The Question Behind the Part Proxy Debate
On 19 May 2025, CADTALK’s Integrate Intelligently podcast published an episode with Prof. Dr. Jörg W. Fischer of Karlsruhe University of Applied Sciences. It gave a catchy English name, the “Part Proxy,” to an idea Fischer has developed since 2011: one engineering part in PLM giving birth to multiple plant-specific materials in ERP. In January of this year, Oleg Shilovitsky published his analysis of why Form-Fit-Function decisions fail without product memory. Two sharp thinkers, two angles on the same wound.
I have followed this discussion with more than average interest because it touches the core subject of my book, The Essential Guide to Part Re-Identification. And to be honest, I think the industry is having this debate one layer too low. We argue about part numbers, material numbers, and BOM structures as if “interchangeable” were a settled term. It is not. Define it properly, and an answer appears that has been sitting in the CM2 framework all along: the equivalent item record. In this article, I will put that answer next to the Part Proxy and weigh both honestly. It is also the first of two articles on this subject: this month, the identification question, what the claim is and where to record it; next month, the governance question, no data model can settle, who actually validates it.
What Does a Part Number Actually Promise?
A part number is not a fact about a part. It is a promise: any unit that carries this number fully replaces any other unit that carries it, in any application, for anyone who will ever touch it.
Here is that promise under load. Two plants build Part #7842 to the same engineering specification, one in Bavaria, one in Baja California. Different suppliers, different processes, different raw materials; both approved after a Form, Fit, and Function check. A field technician needing a replacement pulls #7842 from local stock, installs it, and the machine works. The promise is kept. Six months later, the same machine needs an overhaul, and this time the technician does not swap the part. She goes inside it. The internal layout is unfamiliar, the seal dimensions differ from the service manual, and the diagnostic port speaks a different protocol than her tablet expects. The part number says she is holding the same component. The construction says otherwise. And nobody can answer her most basic question: which plant made this unit?
Note that nothing failed procedurally in that story. The failure is in what the checks never covered, which is why Form, Fit, and Function alone cannot carry the promise. Hold on to this scenario; we will run it through a proper definition shortly.
Fischer’s Answer: Separate the Material
Why is the promise getting harder to keep? Because the pressure to let plants build the same part differently is structural: supply chain disruptions forced manufacturers to qualify regional alternatives, and local content rules mandate local sourcing. Forcing one material number across all plants creates a conflict every CM professional will recognize: the moment Plant Monterrey wants to substitute Supplier Y for Supplier X, a change order fires, weeks pass, the plant runs short, someone makes an informal decision, and the data stops being true. The cost is measurable: one 2025 review reports that ineffective EBOM-to-MBOM processes contributed to 32-41% of new-product-introduction delays.
Fischer traced this class of problem to its root: in his Structure Theory, information models are the crystallization seeds of a company’s business processes. His structural answer is the Object Type Cascade, one of his seven keys to solving PLM and ERP integration, a.k.a. the Part Proxy: one PLM part acts as the master specification, and from it multiple ERP materials are born, one per plant, each carrying its plant’s local supply chain decisions. The cascade rests on Fischer’s unhedged rule: “Material numbers must be revision-free!”, because for the customer, the same material must always be 100% interchangeable.
Fischer’s diagnosis is correct, and the Object Type Cascade is a valid solution. Having said that, I do not believe it is the only one, but to see the alternative, we first have to take “fully interchangeable” seriously.
What Does “Fully Interchangeable” Actually Mean?
The conventional definition is Form, Fit, and Function. That view is incomplete. In my book, I argue that true interchangeability requires a fourth dimension: Interface, the interaction between parts and between parts and their environment and users, from communication protocols to APIs. Form, Fit, Function, and Interface (F3I). The definition I favor, building on IpX’s, is deliberately spare:
Parts that are capable of fully replacing each other in any application. Fully: achieving the same results without any special measures, except for normal adjustment.
Two words in that definition carry more weight than most FFF checklists assess. “Any application” means every parent and dependent part ever issued that is not yet end of life. And “fully” reaches far beyond the moment of installation. In my book, Fit explicitly includes serviceability requirements: accessibility, replaceability, modularity, labeling, diagnostics, and documentation that matches the part you are holding. Upgradeability belongs in the same column. Repair appears again under Function, as part of a product’s sustainability requirements. In other words, a part is not fully interchangeable because it installs and runs. It is fully interchangeable when replacement, service, repair, and upgrade all achieve the same results without special measures.
Now run the Bavaria and Baja California variants through that definition. As sealed units, they replace each other perfectly. Below the output level, the differences surface: different internals, different seals, a different diagnostic protocol. Service cannot use one manual, one spares kit, or one procedure for both. Judged across the full scope of F3I, these are not two instances of the same part. They are two different parts that happen to be mutually replaceable.
That is the conclusion the whole debate keeps circling without landing on: the honest identification was never one part number.
Two Parts, One Promise: The Equivalent Item Answer
Category 3.0 of the CM2 framework, Naming, Numbering and Reuse, states the identification principle in one line: identification numbers are used to control and signify interchangeability. Read it in both directions. A shared number must guarantee full interchangeability, and where full interchangeability ends, so must the shared number. Our two constructions differ in ways that matter for service, repair, and upgrade, so each gets its own part number. The number becomes a true statement about construction: the technician reads it and knows which internals, which manual, which spares.
But what about replacement, where the two parts do substitute perfectly? CM2-500 provides the instrument:
“Items that have different identification numbers, but remain fully interchangeable, are cross-referenced in an equivalent item record.” (CM2-500, Section 3.06)
Note that this resolves the earlier verdict rather than contradicting it. What failed the F3I assessment was never the exchange of the units; it was one number claiming two constructions. Give each construction its own identity, with its own manual, spares, and diagnostics attached, and each number keeps its promise. The equivalent item record then certifies the one claim that holds across both: as complete units, the two parts fully replace each other in every application.
This is the position I want to put on the table: re-identification plus an equivalent item record achieves what the Part Proxy achieves, inside standard CM identification rules. Plant Monterrey’s supplier switch is assessed against Monterrey’s part number alone, and no change order fires against the other plants. Replacement freedom is preserved: the equivalent item record tells every planner, stocking system, and technician that the parts substitute in every application. And manufacturing-origin traceability no longer needs a separate record, because it is the part number itself: which plant made this unit is printed on the label.
Note that CM2 also covers the case where the claim is not universal. An approved alternate item record covers “an item used on a temporary basis in place of a preferred item,” cross-referenced “with the Change Notice authority for its use” (CM2-500, Section 3.07). Equivalent means the promise holds everywhere, permanently. Alternate means it holds where, and for as long as, the change authority says so. Two parts can keep one promise, as long as each carries its own number.
Does Re-Identification Not Explode Engineering’s Workload?
The standard objection is cost: every new part number means a new drawing and new documentation, multiplied across plants. That objection assumes the part and its CAD model are the same object. They are not. PLM systems already support separating the CAD part from the part: the CAD model is a reusable definition of geometry that one or more parts reference, the digital descendant of the tabulated drawing, where one drawing defines a whole family of part numbers.
Engineering already works this way in cases nobody argues about. An anodized bracket offered in three colors is one shape, one CAD model, and three part numbers. A cable assembly cut to different lengths is one parameterized model with a table of lengths, each length has its own number. Two housings with identical geometry, one machined from a standard alloy and one from a marine-grade alloy, are one model with two material specifications and two part numbers. In each case, the shape is shared, and the identity is not, because the parts do not replace each other and the numbers must say so.
Apply the same mechanism to our plant variants. Bavaria and Baja California reference the shared geometry where their constructions are common and carry their own construction-specific datasets where they diverge. A new part number then consumes a number and a cross-reference, not a new documentation set. The administrative weight lands where it belongs: on the interchangeability assessment, not on document production.
There is a second reason to keep identification in engineering hands. Design for Manufacturing and Design for X (DfX), more broadly, requires design engineers to treat requirements from manufacturing, logistics, and service as design inputs, not downstream complaints. Engineering-governed sibling parts keep the diverging realizations in the design engineer’s field of view; plant-owned materials move that knowledge out of view, and risk institutionalizing the wall DfX exists to tear down.
So, Which Solution Should You Choose?
Each has its pros and cons, and I want to be fair to both.
The Part Proxy’s great strength is structural enforcement. The boundary between engineering definition and manufacturing realization is built into the data model, so a local supplier switch physically cannot fire an engineering change order. It works even where change discipline is weak, and it matches how ERP already thinks. Its weaknesses follow from the same design. Interchangeability is asserted by construction, never assessed: every material born from the same part is declared 100% interchangeable, with no place to record that constructions diverged below the output level. There is no vocabulary for scoped or temporary substitution. It requires re-architecting integrations that mostly assume one part maps to one material. And it runs two vocabularies across the lifecycle, part numbers in engineering and material numbers in operations, so the technician’s question is only answered if material numbers reach the field.
Re-identification with equivalent item records is the mirror image. Every claim is explicit, assessed, and auditable; one vocabulary runs from design to recall letter; the alternate record expresses the scoped claims that the proxy cannot. Its weakness is that nothing in the data model forces re-identification when constructions diverge. It substitutes process for structure, and where governance is immature, the assessment simply will not happen. Equivalence records also need maintenance: every change to a sibling reopens the question.
Note that both solutions need the same ERP machinery to let one plant’s output satisfy another plant’s demand. The proxy does not eliminate equivalence logic; it relocates it into ERP configuration, usually with less governance attached.
Having said that, the two are not mutually exclusive. A hybrid is coherent: proxy-style materials where constructions are genuinely identical, re-identification with an equivalent item record the moment they diverge. I do not believe there is one right solution for every company. The choice hinges on governance maturity, systems landscape, and aftermarket obligations. But nobody should claim a separate ERP material is the only way out.
Who Signs Off on the Equivalence?
Whichever mechanism you choose, someone has to validate the claim it records. This brings us to what I consider Fischer’s most durable contribution, more than the Part Proxy itself: the Form-Fit-Function Corridor, the corridor a part’s interchangeability claim must hold across its whole lifecycle, from first design intent to final service action. Who signs off that two constructions are truly equivalent, in every application, across all four F3I columns? No single department can. Engineering rarely reaches the serviceability requirements inside Fit or the Interface dimension, which FFF does not even name; manufacturing sees process constraints; service sees diagnostic protocols and field tooling. And note that the question arrives before any record is written: the re-identification decision itself is of the same nature as the impact analysis on a change request, a cross-functional decision, not an engineering-only one.
I am deliberately not going to resolve that question in a closing section, because it cannot be resolved in one. It deserves an article of its own, and it will get one: next month I take the governance question apart, starting from the oldest interchangeability demonstration on record. For now, hold on to this much: an equivalence claim nobody validated is not a claim, and no data model can turn it into one.
What Should You Do First?
Do not buy a tool. Three commitments come first.
First, define “fully interchangeable” across the complete F3I model, including serviceability, upgradeability, and repair, with all lifecycle stakeholders at the table. Re-identification is not about following a decision tree; it is about understanding the full scope of interchangeability before making the claim.
Second, separate the CAD part from the part, so that honest identification stays affordable. Shared shape stays one model; distinct constructions get distinct numbers without spawning distinct documentation sets.
Third, when constructions diverge below the output level, re-identify and record the replacement relationship deliberately: an equivalent item record where the claim is universal, an approved alternate record with its Change Notice authority where it is not.
Closing Thoughts
If your plants build different constructions under one part number, your service organization is already paying for a promise that number cannot keep. The Part Proxy debate treats that as a data architecture problem. I see it as an identification problem, and configuration management solved it long before we gave it a catchy English name. Interchangeable does not mean identical, and it never did.
Next month, part two: we go back to a workshop near Paris in 1785, where interchangeability was proven the honest way, and ask who plays the witness in your organization today.
So stop asking whether two parts are the same. Ask instead: is the promise between them universal or scoped, who signed off on it, and does your data model make that claim visible or hide it? I am curious how your organization records equivalence today. Let me know what you think.