Component sourcing for small hardware production runs
Component sourcing for small hardware runs: lifecycle checks, second sources, consignment and how to avoid counterfeit parts before you order.

Most small hardware projects do not stall on the schematic. They stall six weeks before a build, when someone finally checks stock and finds that the radio module has a long lead time, the fuel gauge is marked not recommended for new designs, and the only seller with the right regulator is a broker you have never heard of. Component sourcing is the part of product development that nobody owns until it hurts.
This post is for founders and product managers planning runs of tens to a few thousand units, where you are too small for a manufacturer to manage supply for you and too large to keep buying parts on cut tape from a distributor’s website. Here is how we approach it.
Why component sourcing belongs in the design phase
The cheapest moment to fix a sourcing problem is before the footprint is placed. Once a part is routed, swapping it means a new layout spin, new firmware drivers in some cases and another round of testing. Before layout, it is a different line in the bill of materials.
So we treat availability as a design input, alongside current draw and cost. Every part that is hard to replace gets checked when it is chosen, not when the order is placed. In practice that means a small number of parts get real scrutiny: the microcontroller or radio module, sensors, power management ICs, connectors and anything mechanical that is custom or single-vendor. Resistors and capacitors in common packages rarely cause trouble, as long as the values and ratings are ordinary.
What to check before a part goes on the schematic
For every part that matters, we look at the same handful of things on the manufacturer’s site and across the major authorised distributors:
- Lifecycle status. Active is what you want. Not recommended for new designs (NRND) means the manufacturer has already told you to look elsewhere. Anything with a last-time-buy notice is out.
- Stock across several authorised distributors. One distributor holding all the stock is a warning sign. Healthy parts are stocked in quantity by several.
- Factory lead time. Stock can vanish in a week. The factory lead time tells you how long a reorder takes once it does, and it can range from a few weeks for commodity parts to many months for specialised ICs in a tight market.
- Packaging and minimum order quantity. Some parts only ship in full reels of several thousand. That is fine for production and painful for a prototype batch.
- Package and footprint family. A part in a common package with pin-compatible alternatives from other vendors is far safer than a unique package from one supplier.
- Moisture sensitivity level. Higher MSL parts need dry storage or baking before reflow (J-STD-033 covers this), which matters if parts will sit on a shelf between builds.
None of this takes long per part. It takes discipline to do it for every critical line before the schematic is frozen.
Designing in second sources
The single most useful habit is choosing parts that have a drop-in alternative, and recording that alternative in the BOM from day one. Some categories make this easy. LDO regulators in SOT-23-5, common op-amps in SOIC-8 and many MOSFETs in standard packages have pin-compatible equivalents from several manufacturers. Passive components in 0402 or 0603 with standard tolerances are almost fully interchangeable.
Other categories do not. Microcontrollers, radio SoCs, fuel gauges and most sensors are effectively single-source, because firmware and calibration depend on the exact part. For those, the options are different:
- Choose a mainstream family. A widely used part such as an STM32L4, an nRF52840 or an ESP32-S3 is stocked heavily, has several package and memory variants and is less likely to disappear abruptly.
- Leave room for a sibling. Picking a family with a larger-memory variant in the same package gives you a fallback if the exact part number runs dry.
- Consider a certified module. A pre-certified radio module often costs more per unit but is commonly available from several distributors and saves you intentional-radiator certification work under CE RED or FCC Part 15.
- Buy ahead. For a truly irreplaceable part, buying a year of expected demand up front is often cheaper than a redesign.
Where a footprint can accept two different parts, we sometimes lay out a dual footprint. It costs a little board area and saves a spin if one part becomes unavailable.
Turnkey, consigned or a mix
When a contract manufacturer assembles your boards, there are broadly three ways to get parts onto their line. The right one depends on volume and on how much of your BOM is unusual.
| Model | How it works | Good for | Watch out for |
|---|---|---|---|
| Turnkey | The manufacturer buys every part | Simple BOMs, repeat orders | Mark-up on parts, silent substitutions if not controlled |
| Consigned | You buy and ship all parts | Tight control, unusual parts | Your team carries purchasing, logistics and shortages |
| Partial consignment | You supply critical parts, they buy the rest | Most small runs | Clear split in the BOM so nothing is bought twice |
For small runs we usually prefer partial consignment. The client or we secure the few critical parts early, and the manufacturer buys the commodity passives they already stock. Whatever the model, the BOM should state which alternates are approved, so the manufacturer cannot substitute a part you have never tested.
Plan for attrition too. Pick-and-place machines drop and waste parts, especially tiny passives, so manufacturers ask for overage on each line. The exact amount varies by part and by line, so ask your manufacturer for their figures rather than guessing.
Brokers, counterfeits and grey-market stock
When authorised channels run dry, brokers will often have stock. Sometimes it is genuine excess inventory. Sometimes it is remarked, recycled or simply the wrong part. On a small run, one bad reel can cost more than the whole parts budget once you count debugging time.
Our rule is simple: buy critical ICs from the manufacturer or its authorised distributors, and treat any broker purchase as a risk to be tested. If broker stock is unavoidable, we inspect markings and packaging, check date codes against the datasheet revision, and build and test a few boards before committing the full quantity. The same functional checks we describe for end-of-line testing on a first production batch are a good way to catch a part that is not what it claims.
Keeping the BOM alive between builds
A BOM that was healthy at prototype stage can be in trouble by the time you reorder. Between builds, we keep a few things current:
- A lifecycle and stock check on critical parts before every build, not just the first.
- Approved alternates recorded per line, with notes on what was tested.
- A record of what was actually fitted to each batch, so field issues can be traced to a lot or a substitution.
- Product change notifications from manufacturers routed to someone who will read them.
This fits naturally into the build stages we cover in EVT, DVT and PVT for small hardware teams. Each stage is a natural point to re-check availability, because each one is a point where a part change is still affordable.
What we would do before the first order
If you are a few weeks from ordering boards, here is a short checklist:
- List the ten parts that would hurt most if they disappeared, and check lifecycle, stock and lead time for each.
- Make sure every line has an approved alternate, or a written reason why it cannot.
- Decide which parts you will secure yourself and buy them now, including enough for attrition and a rework batch.
- Agree in writing with your manufacturer that no substitution happens without approval.
- Keep a copy of the exact BOM fitted to each batch.
Good component sourcing is unglamorous work. Done early, it is the difference between a build date you can promise and one you can only hope for.