Four feeds, one join

What this is, and what it is not

Ingestion of component lifecycle data from SiliconExpert, Z2Data, Accuris and Octopart, matched against the manufacturer parts on your approved manufacturer list and joined to the products those parts actually ship in.

It is not a parts search engine. Looking up a component is what the vendors' own portals are for, and they do it well. What none of them can do is the join, because none of them know your internal part numbering, your AML, or your bills of materials.

It is also not a substitute for a subscription. You bring your own feed credentials; Manufacturing PLM ingests what your contract entitles you to. A PLM reselling component data would be adding a margin to somebody else's product and a dependency you did not choose.

The mechanism: match, then disagree honestly

Matching runs on manufacturer part number and nothing else. Not descriptions, not internal numbers — the MPN is the identifier the outside world shares, and it is the only one a feed has ever heard of.

An internal part typically carries two or three approved manufacturer parts, each with its own lifecycle state. That is why a single status field on a part is never quite right: one source is end-of-life, one is active, and the honest answer depends on which you actually buy.

Where feeds disagree, the disagreement is shown. Two sources reporting active and one reporting not-recommended-for-new-designs is real information — it usually means the manufacturer has signalled something one aggregator has picked up and the others have not. Collapsing that into a single score throws away the part a buyer would act on.

For alerting, the earliest date wins. If one feed says last-buy 2027-03 and another says 2027-09, the alert uses March, because being early costs a conversation and being late costs a line stoppage.

The failure it prevents

A company subscribes to a good feed. The data is accurate and arrives reliably. Somebody exports a spreadsheet monthly and cross-references it against a BOM export.

The cross-reference works on part descriptions, because the export does not carry MPNs consistently. It produces forty possible matches, most of which are wrong in ways that take a few minutes each to dismiss, and the exercise gets done quarterly instead of monthly. Then it gets done annually. Then somebody leaves.

The data was never the problem. The join was, and a join that costs a day is a join that stops happening — which is why the notice about a part in two shipping products arrives eight months after it was issued.

How it meets the rest of the product

Exposure is computed through resolved structures, so a part present only in a superseded revision is not reported as live risk. Reporting everything trains people to skim, which is precisely how the alert that mattered gets missed.

House numbers with no AML entry match nothing, and are surfaced as unmatched rather than matched to something plausible. A wrong match is worse than a missing one, because nobody goes back to check it. The unmatched list is also the most concrete argument for maintaining the AML that anyone has ever produced.

Risk scoring stays four steps — ok, watch, at risk, critical — with the contributing signals named. Engineers stop trusting a scale they cannot hold in their head, and a five-point scale invites arguments about the middle band.

The Supplier Agent works this surface: triaging the week's notices by what they block, finding alternates already qualified on other products, and drafting the change where one exists. Every claim cites the notice, the manufacturer part and the products.

How it meets your ERP

Feeds supply the risk; your ERP supplies the urgency. On-hand, committed, open purchase orders and lead time read back and sit on every alert, so it reads as “three shipping products, eleven weeks of runway, no qualified alternate” rather than as a status change on a component.

The supplier record itself masters in your ERP. What Manufacturing PLM adds is which manufacturer parts engineering approved and their qualification state — the AML — which is the thing the feeds join against and the thing your ERP has no opinion about.

Nothing transactional is written. Manufacturing PLM does not raise a last-time-buy order or reserve stock; it produces the exposure that makes the decision well-informed, and the buyer acts where they already work.

Where the boundary is

Manufacturing PLM cannot see risk in parts it has no data for. A component bought under a house number with no approved manufacturer entry behind it has nothing to match against, and the honest report is that nothing was found rather than that nothing exists.

It also does not forecast. What you get is what four commercial feeds report, disagreements included, joined to what you build. An agent or a scoring model that predicted obsolescence would be selling confidence rather than information, and the first wrong prediction would cost more than every right one saved.