What actually counts as a second source
The test is simple and most dual-source arrangements fail it: if the disruption you are insuring against occurred, would the second source still be able to ship?
| Arrangement | Independent of | Still exposed to | Real resilience |
|---|---|---|---|
| Two distributors, same mill | Commercial failure of one distributor | Mill capacity, country risk, export licensing, alloy supply | Almost none |
| Two mills, same country | Single-facility events — fire, capacity, quality hold | Country-level export controls, tariffs, regional disruption | Moderate |
| Two mills, different countries | Country-level regulatory and tariff events | Upstream alloy and rare earth concentration | Good |
| Two mills, different countries, different alloy supply | Most single-point failures | Systemic rare earth market events | Strong — and expensive |
| Domestic stock plus an overseas mill | Border and transit events on the buffer quantity | Everything beyond the buffer duration | Good for the covered window |
Verify the mill, not the vendor
Two distributors quoting the same part at similar prices are frequently sourcing from the same facility — the magnet manufacturing base is more concentrated than the vendor base. The check takes one email: ask each supplier which facility presses and sinters the part, by name and address, and compare the mill named on the material certificates you already hold. If they match, you have duplicated your purchasing overhead and insured against nothing.
The three things that make magnets harder than most components
Deciding which parts justify it
Second-sourcing every magnet part number is neither affordable nor necessary. Score the portfolio and spend the qualification budget where interruption actually hurts.
| Factor | Raises priority when | Lowers priority when |
|---|---|---|
| Consequence of stockout | Stops a production line or a shippable product | Buffered by inventory or substitutable |
| Grade | SH, UH, EH, AH or SmCo — export-licence exposed | Standard N-series, widely produced |
| Geometry | Custom tooled, unusual, or tight tolerance | Close to a catalogue size available from many sources |
| Qualification burden | Low — industrial part, simple acceptance | High — full PPAP with customer approval required |
| Volume | High enough to interest a second mill | Below a viable MOQ at a second source |
| Program life | Years of remaining production | Approaching end of life — a last-time buy is cheaper |
The cheaper alternative, where it fits
For many parts, an inventory buffer is a better purchase than a second source. Qualifying a second mill costs tooling, samples, engineering time and often a customer approval cycle, and it delivers nothing until it is used. Holding an extra ten weeks of cover costs carrying charges and delivers immediately. Second-sourcing wins where the risk is long-duration — export controls, capacity loss, a supplier exiting — because no realistic buffer covers those. Buffering wins where the risk is a delivery slip. Most portfolios need both, on different parts. See inventory programs for buffer sizing.
The qualification protocol
Six stages. The discipline that matters is that each gate is passed before the next begins — the common failure is jumping from samples straight to production because a shortage arrived first.
Stage 1 — Paper qualification
- Supplier audit to the depth the part warrants — see the audit checklist.
- Confirm capability specifically: this grade, this geometry, this magnetization pattern, this coating, this tolerance. Approve to a scope, not in general.
- Confirm certification scope covers the site making the part.
- Establish country of origin and export licensing exposure for the grade — a second source in the same regulatory jurisdiction may not diversify the risk you care about.
Stage 2 — Tooling
- Provide the drawing package, not a sample to copy from. Reverse-engineering an existing part perpetuates any drift already present in the incumbent’s output.
- Establish tooling ownership in writing before the die is cut.
- Expect three to six weeks for a press die and a magnetizing fixture where the pattern is not standard.
Stage 3 — First article and cross-comparison
The core technical gate. Measure incumbent and candidate parts on the same equipment, on the same day, at the same temperature. Comparing the candidate against the incumbent’s certificate rather than against incumbent hardware measured yourself is the most common methodological error here.
Stage 4 — Functional validation
- Build assemblies with candidate parts and test against the same acceptance criteria as production.
- Include thermal cycling to the worst-case service profile — irreversible loss differences between sources appear here and nowhere else.
- Verify assembly processes: press-fit forces, adhesive compatibility with the candidate’s coating, insertion behaviour.
Stage 5 — Capability
- A production-representative run on production tooling with production operators, not a hand-built lot.
- Capability indices on critical dimensions and on the magnetic acceptance characteristic.
- Gauge R&R where the candidate’s measurement system will be relied upon.
- PPAP submission where the program requires it, with customer approval obtained before release.
Stage 6 — Release and maintenance
- Update the drawing and the approved supplier list to show both sources, with the acceptance limits that apply to each.
- Allocate real volume — see the next section.
- Set the trigger conditions for requalification.
The cross-comparison matrix
What to measure on both sources, and what an acceptable difference looks like. Some variation between mills is normal and expected; the point of the exercise is to establish that it sits inside your design margin rather than inside your ignorance.
| Characteristic | Method | Typical acceptable difference | Why it matters |
|---|---|---|---|
| Total magnetic moment | Helmholtz coil and fluxmeter | Means within a few percent; candidate spread no wider than incumbent | The primary acceptance characteristic — a systematic offset shifts every downstream calibration |
| Br, Hcb, Hcj, BHmax | Per-lot material certificate | Both above grade minimums; note where each sits within the band | A source running near the minimum has less margin against thermal excursions |
| Demagnetization behaviour | B-H curve at maximum service temperature | Knee position outside the operating region for both | Two parts meeting the same room-temperature grade can behave differently hot |
| Critical dimensions | Metrology per ballooned drawing | Both within tolerance; compare where each centres in the band | Two sources centred at opposite ends of a tolerance make a mixed-lot assembly problem |
| Coating thickness per layer | XRF | Both within specification | Affects press fits, bond strength and corrosion life |
| Coating adhesion | Cross-hatch, thermal shock | Both pass | Adhesive bonding behaviour differs between platers |
| Corrosion resistance | Salt spray per ASTM B117 | Both meet the specified hours | The most common real difference between mills |
| Pattern accuracy | Rotary Hall pole scan | Pole placement and balance within the same limits | Encoder and sensor accuracy depends on it entirely |
| Mechanical integrity | Chip and crack rate over the sample | Comparable; grinding practice differs materially between mills | Drives assembly line scrap |
| Thermal exposure response | Flux measured before and after worst-case cycling | Irreversible loss comparable and within design margin | The test that catches a source running thin on coercivity |
Do not set acceptance limits from the incumbent alone
Limits derived only from the incumbent’s output encode that supplier’s process centring as though it were a requirement. A candidate that is entirely conforming then fails, and a legitimate second source is rejected for being different rather than for being unacceptable.
Set limits from what the design needs, established by analysis and validated against both sources. If that exercise reveals the design has been relying on the incumbent running consistently above the minimum, that is an important finding about your design margin — and it is better to learn it during a qualification than during a shortage.
What triggers requalification
Qualification is a state that decays. These are the events that should reopen it, and the change-control clause in your supply agreement is what makes sure you hear about them.
| Change | Requalification scope | Rationale |
|---|---|---|
| Different manufacturing site | Full | Effectively a new source, whoever owns it |
| Alloy supplier change | Material properties and thermal validation | Composition and grain structure shift |
| Grade or composition change | Full magnetic and thermal | Includes reduced-Dy reformulations — see alternatives |
| Plating line or chemistry change | Coating thickness, adhesion, corrosion, and any bonded-joint validation | Adhesive behaviour and corrosion life both change |
| New or repaired tooling | Dimensional and magnetic first article | Dimensional shift and pattern change |
| Magnetizing fixture change | Magnetic and pattern verification | Saturation level and pole placement |
| Extended production gap | Abbreviated — first article on resumption | Process drift and personnel turnover |
| Significant nonconformance | Scoped to the failure mode | Confirms corrective action actually worked |
The clause that makes this work
None of the above helps if changes arrive silently. The supply agreement should require written notification before any change to manufacturing site, alloy source, grade composition, coating process, tooling or magnetizing method, with an agreed notice period and your right to requalify before the changed parts ship. Without it, the first indication of a mill change is a batch that measures differently, discovered by whatever sampling plan happens to be running. The full term sheet is in commercial terms.
Running two sources in production
A qualified second source that never ships decays into a name on a list. Keeping it viable requires deliberate allocation and a little discipline about lot control.
Allocation
| Split | Effect | Suits |
|---|---|---|
| 100 / 0 with a dormant approval | Cheapest; qualification decays and the second source has no tooling warm | Low-risk parts where the approval is a formality |
| 90 / 10 | Keeps tooling exercised and the relationship live at minimal cost | The practical default for most critical parts |
| 70 / 30 | Second source can scale quickly; meaningful pricing tension | High-volume, high-consequence parts |
| 50 / 50 | Maximum resilience; loses volume pricing at both and doubles quality oversight | Rarely justified outside safety-critical supply |
Ten percent is usually the right number
A minority allocation large enough to require a real production run once or twice a year keeps the tooling exercised, the operators familiar, the account commercially interesting and the qualification current — at a small premium over single-source pricing. It also gives you continuous evidence that the second source still performs, rather than an assumption you will test for the first time during an emergency.
Lot control across sources
- Never commingle sources in a bin. If a magnetic problem appears, source identity is the first thing an investigation needs, and it is unrecoverable once lots are mixed.
- Consider distinguishing the parts where it is practical — a marking, a packaging colour, a distinct internal part number suffix. Sintered magnets from two mills are visually identical.
- Track acceptance data by source. Two sources trending differently is useful information; blended data conceals it.
- Watch mixed-lot assemblies. Where several magnets go into one assembly and matching matters — multi-pole arrays, balanced rotors — specify that all magnets in an assembly come from one source and one lot.
The scenario worth rehearsing
Your primary source is interrupted. Before you switch, confirm four things: the second source’s tooling is present and serviceable, its qualification has not been invalidated by an unnotified change, its current lead time from a standing start, and whether your customer requires notification of a source change. Every one of these is faster to establish now, on a quiet afternoon, than during the interruption — and a second source that fails any of them is not the insurance you believed you had bought.