Second Sources That Work: A Verification Order You Can Actually Run

Sourcing & channels · Published 2026-08-22 · Updated 2026-08-22

"Alternate available" in a BOM has no binding power. Rank parts by position, then verify parametrics, footprint, software, samples, pilot runs and supply commitments in order.

The cost of a second source is not the price gap between two parts. It is engineering hours and the window in which a change can still be injected into production. Writing "alternate available" on a BOM line binds nobody: it does not say how far the substitution reaches, nor what evidence closes it. A workable policy has two layers — rank every line by the position the part occupies, then run six gates in a fixed order: worst-case parametrics, footprint, software, board level, pilot run, supply commitment. Ranking decides whether the hours are worth spending; the order decides how many times you redo them. Substitutions that skip a gate surface at the pilot batch or the first production run.

Rank the BOM before hunting for alternates

Fewer than half the lines in a typical BOM deserve a real second source, yet nearly all critical lines need a supply answer in advance. Ranking spends verification hours where a stockout damages the product, not where it only annoys purchasing.

The first group cannot be swapped casually: references, isolation and gate drivers, power devices, safety-rated and automotive parts. Their failure mode is a change in system behaviour, not a parameter drift — permitted delay in an isolator and dead-time compensation in a driver both rewrite the reliability of the power stage. Parts from Novo Sense and Texas Instruments sit in those positions. The second group is structurally swappable but must be retested: general MCUs, memory and interface devices, where the pinout lines up and the differences hide in peripherals and firmware. The third is low risk: resistors, capacitors, standard logic, commodity connectors — match rating and landing pattern. The fourth gets written into BOMs by mistake: dedicated ASICs, private protocols, parts with custom firmware. There is no second source for them, only a second capacity commitment; the honest conclusion is to lock capacity, lot and change policy instead.

Six gates, and the sequence is the point

The order is not tidiness. Each gate can invalidate the one before it, and skipping one always pushes the cost to the last step.

1. Worst-case parametrics, not headline typicals: full temperature range, supply range and lot spread. Read drift and noise on references, permitted delay on isolators, dead time on drivers, thermal resistance on power devices.
2. Package, pad and stencil. The same package name is not the same pad; body length, creepage or aperture changes force a new reflow profile and solder volume.
3. Software and register compatibility. Count peripheral instances, walk the clock tree, check device IDs, power-on defaults and the programming flow. Pin-compatible on an MCU has never meant firmware runs unchanged.
4. Board and system testing. EMC, thermal drift, ageing and real load conditions. A part passing alone is not the assembly passing.
5. Pilot run and process window. Build a batch and read placement yield, rework rate and lot consistency, not whether two samples light up.
6. Supply commitment and change policy. EOL and PCN notice periods, lot traceability, and whether fab and assembly site sit inside the same written promise.

Split the gates across different channels

Samples and small quantities belong to catalogue distribution, where authorisation and lot records are legible and a candidate set lands on one bench. A house such as DigiKey, with tens of millions of SKUs on file, is useful for breadth; Mouser keeps candidates refreshed through 1100-plus authorised brands. Urgent top-up belongs to in-house stock: the ICNAV record for LCSC is 810,000 self-stocked SKUs with dispatch in as little as four hours from Shenzhen, Dongguan and Suzhou — a real advantage for gate five, a real limit for the imported long tail. Price comparison across sources belongs to BOM aggregation: ICKey quotes whole BOMs against multiple suppliers and its substitution lookup generates candidates fast. A candidate list is an input to gate one, never a verdict for gate six.

What auto-kitting quietly decides for you

The classic mistake with BOM tools is reading "fully kitted" as "interchangeable". A kitting engine optimises completeness and price. It will silently drop another part into a reference, isolator or driver position, and the design becomes a different design without anyone signing for it. Freeze those lines in the exported file. The second half of the risk sits in the promise layer: for new imported and automotive parts, platform stock is not capacity. Stock describes a warehouse today, not lot consistency two years out or who must notify you of a change — only the manufacturer or an authorised chain answers that in writing.

Notes for buyers

• Finish the ranking first: label every line prohibited, retest-required, swappable or unswappable, and open no substitution project without that table.
• Give each critical position one owner and a signed test record for the six gates.
• Compare vendors under one shared test condition; isolation delay and short-circuit withstand are measured differently across datasheets.
• Samples from catalogues, top-up from in-house stock, pricing from BOM platforms, commitment from manufacturer or agent; never ask one channel all four questions.
• For the fourth class, sign capacity instead of chasing alternates: EOL notice period, last-time-buy window, who carries safety stock.
• Freeze critical part numbers before running any kitting tool; tool output is a candidate, not an approval.

Data caveats

SKU counts, dispatch times, locations and brand-authorisation figures follow the ICNAV company database, from company statements and public material; not live quotes. The ranking model and the six gates are editorial method drawn from this site's distribution and channel category, not measured from orders. Not procurement or investment advice.

Written by ICNAV 编辑部 · ICNAV editorial team. References: the ICNAV company directory and public industry sources.

More articles

• Ordering Through a Shortage: Four Negotiable Variables Behind Lead Time
• Silicon or SiC: Where the Break-Even Line for Power Devices Actually Sits
• In SoC Selection, the Ecosystem Decides Before the Compute Does
• Controllers, Dies and Modules: Three Roles Inside the Embedded Storage Supply Chain