Jos Voskuil's Weblog

This time a more theoretical post about classification in PLM. A topic that always has been around for more than a decade. Recently classification also came up in some discussions both with customers and on discussion groups on the Internet. So what I will try to do in this post, is to explain the goals of classification, the ways classification is implemented and finally how I see classification will evolve. As always feel free to comment or extend my post.

The goals

Classification is a generic understanding, so to start I want to narrow classification to item classification. Companies might use classification in various ways, for products, for knowledge, for supported functions and more. The most discussed topic in the context of PLM however is item classification. The idea of item classification is twofold: understand what you already have (designed) and promote reuse. Historically the item definition has been stored in the ERP system and reuse was mainly based on recognizing parts of the description. Sometimes the ERP system also supported a kind of classification id to group certain parts – like fasteners, frames, base materials etc.

With the introduction of electronic parts this rough classification as defined in ERP became insufficient as already the description and classification id were not enough to really understand if an item could be reused. During that same period of time more and more companies where merging or acquiring other companies and they want to understand and benefit from items already used in one of the companies.

So this brings back the challenge for the two goals mentioned:

• How can I make sure my engineering reuse existing items in future products ?
• How can i consolidate and understand items i have used in my company ?

Item reuse

In order to promote item reuse companies have used various classification systems in engineering to promote reuse and standardization within the company. Design catalogs with standard purchase parts, extended with company standard parts were implemented to limit the variety of choices for a designer. Companies & Products  like Trace Parts, SolidWorks Toolbox, Inventor Content Center address this need.

Additional mostly in the German speaking countries a classification standard exists, called sachmerkmahl leiste (sorry only in German) or often referred to in the context of the DIN 4000 standard. This is also a standard classification standard, less CAD design centric. Interesting to analyze why this standard does not exist in other countries.

One of the reasons might be that classifying all your engineering data takes a lot of time – specially when you haven’t done it from the start. I worked with some companies where more than a man-year was spent on classifying information. This work had to be done by someone with engineering knowledge, so you can imagine the investment for classification, beside the software was huge. Main question is, what will be the (expected) Return On Investment ?

In this area I think that a cultural difference plays a role here. Some countries invest more in their working methodology and processes than others where the focus might be only on the single result. From my global experience to be fair, i have not seen and heard the real benefits of this type of classification for reuse. I am looking forward for statements from companies that have measurable result here. Like many IT projects we have the emotional feeling that this approach should bring benefits

Item Consolidation

In the mid nineties when companies started to merge, PDM became PLM, CRM became important, also another trend became visible. The need for item classification systems, more on the inventory side, for companies to understand which items they were using around their (merged) enterprises. One of the first companies that time was Aspect Development Inc, later in 2000 merged with I2. Customer case studies learned us that in some of these enterprises a single item could exist with 100 different ID’s, all described or classified in various departments a little different, so hard to reuse. Only by classifying items within an enterprise based on their specific characteristics, people start to recognize identical items. Also in smaller mid-market companies I have seen situations where items have been named or identified just a little bit different, although they were the same.

Here benefits of item consolidation can be easier justified. I assume most companies can estimate what is the total cost of handling an item through its lifecycle and what are the purchase benefits by consolidating for example 10 different named into a single item to be purchased in a much bigger quota.

The benefits really come when you control your inventory and from this base feed the engineering department with an optimized selection of validated items for reuse. And this is to my opinion the most important goal of classification

How to implement classification ?

As described above classification is needed to promote reuse of engineering knowledge and to standardize on inventory (purchased items).

To address the first need I believe PLM offers various ways to support a classification. Some might believe DIN 4000 is a useful standard. From my experiences with companies it appears that it is important to bring rational to what you classify. Where is the ROI. Classification brings a lot of constraints and overhead to the engineering department – all parameters needs to be mandatory managed for each part, otherwise your classification looses its value. Probably you will realize that classifying metal strips does not bring the reuse value as the overhead for maintaining the classification is higher then the cost of producing a new strip. So I am not so convinced about classification for this need.

For the second need – inventory optimization – here i believe the classification brings a measurable ROI, specially when the company uses  a New Item Approval process  or Standardization Process, where every new item will be reviewed (and classified) to guarantee its unique need. Of course it depends very much on the type of industry and main business process if this approach brings value. Listed in a more relevance order: Engineering To Order / Configure To Order / Design For Manufacturing

Folksonomy versus Taxsonomy

A new trend for classification is the way search engines work on massive unstructured data. No one tries to classify all the web pages that exist (although there might be a standard for that). It is easier to perform a context search and specially with new web development you see that tagging information becomes more and more important for retrieval. For example I tagged this article with PLM, ERP, Classification, Item Reuse and Item Consolidation. These tags will be used by search engines and I do not have to worry on which level and where Item Reuse is stored. As a creator of this text part I tag my information for reuse.

This is called Folksonomy, this in contrary to Taxsonomy, the classical method for ordering information. See for more background the related wiki hyperlinks.

Conclusion

Implementing Folksonomy in a PLM environment depends on the type of PLM system you are using (in case you use a PLM system). It requires a way to tag information in an user-friendly way and to retrieve information by tags in an easy way – the ease of use of a search engine. In case it is too futuristic this approach, evaluate your engineering classification needs based on your expected ROI and goals, keeping in mind in the classical way of classification will evolve.

Do you have examples of classification with a proven ROI for engineering, let me know