Disclaimer: This topic is not interesting, not even to me. It’s an engineering process rant.
The problem centers around the fact that first principles, while a good place to start, are not good at resolving problems to which first principles have been applied and repeatedly failed.
In a modern approach to design, you have a team of mechanical engineers working to design parts. You’ve also got electrical engineers incorporating electronics, industrial designers managing the form, marketing guys designing the packaging, manufacturing engineers making sure it can all be produced and assembled in a cost effective manner, and managers to try and keep things on schedule.
There are also software engineers, but this argument doesn’t entirely apply to them, for I’ll leave them out of the example for the moment.
Very early on in a process like this, it becomes evident that every part that is designed needs to have a unique identifier. Otherwise, how do you know what all you need to order and build?
The problem arises when people try to utilize that part number to describe the part itself. Lets go through the exercise of developing a smart part numbering system.
1st design consideration: At least in the US. studies have shown that when the number of digits surpass 10, the probability of a transcription error is greater than 50%. We have an average brain buffer of 10 digits. So, any part number methodology you assign, you’ll want to keep to no more than 10 numbers.
2nd design consideration: The methodology must be simple, and relatively intuitive. When new employees are brought on board, you don’t want to spend hours/days/weeks teaching them how your proprietary method works. For one that’s time that they could be doing work that will make you money. For another, given the transient nature of so many jobs, contractors and new employees are not always enthusiastic to learn a skill that will not make them more marketable to another company. If you have to write a manual on your part numbering system, you’re on the wrong track.
“Now wait a second!” you might ask. “Why not use one of the established part numbering systems already in use for a given industry?” Great idea. Go find one. I’ll wait….
Actually, I won’t. I’ve already written this by the time you’re reading it, and this isn’t a rant to go into the ramifications of reading something that’s not yet been written. What is telling is that companies with massive inventories that have to index millions of unique products have opted to use a system that does not rely in intelligent part numbering systems. Examples: Amazon.com, and McMaster-Carr. Alternately, Stockdrive parts/Sterling Instruments has what appears to be an intelligent numbering system. The numbers are twice as long, and good luck figuring out their methodology. To say nothing of the fact that the part often has multiple numbers, none of which are ever in stock (no I’m not bitter (well, okay, maybe a little bit)).
3rd design consideration: Think about all of the things different departments that I listed previously. Here’s a short list.
If you’re a mechanical engineer: weight, color, material, dimensions, general tolerances, is it an assembly?, is it a part?, is it an assembly but you buy it as a single module that you can treat as a part?, what project does it fall under?, etc.
If you’re an electrical engineer: what is the resistance?, what is the capacitance?, is it a microprocessor?, how much I/O does it have?, how does it mount to a PCB?, what is it’s thermal profile?
If you’re a manufacturing engineer: fabrication method, number of post processes, is it heat treated?, is it toxic to handle?, what is the shelf life?, is it reactive?
If you’re an industrial designer: what is the acceptance of marring, I can probably think of others, but I don’t feel like it.
Okay, lets say you’ve managed to accomplish all of the above. The system is simple, it’s running, people can use it, and learn it quickly. You’ve got 4 screws sized 1, 2, 3, and 4.
But the inevitable happens, someone comes out with a size 1.5, and your designers need to use it. 1’s not strong enough, 2 is too expensive. Prudence demands that you append a digit and create sizes 1.0, 1.5, 2.0, 3.0 and 4.0. Hopefully you designed that system to have just 9 numbers this kind of contingency. Sadly, that means you have update every number of every part, which is time consuming and expensive. You also have to pray that it won’t happen again, which it will.
Alternately, you create an exception and make size 1.5 just call it 5. You’ve broken the scheme, though, which means you have to update the documentation explaining how the rules work, with an appendix of exceptions to the rule. This can grow pretty fast, which violates the design goal of simple to learn and intuitive.
A third option, of course, is to not allow the engineers to use a 1.5 size screw because the costs associated with updating the part numbering scheme make it prohibitively expensive. Congratulations! Your numbering scheme has just reduced the profitability of your product. I’m sure that will look good on the performance review.
Smart part numbering systems are traps. They designed like wikis, where people start with a simple setup that is allowed to grow organically over time. It’s another database that’s easy to add data in. You can add more rules and complexity to maintain the original methodology, but the simplicity is lost within just a few generations of the numbering scheme, which can happen in less than a year for fast growing companies. I say it’s a trap because people believe that it’s easy to get the information back out. There’s a method to the scheme, so people can figure out the part number based upon their needs, right?
The more you add intelligence to a part numbering scheme, the more you reduce its flexibility, in terms of part reuse and cost out opportunities. The original purpose of the part number is to give each part a unique identifier. We’ve not even covered the possibility of multiple parts accidentally deriving the same number under an intelligent scheme.
This is a problem no one has ever been able to solve. First principles have been abused repeatedly in this respect. Moreover, the market has moved away from intelligent numbering systems. Consider Google and the internet. Websites are assigned a unique IP address by which to reach them. There is no intelligence related to the IP address and the content within the site. Once you have an IP address, you can put anything there you want. Rather than assigning an intelligent numbering scheme to IP addresses, it’s better to allow for random assignments, and then services like Google will trawl the content and index them for fast and easy searching.
So, when thinking about smart part numbering schemes, remember the words of Admiral Ackbar
No comments:
Post a Comment