AEE Engineering6 min read
Every electronic product is a bet that its parts will keep existing. Vendors do not honor that bet: components get discontinued when fab lines retire, acquisitions rationalize catalogs, or a package falls out of fashion. A product designed for a fifteen-year life will, statistically, watch a large share of its BOM go end-of-life along the way.
None of this is preventable. All of it is manageable — and the difference between managed and unmanaged obsolescence is roughly the difference between routine maintenance and an emergency respin with the production line down.
How parts actually die
Discontinuation is a process with warnings, not an event. A part typically moves from active to not-recommended-for-new-designs (NRND), then to a product change notice (PCN) or end-of-life (EOL) notice announcing a last-time-buy (LTB) date, then to discontinued. The window between notice and last order is commonly six months to a year.
That window is the entire game. Inside it, every option is open and cheap. After it closes, the remaining options are the secondary market and redesign, and both cost multiples more.
Why teams miss the window
The information is public — vendors publish notices, distributors relay them — so the failure is rarely informational. It is organizational: nobody's job is to read notices against your specific BOMs. The design team moved to the next project, purchasing tracks price and stock rather than lifecycle, and the notice lands in an inbox nobody owns. Eighteen months later, the buyer discovers the part is gone.
An obsolescence practice fixes the ownership problem before anything else: someone, internal or contracted, whose explicit job is watching your parts.
The four habits of a working practice
1. Monitor the whole BOM, continuously. Component lifecycle databases score every commercial part's status and projected availability; PCN feeds catch changes between scores. The mechanics take hours per month once set up. What matters is that the coverage is complete — the part that kills you is always the one nobody flagged.
2. Score risk, not just status. A discontinued resistor is a non-event; a discontinued microcontroller with your firmware on it is a project. Rank each line by replacement difficulty: how many qualified alternates exist, whether they are drop-in, how deeply the part is entangled with firmware and layout. Attention goes to the short, dangerous end of the list.
3. Act inside the window. For each flagged part there are three moves — stock up before the deadline, qualify a replacement, or redesign the affected section. Choosing between them is a real engineering-economics decision; we break down that choice on the sustaining engineering page. What matters here is when the choice happens: inside the notice window, all three options exist. After it, the first one is gone and the second gets harder.
4. Prevent at design time. The cheapest obsolescence management happens before the schematic is finished: prefer parts early in their lifecycle, require second sources for anything critical, and confine single-source exotics to places where they earn the risk. A day of lifecycle-aware part selection during design saves respins a decade later.
The broker trap
When a part is gone and demand remains, the secondary market appears — brokers with stock of discontinued components at spot prices. Sometimes that stock is genuine surplus. Sometimes it is relabeled, refurbished, or counterfeit, and counterfeit semiconductors range from cosmetically identical parts that fail early to empty packages with the right markings.
Broker stock is occasionally the right bridge, but treat it as a controlled substance: reputable sources, authenticity testing on arrival, and never as the long-term plan. A product whose BOM depends on brokers is a product that has already lost; it just has not stopped shipping yet.
When the documentation is gone too
Obsolescence has an advanced form: the product whose parts and design knowledge have both decayed. The schematics are missing, the last engineer left, and production survives on folklore. Managing part risk requires knowing what the parts do, so documentation recovery comes first — often via reverse engineering the product's own boards back into editable CAD.
The general lesson is the same at every scale: obsolescence rewards whoever moves first. An afternoon spent auditing your BOM's lifecycle status this quarter is the cheapest engineering you will buy all year.