Parts Classification in Windchill (PartsLink)

In short: parts classification in Windchill (PartsLink, part of PDMLink) sorts every part into a tree of categories with defined attributes, so an engineer can search “bearing, 20 mm bore, stainless” and find the part that already exists instead of creating a new one. The software is the easy part. The taxonomy design, the governance over who creates parts, and the cleanup of legacy data decide whether it works.

What parts classification is

Most companies find parts by part number or description. Both fail at scale. Numbers carry no meaning, and descriptions drift: one engineer writes “transmitter”, another “xmtr”, a third “TRANSMITTER ASSY”. Search misses half of them, so people create a new part, and the duplicate count grows every year.

Classification fixes that by giving every part a place in a structured tree and a set of attributes that describe it the same way every time. In Windchill this is PartsLink, part of PDMLink.

Key fact	Detail
Windchill capability	PartsLink (classification), part of PDMLink
What it organizes	Parts, and optionally documents, into classification trees
What users get	Browse by category, parametric search on attributes, “find similar”
What it enables	Parts reuse, fewer duplicates, standard part names
Who owns it	A classification administrator plus component engineers

Classification and reuse are two different things

Classification is the structure: categories, attributes, naming rules. Reuse is the outcome: an engineer finds a suitable part that already exists and uses it. You can have a perfect taxonomy and still get no reuse if nobody is required to search before creating a part. That’s a governance question, covered below.

How PartsLink works

The tree. A namespace holds one root node, branch nodes under it, and leaf nodes at the ends. Parts are normally classified at leaf nodes, for example Fasteners > Screws > Machine screw.
Attributes and inheritance. Each node carries attributes (thread size, length, material). Children inherit their parents’ attributes, so common ones live high in the tree and specific ones low.
Valid values. Attributes can be limited to a fixed list or range, which keeps data consistent and makes search work.
Classifying a part. A classification attribute on the part type links the part to a node. Engineers pick the node while creating the part, then fill in that node’s attributes. A part can be classified on more than one node if it genuinely belongs in two.
Search. Users can browse the tree (with images on each node), search by attribute values and ranges, narrow results with filters, and use “find similar” from an existing part.
Auto-naming. A naming rule per node can build the part name from the node and its attribute values, so names come out standard, and that standard name can flow to ERP.

Designing a taxonomy people can actually use

These are common guidelines from Windchill implementations, including PTC’s own best-practice material:

Three to seven children per parent. Fewer makes the tree deep and click-heavy; more makes long lists nobody scans.
No more than about six levels deep.
No “Miscellaneous” or “Other” nodes. They fill up fast and defeat the purpose.
Singular names for leaf nodes, plural acceptable for parents. Keep abbreviations and punctuation to a minimum.
An image on every node, and a schematic where the attributes aren’t obvious.
Node or attribute? If you’ll have hundreds of a thing that people browse separately, make it a node. If you’ll have a few dozen, an attribute with valid values is easier to maintain.

Start from the predefined classification that ships with Windchill for common mechanical and electrical components, then extend it for your own parts.

Governance: who creates and classifies new parts

This decision matters more than any setting. There are three common models:

Model	How it works	Tradeoff
Components group, full control	Engineer searches, then requests a part; a commodity expert creates, classifies and releases it	Cleanest data and most reuse; slowest for engineers and needs dedicated staff
Components group oversight	Engineer creates and classifies; the group reviews and releases	Engineers keep moving; still needs a reviewer with time
Self-management	Engineer creates, classifies and releases	Fastest; nothing forces a search first, and misclassified parts are hard to fix after release

Whichever you pick, name one person as the classification administrator who owns the tree, and give each business unit a representative who watches new part creation.

A sensible order to roll it out

Search first. Let engineers find what already exists before anything else changes.
A new-part request process, so nobody creates a part without searching.
Classification: build the tree, attributes and naming rules.
Migrate and cleanse legacy part data: extract from ERP, spreadsheets and old systems, normalize descriptions and attributes, merge duplicates, then load and validate.
Manage sources: manufacturer and vendor parts, approved manufacturer and vendor lists.
Maintain the catalog: obsolescence notices, spec updates and changes to the tree itself, through your change process.

Where classification projects go wrong

From our implementation work:

Loading the mess. Migrating thousands of legacy parts without normalizing descriptions and merging duplicates just makes the duplicates searchable.

A tree designed by one person. If engineers didn’t help name the nodes, they won’t find anything in it. Test the tree with the people who’ll search it.

No rule to search first. Without a request process or a review step, engineers keep creating parts because it’s faster than searching.

Redesigning in place. You can move nodes and reclassify parts later, and attribute values carry over. But don’t delete nodes or attributes that parts already use, and don’t change a unit of measure that’s in use.

Frequently Asked Questions

What is parts classification in Windchill?
It’s organizing parts into a structured tree of categories, each with its own attributes, so engineers can find an existing part by what it is instead of guessing its number or description. In Windchill this is done with PartsLink, part of PDMLink.

Is parts classification the same as parts reuse?
No. Classification is the structure: categories, attributes and naming. Reuse is the payoff: engineers find and use an existing part instead of creating a duplicate. Classification makes reuse possible; a governance process makes it happen.

Should we add a new classification node or use an attribute?
It depends on volume and search. If you expect hundreds of parts that people will browse separately, a node helps. If there are only a few dozen, an attribute with a fixed list of values is easier to manage and still searchable.

Who should be allowed to create and classify new parts?
There are three common models: a components group creates every part, a components group reviews parts that engineers create, or engineers manage it themselves. Central control gives the cleanest data but can slow engineers down; self-management is fastest but lets duplicates and misclassified parts in.

Can we reorganize the classification tree after parts are classified?
Yes, with care. You can move nodes and reclassify parts, and attribute values are kept. Avoid deleting nodes or attributes that parts already use, and don’t change a unit of measure that’s in use.

Setting up classification in Windchill?

Element Consulting has implemented, upgraded and rescued Windchill since 2003, across 100+ PLM implementations. We’ll help you design a tree your engineers will actually search.
Talk to a Windchill Expert

Parts Classification in Windchill: How PartsLink Works and How to Set It Up