Component Obsolescence Management: DMSMS, Lifecycle Risk Scoring and Last-Time-Buy Decisions in PLM

A product that ships for fifteen years is built from parts that live for three. That mismatch is the entire problem. A microcontroller goes end-of-life, a supplier issues a discontinuance notice, and the clock starts on a decision that can cost either a few thousand dollars in stock or a six-figure redesign. Component obsolescence management is the discipline of making that decision early, with data, instead of late, in a panic.

It matters more in 2026 than it did a decade ago because bills of materials are longer, silicon nodes churn faster, and the products that suffer most (industrial controllers, medical devices, avionics, rail, energy gear) are contractually obliged to ship and service for ten to thirty years. Your PLM system already knows what is on every BOM. What it usually lacks is a feedback loop from the supply side.

This post builds that loop. You will leave with a risk-scoring model you can implement, a worked last-time-buy (LTB) calculation with runnable Python, and a pattern for turning lifecycle alerts into engineering change requests.

What this covers: the causes and vocabulary of obsolescence, the DMSMS standards landscape, a reference architecture that connects lifecycle data to PLM, a scored BOM-health example, LTB sizing math, form-fit-function substitution, counterfeit exposure, and the failure modes that sink real programs.

Context and Background

Obsolescence is not one event. It is a family of supply-side losses: a manufacturer discontinues a part, a fab line closes, a die shrink retires an older process, a raw material becomes unavailable, a software tool chain loses support, or a sole-source supplier is acquired and rationalises its catalogue. In the US defence community the umbrella term is DMSMS, short for Diminishing Manufacturing Sources and Material Shortages. The Defense Standardization Program Office describes a DMSMS issue as the loss, or impending loss, of manufacturers or suppliers of items, raw materials or software. The same office publishes the SD-22 guidebook, “DMSMS: A Guidebook of Best Practices and Tools for Implementing a DMSMS Management Program,” which its site lists as updated in January 2026 (DSP DMSMS program).

The vocabulary of notices matters because each one starts a different clock. A Product Change Notification (PCN) announces a change to form, fit, function, process or materials. A Product Discontinuance Notice (PDN, sometimes called an end-of-life or EOL notice) announces that a part will stop being sold. JEDEC publishes the conventions for solid-state suppliers: JESD46 covers customer notification of product and process changes, and JESD48C, “Product Discontinuance” (December 2011, revising the May 2005 B edition), governs discontinuance. Clause 3.1 of JESD48C sets the minimums most engineers quote: final orders (the last-time-buy window) at least six months after the notice, and final shipments at least twelve months after it (JEDEC JESD48C, as hosted by a supplier). Individual suppliers may offer more, and non-JEDEC suppliers may offer less, so treat those figures as a floor you verify per vendor rather than a guarantee.

Two things follow from those numbers. First, a part with a twelve-month last-ship date can be perfectly healthy in your ERP stock report and still be dead to new orders in six. Second, a six-month window is shorter than the redesign cycle of most regulated products. A medical device that needs a verification and validation cycle, or an avionics board that needs re-qualification, cannot redesign in six months. For those products the realistic options after a PDN are a last-time buy or a drop-in substitute, which is why detection speed dominates every other variable.

Standards give the management layer. IEC 62402:2019 (second edition, published 29 May 2019) is titled “Obsolescence management” and sets requirements and guidance covering policy, infrastructure, an obsolescence management plan, design strategies to reduce obsolescence, choosing an approach, selecting and implementing resolutions, and measuring outcomes (IEC webstore). On the US side, SAE STD0016A, “Standard for Preparing a DMSMS Management Plan,” was revised on 12 December 2023; the original STD0016 dates from 1 August 2011, and the standard lineage is commonly cited as GEIA-STD-0016 (the SAE page lists the 2011 issue as the historical predecessor but does not itself use the GEIA name, so confirm the designation your contract cites). SAE states that following its Section 3 requirements helps a plan owner meet obsolescence requirements in EIA-STD-4899, AS5553 and DFARS 252.246-7007 (SAE STD0016A).

Commercial data comes from lifecycle intelligence providers. SiliconExpert, Z2Data, and the IHS Markit lineage (now part of S&P Global) are commonly named examples, alongside distributor and manufacturer feeds. This post does not endorse any vendor; coverage, update latency and match accuracy differ, and you should benchmark them on your own BOMs. For related change workflow, see our guide to engineering change management with ECR, ECO and ECN, because every obsolescence resolution ends as a change order.

A Reference Architecture for Obsolescence Management in PLM

Component obsolescence management works when a status change on a manufacturer part number becomes a prioritised, assigned engineering task within a day, and fails when it becomes an email that sits in a shared inbox. The architecture below makes that handoff mechanical: ingest, match, locate, score, route.

Figure 1: Lifecycle signals are matched to the PLM part master, expanded through where-used, scored, and routed by risk tier into change management.

The diagram shows three input streams converging on an aggregation layer, a match against the part master, a where-used lookup across every released and in-work BOM, a scoring engine, and a three-way branch by risk tier. High-risk parts open an engineering change request (ECR) carrying the LTB-or-redesign decision, which later becomes an engineering change order (ECO) with effectivity. Low-risk parts are simply monitored on a schedule, so engineers see a short, ranked list instead of a firehose.

Ingest and normalise lifecycle signals

Three signal families matter. Manufacturer notices (PCN, PDN, and end-of-life announcements) are authoritative but inconsistent in format and often arrive as PDF attachments. Lifecycle data providers normalise those notices into structured fields such as lifecycle stage, estimated years to end-of-life, number of known sources, and cross-reference candidates. Distributor signals, such as stock depth, lead-time drift and price spikes, are not notices at all but they are the earliest warnings you get: a lead time that doubles over two quarters often precedes an allocation or discontinuance announcement. None of these is complete alone, so the aggregation layer should keep provenance per field and never overwrite a manufacturer-stated date with a model-estimated one.

Data quality is the first failure point. Lifecycle estimates are predictions, not facts. A provider’s “estimated years to end-of-life” is a statistical forecast and should be stored as a distribution or at minimum with a confidence flag. Treat it like a weather forecast: useful for ranking, dangerous as a contractual date.

Match on the part master, not on strings

The match step is where most integrations leak accuracy. Your PLM part master contains your internal part numbers, each linked to one or more approved manufacturer parts (the AML or AVL, approved manufacturer list or approved vendor list). Lifecycle feeds key on manufacturer plus manufacturer part number (MPN). If the PLM stores MPNs as free text with inconsistent suffixes, packaging codes and temperature-grade variants, the join silently drops parts. Normalise MPNs on ingest, store the manufacturer as a controlled entity, and record packaging and grade as separate attributes. A part that matches on base number but not on the ordering code can be a different qualified item with a different lifecycle, so the match should be explicit about its confidence level.

Where-used turns a part event into a product impact

A discontinuance of one capacitor means nothing until you know it sits on eleven assemblies, four of which are in production and two in a customer’s installed base under a ten-year support contract. The where-used query must traverse released revisions, in-work revisions and, crucially, configured variants. If your product uses a 150 percent BOM with option-driven variants, see our walk-through of CPQ configurators and variant BOM structures for how the same part can appear conditionally. Weight each hit by the volume and remaining life of the assembly, not just by the count of assemblies.

Route by tier into change management

The routing rule should be deterministic and documented. For example: high tier opens an ECR within one business day and assigns a component engineer; medium tier triggers an alternate-qualification task with a due date; low tier is reviewed quarterly. The ECR should be created by the system with the where-used report, the risk score components, the supplier’s notice date and the computed LTB-versus-redesign comparison attached. That pre-population is what saves the week that people otherwise lose gathering facts.

Reactive, Proactive and Strategic Management

Programs sit on a maturity ladder. Reactive management waits for the PDN and then scrambles. Proactive management monitors the BOM continuously and acts on forecasts. Strategic management designs obsolescence resilience into the product before the BOM is frozen. The SD-22 guidebook is explicitly about enabling proactive management, and IEC 62402 includes design strategies as a distinct step for the same reason.

Figure 2: The three management postures, with the typical practice at each level.

The figure shows the ladder from left to right, with an example practice under each rung: panic LTB and broker purchases at the reactive level, BOM-health monitoring and pre-qualified alternates at the proactive level, and preferred-parts lists plus supplier longevity commitments at the strategic level.

What reactive really costs

Reactive management is not merely expensive; it is risky in a specific way. When a PDN lands with a six-month order window and no alternate is qualified, the options collapse to a large LTB or to the grey market. The LTB quantity is a forecast made under pressure, with no time to validate the demand model. The grey market is where counterfeit exposure concentrates, a point we return to below. Teams in this mode also tend to over-buy, because the penalty for running short (a line-down event) feels worse than the penalty for carrying surplus. That asymmetry is rational, but it should be priced explicitly rather than expressed as a safety factor chosen by gut feel.

What proactive management changes

The proactive posture uses lifecycle data to move the decision earlier. A part flagged as approaching end-of-life two or three years out lets you pre-qualify an alternate at low cost, fold the swap into a planned board spin, or negotiate a buffer inventory while suppliers are still cooperative. The economic argument is simple: the same mitigation (qualifying an alternate) costs less when it is scheduled than when it is an emergency, and the option value of time is highest before the notice. Studies and guidebooks consistently argue this direction, but published cost multipliers vary and depend on the sector, so build your own estimate from your engineering rates rather than quoting a headline number.

What strategic management adds

Strategic management acts at design time. Preferred-parts lists steer designers toward components with long-life commitments and multiple sources. Architecture choices isolate volatile parts: putting a processor on a standardised module (a system-on-module with its own supplier-managed lifecycle) confines a CPU obsolescence event to one replaceable item. Programmable logic can absorb discrete-logic obsolescence. Abstraction layers in firmware reduce the cost of changing a peripheral. Each of these costs something up front, so they pay back only on long-lived products. A consumer gadget with an eighteen-month life should not pay for them; a rail signalling controller should.

Building the BOM Health Model and Part Lifecycle Risk Score

BOM health is the aggregate expression of part lifecycle risk across a bill of materials. A useful model is transparent, tunable and auditable: weighted factors, normalised to a zero-to-one-hundred scale, combined into a part score and rolled up to an assembly score.

The scoring model

Choose factors that you can actually measure and that map to action. A defensible starting set is: availability and lifecycle stage (how close to end-of-life), source count (how many manufacturers can supply the item or an equivalent), market trend (lead-time and price movement), substitute difficulty (how hard a drop-in replacement is), and criticality (the consequence if the part cannot be supplied). Weights are a policy choice. The numbers below are illustrative, not an industry standard; calibrate them against your own history of past obsolescence events.

Factor	Weight	Scale meaning (0 low risk, 100 high risk)
Lifecycle stage	0.30	0 active and growing, 50 mature, 80 NRND or last-time-buy announced, 100 discontinued
Source count	0.20	0 for three or more sources, 40 for two, 100 for sole source
Market trend	0.15	Based on lead-time and price drift over trailing two quarters
Substitute difficulty	0.15	0 pin-compatible alternate exists, 100 no equivalent, requires redesign
Criticality	0.20	Impact class of failure to supply or qualify

NRND stands for “not recommended for new designs,” a status many suppliers apply before a formal discontinuance. It is an early warning, and a part in NRND status should already have a nonzero score even though it is still orderable.

A worked example

Take three parts on an industrial gateway board. Part A is a sole-source, specialised motor-control microcontroller whose manufacturer has announced a PDN. Part B is a commodity ceramic capacitor with several sources. Part C is a mature interface transceiver with two sources and a pin-incompatible successor.

Part	Lifecycle	Sources	Trend	Substitute	Criticality	Weighted score
A: MCU, sole source	80	100	60	40	100	79
B: capacitor	30	40	20	20	50	33
C: transceiver	90	60	80	100	20	70

The arithmetic for Part A is 0.30 times 80 (24.0), plus 0.20 times 100 (20.0), plus 0.15 times 60 (9.0), plus 0.15 times 40 (6.0), plus 0.20 times 100 (20.0), for a total of 79. Part B sums to 33 and Part C to 70. With tier thresholds of 65 and above high, 40 to 64 medium, and below 40 low, A and C both open an ECR and B is monitored. Notice that C scores high despite low criticality, because the lack of a drop-in alternate and a late lifecycle stage dominate; that is the right behaviour, since low-criticality parts are exactly the ones that get ignored until a line stops.

The same weighted sum aggregates to assemblies. Do not simply average part scores, because one failing sole-source part is worse than ten healthy ones. A better roll-up uses the maximum, or a high percentile, plus a count of high-tier parts. Many teams report two numbers per assembly: the worst part score and the percentage of BOM lines in the high tier. Some also publish a cost-weighted exposure, which sums extended cost of at-risk lines against the annual product margin.

Making the score actionable

A score is only useful if it names the next step. Attach a default action to each tier and make the system propose it: monitor, qualify alternate, or decide LTB versus redesign. Record every override with a reason. Over a year, those overrides are training data: if engineers repeatedly downgrade a factor, the weight is wrong. Also store score history per part so you can plot trend; a part drifting from 45 to 62 over two quarters is a more urgent story than one that has sat at 62 for years.

Responding to a Notice: LTB Versus Substitute Versus Redesign

When a PDN arrives, three resolutions compete: buy enough stock to outlast the product, swap in a substitute, or redesign around the loss. Which wins depends on remaining demand, the cost of change, and how much risk you can stomach. The Defense Standardization Program and IEC 62402 both describe a menu of resolutions and ask the owner to select by cost-effectiveness; the decision flow below is a practical encoding.

Figure 3: After a PDN, check for a form-fit-function drop-in first, then compare remaining demand against redesign cost.

The flow starts at day zero with the notice, matches the part to every BOM within a day, and then branches. If a form-fit-function drop-in exists, qualify it and file the ECR. If not, compare the lifetime cost of buying and storing against the cost of redesign, and either size and place the LTB before the last order date or start the redesign with a bridge buy to cover the gap.

Form, fit, function substitutes

A form-fit-function (FFF) substitute matches the original in physical footprint and mechanical envelope (form and fit) and in electrical and operational behaviour (function). The catch is that “function” has more dimensions than the datasheet headline suggests. Check absolute maximum ratings, not just typical values; timing margins; thermal resistance and derating; power-up sequencing and reset behaviour; errata; and, for programmable parts, the silicon revision and boot ROM. A pin-compatible microcontroller from the same family may still change interrupt latency or peripheral register layout. For regulated products, even a true FFF swap usually requires a documented equivalence assessment and sometimes regression testing, so budget the qualification effort rather than assuming zero.

Also check the substitute’s own lifecycle. Replacing a part with its own successor’s older sibling just moves the problem two years out. Run the candidate through the same scoring model before approving it.

Last-time-buy quantity math

An LTB is a bet that you can forecast demand over the remaining life of the product. The core formula is demand over the horizon, inflated for scrap, spares and uncertainty, minus stock on hand and on order, rounded up to the supplier’s minimum order multiple. Here is a worked case with illustrative numbers.

Assume annual consumption of 1,200 units, an eight-year support horizon, flat demand, a 3 percent production scrap rate, a 5 percent allowance for field-service spares, 2,000 units already on hand, an annual demand standard deviation of 150 units, a one-sided service level of 95 percent (z of 1.65), and a minimum order multiple of 250.

Horizon demand is 1,200 times 8, which is 9,600. Adjusting for spares and scrap gives 9,600 times 1.05 divided by 0.97, or about 10,392. Safety stock for demand uncertainty across the horizon is z times sigma times the square root of years: 1.65 times 150 times the square root of 8, or about 700. The gross requirement is therefore roughly 11,092; subtracting 2,000 on hand leaves about 9,092, which rounds up to 9,250 at the 250-unit multiple. At an illustrative unit price of 4.20 the buy is 38,850.

The code below reproduces this and is meant to be adapted to your own demand model.

import math

def ltb_quantity(annual_demand, years, growth=0.0, scrap=0.03,
                 spares=0.05, on_hand=0, on_order=0,
                 z=1.65, sigma_annual=0.0, moq=1):
    """Return (buy_qty, horizon_demand, gross_need).

    Assumptions: independent annual demand errors, so uncertainty
    over the horizon scales with sqrt(years). Scrap is applied as a
    yield divisor; spares as a multiplier.
    """
    whole = int(years)
    demand = sum(annual_demand * (1 + growth) ** t for t in range(whole))
    demand += annual_demand * (years - whole) * (1 + growth) ** whole
    need = demand * (1 + spares) / (1 - scrap)
    need += z * sigma_annual * math.sqrt(years)
    net = max(0.0, need - on_hand - on_order)
    buy = math.ceil(net / moq) * moq
    return buy, demand, need

def carrying_cost(buy, unit_cost, annual_rate, years):
    # Linear drawdown from buy to zero: average inventory is buy/2.
    return buy * unit_cost * annual_rate * years / 2

buy, d, n = ltb_quantity(1200, 8, growth=0.0, scrap=0.03, spares=0.05,
                         on_hand=2000, z=1.65, sigma_annual=150, moq=250)
print(buy, round(d), round(n))        # 9250 9600 11092
purchase = buy * 4.20                  # 38,850
carry = carrying_cost(buy, 4.20, 0.08, 8)   # 12,432
print(purchase, carry, purchase + carry)    # 38850.0 12432.0 51282.0

The carrying cost line deserves attention. At an 8 percent annual holding rate, a linear drawdown over eight years averages half the purchase in inventory, which adds about 12,432. The all-in cost of the LTB is then roughly 51,282, before storage conditions, shelf-life management or the cost of capital beyond the holding rate. If a redesign would cost 85,000 in engineering and requalification, the LTB wins by a wide margin on this single part; if the same part were used across a volume five times higher with a longer horizon, the balance could flip. Run the comparison, do not guess.

What the simple formula hides

Several effects distort real LTB decisions. Demand is rarely flat: a product in decline needs fewer parts, and a product that wins a large contract needs many more, so run scenarios rather than a point estimate. Shelf life matters: moisture-sensitive packages, tantalum or electrolytic capacitors, batteries, and parts with solderability limits degrade in storage, and manufacturers specify storage conditions and re-baking or re-tinning rules. Date codes and lot traceability must be preserved for regulated sectors. Finally, a deterministic uncertainty term understates fat-tailed risk; if a single large customer order can double a year’s demand, use a scenario table instead of a normal approximation.

There is also a cash dimension. An LTB converts a stream of small purchases into one large, early expense, and finance will want it expensed against the product’s future margin. Present the LTB with the net present value of the avoided redesign, and with an explicit write-off risk if the product ends early. Where a supplier allows it, a negotiated buffer arrangement (the supplier holds stock or wafer die on your behalf and releases against orders) converts the capital burden into a service fee and often beats buying everything upfront.

Integrating Alerts into PLM: From Notice to ECO

The point of the earlier architecture is that nothing should depend on a person remembering to check a portal. In practice this means an event flow from the lifecycle feed to the PLM part master, an automatic risk calculation, an alert to a named engineer, and a pre-populated change request.

Figure 4: The alerting sequence turns a status change on an MPN into an ECR and, after approval, an ECO with effectivity.

The sequence has five participants: the lifecycle feed, the PLM part master, the component engineer, change management and ERP procurement. A status change triggers a where-used and score calculation in PLM, which alerts the engineer with the affected assemblies. The engineer opens an ECR with options, change management approves an LTB quantity, ERP records the purchase order and stock position back into PLM, and the ECO is released with effectivity.

Mapping to ECR, ECO and effectivity

An ECR is a proposal; an ECO is the approved, executable change; an ECN distributes it to stakeholders. For obsolescence, the ECR should include the notice reference, last-order and last-ship dates, the options considered and their costs, and the recommendation. If the decision is a substitute swap, the ECO changes the BOM and typically uses effectivity (by date, serial number or lot) to cut over cleanly while consuming remaining stock of the old part. If the decision is an LTB, the ECO may simply annotate the part as “bought out” with the reserved quantity, and set a stock-watch alert. Our deeper treatment of effectivity and the digital thread is in the engineering change management guide.

The EBOM-to-MBOM gap

Engineering BOMs describe design intent, while manufacturing BOMs add plant-specific items, alternates and process materials. Obsolescence tracking has to cover both, because a manufacturing-only consumable can stop a line as effectively as a design part. An alternate approved on the EBOM may not yet be loaded into the plant’s MBOM or the contract manufacturer’s system. For the structural reasons behind this, see our article on EBOM-to-MBOM transformation. In practice, include the MBOM in the where-used scope, and treat the ERP stock position as an input to the LTB calculation so that you do not buy stock you already own in a plant you forgot.

Integration mechanics

Two integration styles dominate. A scheduled batch pull is simple and adequate if the provider refreshes daily and your SLA is a business day. An event-driven webhook gives lower latency but requires idempotency (a repeated event must not open a duplicate ECR) and a dead-letter queue for failed matches. Whichever you pick, log every unmatched MPN to a queue that a person reviews: the unmatched list is itself a BOM-health metric, because a part the feed cannot find is a part you cannot monitor.

OSLC and similar linked-data interfaces help when the lifecycle tool, the PLM system and the ERP are from different vendors, since the ECR can hold a link to the notice and the LTB purchase order rather than copies that drift. Keep the PLM as the system of record for the decision and the ERP as the system of record for stock and purchase orders.

Counterfeit Risk and Long-Lifecycle Sectors

Obsolescence creates the market for counterfeit parts. When an authorised channel is exhausted and a product must keep shipping, buyers turn to brokers and independent distributors, where remarked, recycled or re-sold parts circulate. The defence response is codified: SAE AS5553 covers counterfeit electronic parts avoidance, detection, mitigation and disposition, AS6081 addresses distributors, and DFARS 252.246-7007 and the related 252.246-7008 impose detection and avoidance system requirements on certain US defence contractors (verify applicability to your contract; SAE says following STD0016A helps meet the obsolescence requirements within those documents).

Practical controls are well understood. Buy from the original manufacturer or its authorised distributors wherever possible, and require full traceability back to the manufacturer when you cannot. For unavoidable broker purchases, specify inspection and test levels in the purchase order: visual and marking inspection, X-ray or decapsulation sampling, electrical testing against the datasheet, and a documented quarantine and disposition path. Do not skip these for low-cost parts; a failed 50-cent part in a safety-critical product is not cheap. Record the supply channel in the part history so that later field failures can be traced.

Why long-lifecycle sectors are different

Industrial, medical, aerospace, rail and energy products share three constraints. Their lifecycles are long, often well beyond a decade. Their change cost is high because any change drags verification, validation or certification behind it. And their volumes are low relative to consumer electronics, so they have little leverage with suppliers who optimise for the largest customers. Medical devices face design-change controls; aerospace faces configuration control and airworthiness requirements; rail and industrial safety products face functional-safety assessment. The effect is that the cost of substitution is dominated by qualification, not by the part. That is why LTB remains a legitimate and common strategy for these sectors, and why the quantity model deserves care.

Another feature of these sectors is the need to manage the data as long as the product lives. Retain the BOM snapshot, the approved-parts decisions and the LTB reservations as part of the product’s configuration baseline, so that a field repair in year twelve can find out what was actually built and what stock exists.

Trade-offs, Gotchas, and What Goes Wrong

The most common failure is false confidence in a green dashboard. A BOM-health score is only as good as the match rate beneath it. If 12 percent of MPNs fail to match the lifecycle provider, the “healthy” label silently covers a blind spot. Report the matched fraction beside every score, and treat unmatched high-value parts as high-risk until proven otherwise.

The second failure is conflating forecast with fact. Estimated end-of-life years from a provider are model output. A part may be flagged “ten years remaining” and then be discontinued in two because the supplier is acquired. Reserve high confidence for manufacturer-stated dates, and keep a risk floor on sole-source parts regardless of the forecast.

Third, over-buying. An LTB is irreversible and often paid for from the same budget as the next product. Teams buy for the optimistic demand case, then leave a pallet of expired-shelf-life parts in a warehouse. Use scenario ranges, limit the buy to the committed support horizon, and revisit the unused remainder yearly. Conversely, under-buying leaves the program exposed to a second, more expensive scramble when the stock runs out.

Fourth, ignoring the non-electronic BOM. Connectors, displays, batteries, plastics resins, adhesives, firmware libraries and even compiler tool chains go obsolete. Scoring only integrated circuits leaves the mechanical and software bill unmanaged. Fifth, process drift: a model nobody owns decays. Assign an owner, schedule a quarterly weight review, and track results such as the share of PDNs caught before the last-order date and the number of line-down events caused by supply.

Finally, beware of local optimisation. A component engineer who qualifies an alternate for one product may forget the nine others that use the same part. The where-used step should surface every consumer so that a single resolution covers them all, and so that an LTB is sized on combined demand rather than on whichever product raised the flag.

Practical Recommendations

Start small and instrument everything. Pick one product line with a long life and a realistic BOM, load its approved manufacturer parts into a clean part master, and subscribe it to a lifecycle feed. Measure match rate first; fix MPN hygiene before you tune any scoring weights. Then implement the five-factor score, publish the tier thresholds, and wire the high tier into your ECR workflow with a default owner and a one-day target.

Define your decision policy before the crisis. Write down who may approve an LTB, up to what value, what service level you are buying to, and which options must be considered. Keep the LTB model in version control, with the assumptions you used, so that an auditor or a successor can reconstruct the logic. At the same time, push strategic measures into design reviews: require a lifecycle check on every new part, prefer parts with multiple sources, and isolate volatile silicon behind modules or abstraction layers where the product’s life justifies it.

Finally, connect the loop to the factory and the field. Feed the ERP stock position back into PLM, review reserved LTB stock annually, and archive the decisions with the product baseline.

Normalise MPNs and report the matched fraction of the BOM.
Score every part on lifecycle, sources, trend, substitute difficulty and criticality.
Auto-open an ECR for high-tier parts with where-used and LTB comparison attached.
Qualify alternates for medium-tier parts before a notice arrives.
Size LTBs with scenarios, shelf life and cost of carry, not a single point estimate.
Restrict broker purchases and require inspection and traceability for any grey-market buy.
Review weights, thresholds and overrides every quarter.

Frequently Asked Questions

What is component obsolescence management?

Component obsolescence management is the process of identifying parts that suppliers are discontinuing or may discontinue, assessing the impact on your products, and choosing a resolution before supply runs out. Resolutions include a last-time buy, a form-fit-function substitute, a redesign, or a negotiated supplier arrangement. It combines lifecycle data, a risk-scoring model, PLM where-used analysis and change management, and is defined formally in standards such as IEC 62402.

What does DMSMS stand for and who uses it?

DMSMS stands for Diminishing Manufacturing Sources and Material Shortages. The US Department of Defense uses it for the loss or impending loss of manufacturers or suppliers of items, raw materials or software. The Defense Standardization Program Office maintains the SD-22 guidebook and the related resources. Commercial industries use the same ideas under names such as obsolescence management, and the vocabulary is common across aerospace, medical and industrial equipment makers.

How much notice does a manufacturer have to give before discontinuing a part?

For solid-state devices, JEDEC JESD48C recommends at least six months from the notice to the last order date and at least twelve months to the last shipment. These are industry minimums for suppliers that follow the standard, not legal guarantees, and some vendors publish longer commitments. Always read the specific notice and the supplier’s own discontinuance policy, because the dates in the notice are the ones that bind your purchase.

How do you calculate a last-time-buy quantity?

Multiply annual demand by the support horizon, then adjust for production scrap, field-service spares and demand uncertainty, subtract stock on hand and on order, and round up to the minimum order multiple. For example, 1,200 units per year over eight years, with 3 percent scrap, 5 percent spares, a 95 percent service level and 2,000 units on hand, gives about 9,250 units. Run scenarios and include shelf-life and carrying costs before approving.

What is BOM health?

BOM health is a measure of how exposed a bill of materials is to supply risk. It rolls part-level lifecycle risk scores up to an assembly or product, often reporting the worst-part score and the percentage of lines in the high-risk tier. A trustworthy view also reports the match rate, meaning how many manufacturer part numbers were found in the lifecycle data, because unmatched parts are invisible risk.

Should you ever buy obsolete parts from brokers?

Only as a last resort, and with controls. Broker and independent-distributor channels are where counterfeit and remarked parts concentrate. Standards such as SAE AS5553 and AS6081 describe avoidance, detection and disposition practices. Require traceability, specify inspection and electrical testing in the purchase order, quarantine incoming lots until they pass, and record the supply channel in the part history so field failures can be traced back.