Connecting manufacturing professionals to drive continuous improvement!
Intelligent vs. Non-Intelligent P/N' - what's the better choice?
When migrating from legacy ERP to a modern system such as Oracle or SAP, does it make sense to implement intelligent part numbers or to use an automated system with random part numbers?
Replies to This Discussion
-
Permalink Reply by Frank Michaels on -
This is a debate which has gone on for years. My observations:
1. IT and database folks always push for random arguing that intelligent systems always breakdown when there is enough change in products or components. Their bias is that they are all focused on system efficiency and a random system is much easier to implement and maintain.
2. Planners who use the part numbers all day long are often prefer to use random numbers since they memorize the part numbers they use anyway. Simple random numbers may be easier for them to work with - and data entry errors in planning can have huge implications.
3. Engineers and product oriented folks who don't see the part numbers often enough to memorize them often like some degree of intelligence, arguing it makes analysis easier, and you can spot errors in data more quickly. It's important to listen to them, because they, along with the planners, are real customers of the system.
I've lived with both random and semi-intelligent systems. My experience is that if you really understand your components, sales items and product structures, and can accurate project the changes in the roadmap out many years some degree of intelligence can be useful - distinguishing between catagories of components, BOM levels, etc. But, if you set something up with any degree of intelligence you really have to know the products and technology and anticipate the changes in your scheme, otherwise go with random. It will be a mess if it breaks down.
It's true that modern ERP systems allow you to put enough attributes against a part number in the item master to provide the intelligence you need in a random part number scheme, but keep in mind that you have to apply the same naming convention controls in the item master that you would in an intelligent part number scheme to keep this useful.
-
Permalink Reply by ASRA ISLAM on -
Paul I think it is a choice that can be made on future utlization of data over a period of time. If the company is interested in a robust analysis of material (eg category spend, inventory etc.) then an intellegent p/n may be the way to go. It allows to follow the logic and sequence and makes category management easy. If that is not a concern an automated numbering will be sufficient to meet most demands.
-
Permalink Reply by Greg Galaski on -
Its better to stick with what ever part number was developed with the ERP system of that time. Otherwise creating new part numbers or part numbering schemes will cause lots of problems for legacy bills of material and component inventory notwithstanding further problems involving traceability and accountability for legacy products.
-
Permalink Reply by David L. Hasse on -
As the current CAD guy at a company with legacy documents that originated in England, Germany, Belgium and various acquired locations around the US and considering 25 years of contract engineering exposure I can't recommend "intelligent" part numbers. Assigning numbers randomly doesn't really sound like an improvement. Sequentially without trying to "code" additional meaning into a particular location within the set seems to work best in the long run. Adding dash numbers to define a tab drawing with options generally works with purchasing's cooperation.
-
Permalink Reply by Steve Grube on -
With experience in both system, I like the straight no-information-bearing random next-number-in-order system. Our small manufacturing company (<100 people) uses Syspro (ERP) and it assigns the next 6-digit number and this is just fine. I'm sure there are those that will go 13 rounds over the efficacy of one system or another, but there are bigger issues for doc control systems to wrestle over: timeliness, completeness, accuracy AND compliance.
-
Permalink Reply by Jos Voskuil on -
For me working in a PLM environment, intelligent part numbers are a hell to maintain and a threat for the future. Let me explain.
Intelligent part numbers were invented during times there was no quick access to an IT system (PDM or ERP) - people were working document based and on the shopfloor people could immediatly see (after building the experience) what kind of part they were looking for.
So nothing is wrong with intelligent part numbers as long as nothing changes. And this is the main problem. Companies need flexibility.
Why ?
1. Intelligent part numbers work for experienced people - this means if a company relies on this, finding new people to do the job takes more time (reduces flexibility)
2. Intelligent part numbers become a problem when a company merges with another company. Who's intelligence is going to be dominant ? Or do we need two intelligences. Again it reduces flexibiltiy
3. Intelligent part numbers can create a fixed framework that does not allow companies to work in a different way. It is a legacy they will be carrying on for the rest of their existance.
I implemented PLM in a Swiss company that used intelligent part numbers for more then 100 years. When the system was defined there was not so much complexity. But now when different types of parts are used (electrical / embedded software) this definition fails - they are reluctant to drop it (due to the long history) but it has become a burden to maintain.
My conclusion: As IT has brought us to the level that we can spread information in an electronical way, there is no need for intelligent part numbers anymore. Focus on other attributes to describe better and assure the right understanding (and reuse)
I thouched this topic briefly in my blog:
-
Permalink Reply by Joe on -
as an engineer I would say; if it aint broke don't fix it! Keep the part # scheme the same unless your users are demanding a change.
-
Permalink Reply by Jonathan Scott on -
Much like a previous post, I approach this problem from the PLM and Engineering angle. I believe there are cases to be made for specific compromises/exceptions, but as a rule I believe that part numbers should be non-intelligent. The argument I will put forward (I support the arguments made by several others, including Jos Voskuil and Greg Galaski) is based on the frequency of change in today's business environment.
Back when mergers and acquisitions took years to accomplish, partnerships were forged over years (not months), and employees stayed with a single company for their entire careers, the cost of "deployment" of an intelligent numbering system could be small when compared with the benefits. As others have noted, there is a learning curve to intelligent numbering systems. For the average user operating in medium-sized business with average product complexity, becoming an expert at the numbering system can take 1 - 12 months. Consider that relative to 3-year average employment durations, the proliferating use of contract employees, and trends toward outsourcing manufacturing and design, and it becomes clear that you may spend more time training people on your numbering systems than you do reaping the benefits of their in-brain search system (that's what intelligent numbering amounts to). And of course M&A activity only compounds these issues (reference David Hasse comments). In a high-growth industry, a company could find itself re-inventing / merging numbering systems once or twice a year.
On the PLM front, I'd like to take one of Jos Voskuil's arguments a bit further. Having been in this industry for a while, I have observed that engineers (I count myself guilty here) really want to name their CAD files and other data with the (intelligent) part number. This is fantastic as an ad-hoc search system (absent PDM), but breaks down quickly in the face of large teams, corporate re-use initiatives, and enterprise search requirements. Pigeon-holing a part as a "Boeing 787 Left Far Control Surface 8-pin Power Connector" will never do in today's re-use-it, order-it-in-bulk, drive-to-commodity situations. All of you 3D parametric CAD users out there will recognize that giving CAD and other design files an intelligent part number file name is essentially the same folly. Trying to rename a component once it is "in the system" is as fruitless and un-Lean as trying to reassign a file name in CAD once the file is used anywhere.
So what's my answer? I believe non-intelligent part numbering coupled with rules-driven part naming is a great hybrid. The part number can then be used as the component's "unique key" in every system in participates in without worry for changing the name or intent of the component later. The part name should be driven by rules to avoid searching and reuse problems later. Further classification or tagging can help the situation even more.
I think the key here is "de-coupling" the information and tags from the unique identifier - ask the IT guys about this, they can give you an earful about how they've found this to be helpful over the years :) That being said, once you "de-couple", you need systems to maintain the information in a meaningful way, so beware that you are trading one burden (inflexibility of intelligent numbering systems) for another (administration of management systems like PLM). If I'm correct though, the first burden is much more expensive than the second.
-
Permalink Reply by Menk Slot on -
The question is how far do you want to go with intelligent Numbers. You have to make a difference between Identification (random) and Classification. A lot of companies start intelligent numbers until the number range is not big enough.
-
Permalink Reply by Ed Danzer on -
Having worked with several types of part numbering systems developed by many different manufactures and distributers, the shorter the numbering system the better. This mean you can have a semi significant (semi intelligent) which is what we chose in 1997 after trying several methods, including random, completely significant, and using a specific distributers part numbers.
Even though we are a small manufacture we inventory about 10,000 items and have over 15,000 3D solid models and drawings. The advantage of semi significant is trying to fine items seldom used. The disadvantage is tracking the assigned numbers.
The companies we work with that have random numbers have difficulty finding seldom used part numbers if you don’t remember the exact item description.
Using a dash in the part number just adds another character to type and most production people make a key stroke error every 10 key strokes. When customers order by your part number the less characters the better as there will be fewer mistakes by the person ordering and the people filling the order.
I would recommend using your old numbering system for existing items, then experiment with how easy it is to document both random and semi significant so items can be found in the new ERP system.
-
Permalink Reply by Menk Slot on
-
As you can see in this discussion, at every PLM or ERP implementation the numbering system is always a discussion and can delay the complete project. If you already have a numbering system that is working good, why changing this.
When companies start with a PLM or ERP System they can consider to use a number generator and use Standard descriptions. If you combine this with a classification system it will be much easer to find back your parts. It is also impossible to describe a product in a part number. Some parts will need more characters in a number then others. It is also depending on the department how to find back a part. Manufacturing will have other demands then Engineering for example.
What most companies try to do is to put a classification in there numbering system
This can all be solved by implementing a classification system. At this moment there are already systems who can generate the classification Group and Parameters from the 3D CAD Model.
-
Permalink Reply by Jos Voskuil on
-
It is interesting to read Ed and Menk's responses and perhapse we can combine them.
I am aware that we have the tendency to like intelligent part numbers - this is how our brain works - we like ordering/sorting. The current issue is that we like to do this on the part number.
If you look in most of the it-systems (PLM,ERP even office) many have already internally a GUID (global user ID) or UUID (universal unique ID) and as they are too long to memorize we do not even display them.
I agree with Menk that classification is probably the best way to structure your parts and the 'old intelligent' part number, somehow is a classification code. So is the problem solved ?
The main pain I believe is that most of the current manufacturing companies do not work stand-alone in this world anymore. They are merging / aquired or aquiring other companies and then they need to communicate around shared parts.
Classification is a good way to identify common used parts as when the characteristics are the same, probably the parts are the same (or an extra characteristic is missing).
In implementations where I assist I usually try to implement:
- non intelligent part numbers
- standardized part decscription(s)
- a classicication code
But at the end - the main pains are legacy data - as there is not a global unique classification system (yet) - we will need to struggle to find the best compromise
© 2010 Created by Paul Patterson.
