone numbers memorized. Why? Because it is easier for me to type in 7 or 10 digits to get in contact with someone than it is to scroll through the long list of people in my address book. And maybe, if I could avoid all the typos from fat-thumbing that tiny little screen, filter my contacts list to get to all the W’s, but that still leaves a lot of Wertels in my list that I have to scroll through.

    Is it important to memorize engineering equations anymore? I can just Google them if I ever forget, because apparently looking them up in old textbooks (and thereby getting examples of their proper application as well) is too hard for most people. Or I could just visit one of the many engineering-specific websites that I have bookmarked and use their built-in calculators with little comprehension of the assumptions and possible errors of the tool.

    The same goes for part numbers. I have always found what I’m looking for very quickly when I know what I’m looking for and been able to properly utilize in my application. I have not been a project engineer for several years now, and I can still rattle off the part numbers of the main systems of the product line I was assigned. Most people can memorize a string of 7 characters. Memorizing an automated 32-character hexadecimal index is not feasible, and therefore unlikely to be a fast and responsive way to find exactly what I’m looking for. Metadata searches are only as good as the text strings that HUMANS entered into the system. How many “brakcets” do you have in your PLM system? Brakcets are teh best.

  • Robert

    Come on guys! Every couple of months (it seems) you have this position where part numbers will go away, but because they’re something that can be sold and cataloged as something with specific criteria (number of threads, size, length, speed, pressure, etc.), it’s unlikely that part numbers will ever, ever go away.

    We’re in an era where we make things with more varieties now than ever, and with the advent of 3D printing, we’re going to see more and more customizable features, which means (you guessed it) more classifications, which means more part numbers.

    The only way that part numbers are going to see LESS use is if we start using more of the same parts more frequently (more homogeneity), and I seriously doubt that will happen.

    • Robert

      Like Mr. CAD below, I think you misunderstand. I’m not advocating that unique parts numbers cease to exist, only that they become irrelevant to your interaction. We’re making humans do what is better done by machines, by creating, remembering, ordering, configuring and manually entering part numbers from one system to another.

      We do track those numbers now because we have to, but we only want to make sure we’re working on the right thing all the time. We’re doing too much of the configuration control.

      While you may think you’d rather scribble that part number down on a post-it, so you can find it again, you’re thinking only in the context of the present. The same way we think about still driving our cars and dialing phone numbers. We won’t be doing either for long, and it will happen sooner than you think. Everything will be determined by context. Who you are, what you’re working on, when, on what project, etc. Tony Stark just starts asking questions and working, it’s Jarvis’ job to keep up and configuration manage all his ideas. It might seem like a pipe-dream when you look at the crude reality today, but the technology to enable this will come fast.

      Instead you’ll have familiar identification for the stuff you’re working on that’s independent of whatever the systems use to version, regulate, distribute, purchase, and control. That will be the point of the index, the UID, what the P/N is meant to be in the end.

      As you mention, the sum total of part numbers will only go up, perhaps exponentially based on the highly customized and configurable products that are increasingly demanded by discerning consumers. At that point P/N uniqueness really becomes critical. Duplicate part numbers across different vendors are already a real problem for larger companies with wide supplier bases. The only way to ensure that is to move P/N’s to a proper comp-sci worthy UUID, at which point it will be unsuitable for human interaction, unless you buy your Post-it’s in a foot-long variety.

      • Robert

        It’s not the technology that I think makes this unreasonable. I think it’s the bureaucracy that will prevent it from happening.

        Tony Stark owned his own company, with his own private access to assets as he sees fit. It would be one thing if everyone were a wealthy billionaire with seemingly unending pool of resources that would make using part numbers unmanageable.

        At the end of the day, accounting matters. Purchases orders aren’t going anywhere.

  • Sean

    This is only being looked at in a vacuum. Yes, both the Ironman and Batman probably do not really need part numbers. And if you are an imaginary bazillionaire that likes to make high tech gadgets for your own personal amusement and fighting crime then you don’t either. But most people are developing products that are going to be produced and sold. And almost certainly assembled and/or repaired. As you point out, a UUID is not reasonable. You probably won’t be able to legibly mark it on a small part and good luck getting a customer or technician to read that back to you over the phone when trying to make repairs. Could you imagine trying to layout a BOM/exploded parts diagram with that many characters? Yikes. Unless all your outside suppliers are keyed in on your system then you will have difficulty translating things over to them. If they are, you would be stuck with them. Worse. Using a name schema doesn’t work if you operate internationally as the the language doesn’t translate correctly. And both your customers and probably repair centers don’t have access to your PLM/PDM system nor would you want to train them or give them access. Most small companies can’t afford that either. Try cutting P.O.’s through a purchasing department with something other than just a number. Good luck with that. So as most small companies do, you start with a spreadsheet of rolling parts numbers that get assigned to parts as you go. And that system is more or less perpetuated as the company grows. You will need to support legacy product in the field so changing the scheme later is a real pain as the old one needs to be perpetuated.

    In my opinion, just use numerical part numbers that are as short as possible and always unique. Don’t get cute and try to add a prefix for product classes as you will just introduce redundant numbers that will get confused. And confusion costs time and money. KISS.

  • Joe McGrath

    Maybe engineers’ don’t need part numbers. Everyone else does. How do you purchase a part without a part number? outsource? quote? produce? package? ship? invoice?

    Do you have to send them a picture? Make everyone install proprietary software you need to maintain to do something simple? Sounds like your trying to fix a problem that doesn’t exist.

    Manufacturing DEMANDS part numbers.

    I can tell someone go make 500 PCS of PZZ1584 right now and they can.

    I have several hundred if not thousand memorized. I’m sure many people in manufacturing are the same way.

    The engineer can click a link in their assembly just fine, if that’s what your after. And why not make an algorithm to generate the P/N for you. Just try to keep it short. Three letters and four numbers seems to work well.

    FWIW I do know my IP address (public and NAT) at home and work. I also manually enter common phone numbers until their name pops up.

  • Tim Johnson

    I would say that we’re already there. The technology already exists. When you go to Amazon you don’t search for ML3K2LL/A. Instead you search for something meaningful to the broadest range of users, and begin to narrow down your search to the specific criteria that your current task requires, e.g. Apple, iPad Pro, 128GB, Wi-Fi + Cellular, Space Gray. You are not concerned with the actual Model Number/Part Number unless circumstances dictate, i.e. warranty or service issue. When those rare situations do pop up when the Part Number is needed, there it is, easy to find – just not easy to understand or memorize. In the Engineering world it should be no different. We should be able to search for a Copper [material] Block [shape – prismatic, round, hex, etc.] 20mm x 50mm x 100mm [size-thickness, width, length] with two M5 tapped holes [quantity and type of feature(s)] and any other defining characteristics that are inherent to the design of part that will enable the broadest range of users to find the part based on the criteria that is most meaningful to their specific interaction with the part in question. Remember that engineering is actually a function of laziness… One of the first thing computer programmers developed was Copy+Paste. :)

    • Sean

      “When you go to Amazon you don’t search for ML3K2LL/A. Instead you search for something meaningful to the broadest range of users, and begin to narrow down your search to the specific criteria that your current task requires, e.g. Apple, iPad Pro, 128GB, Wi-Fi + Cellular, Space Gray.”

      So you think every company should be able to build a custom, accessible, computational infrastructure comparable to Amazon? Do you think that is even remotely reasonable? That idea might be fine when shopping for things that you are generally interested in. Parts numbers get you to a specific item that fits in a specific product, probably at a specific rev level. They do so quickly and efficiently. Nothing about engineering was or is derived from “laziness”. The essence of engineering is to make things more reliable and repeatable so we become more efficient as both as individuals and as a society. In general, it’s called being smart. And that is why computer programmers made Copy+Paste. Because manually rewriting things is stupid.

      • Tim Johnson

        Who ever said engineers don’t have a sense of humor? :P

      • I actually do type in the part number in Amazon.

        When I’m working on a repair, like the valve assembly for my water/ice unit in my fridge, I took apart the old one, typed in the part number in Amazon, and got it 2 days later. Exact fit.

        If I were to try to find it by keyword, I would have come across hundreds of non-conforming results to sift through and I still would not have been certain the part I received was going to work until I installed it.

        I prefer to prevent mistakes before they reach production.