Where the money actually is
Before changing anything, understand the proportions. A change that halves a cost element representing 4% of the part price is not a cost reduction program; it is a distraction.
The seven levers, ranked
| # | Lever | Typical saving | Engineering risk | Requires requalification? |
|---|---|---|---|---|
| 1 | Grade & temperature class right-sizing | 10–40% | Medium — needs a load-line check | Usually yes |
| 2 | Geometry & machining reduction | 8–25% | Low to medium | Depends on fit |
| 3 | Tolerance right-placement | 5–20% | Low | Rarely |
| 4 | Coating fit | 3–15% | Medium — environmental risk | Sometimes |
| 5 | Magnetization pattern simplification | 2–12% | Medium to high | Yes |
| 6 | Material substitution | 20–60% | High — full redesign | Yes |
| 7 | Commercial structure | 5–20% | None | No |
Start with 3 and 7
Tolerance review and commercial restructuring carry no engineering risk and no requalification burden. On a part already in production they are almost always the correct first move, and they frequently return double digits on their own. Save the grade and material work for the next design revision, where it can be validated properly rather than retrofitted.
Lever 1 — Grade and temperature class
The single largest and most common overspend in magnet purchasing. It happens because the temperature suffix looks like a safety margin and costs like a material change.
| Class | Max operating temp | HRE content | Relative cost index |
|---|---|---|---|
| N (none) | 80 °C | None | 1.00 |
| M | 100 °C | Low | 1.10–1.20 |
| H | 120 °C | Moderate | 1.20–1.35 |
| SH | 150 °C | High | 1.35–1.60 |
| UH | 180 °C | Very high | 1.60–1.90 |
| EH | 200 °C | Very high | 1.90–2.30 |
| AH | 230 °C | Maximum | 2.30+ |
Two changes are usually available, and they are independent:
Drop the temperature class where the duty allows
The correct method is a load-line analysis at the true maximum operating temperature with the actual permeance coefficient of your geometry — not a comparison of the temperature rating against your ambient. A thick magnet in a closed circuit tolerates far more than a thin one in a large air gap at the same temperature. Many parts specified SH would run indefinitely at H, and some specified H would run at M. The method is in the temperature guide.
Reconsider the energy product
Grade number and temperature class trade against each other. Higher energy product means lower coercivity, so N52 is more vulnerable to demagnetization than N42 at the same temperature. Where field requirement can be met by geometry instead of grade, a lower energy product with better coercivity is often both cheaper and more robust. Field at a working distance is driven far more by magnet dimensions than by grade — a full grade step typically buys only a few percent of field. The numbers are in the grades chart.
The 2026 argument for right-sizing has nothing to do with price
Heavy-rare-earth-bearing grades — SH, UH, EH, AH and all SmCo — fall under Chinese export licensing, which adds roughly nine weeks of regulatory review to the schedule and exposes the part to a control regime that can tighten without notice. Dropping from SH to H where the thermal duty permits removes cost, lead time and geopolitical risk in one change. See lead times and MOQs.
Lever 2 — Geometry and machining
Sintered magnet material is brittle and must be ground with diamond tooling. Every ground surface is a separate operation with its own setup, cycle time and scrap risk, and pressed geometry is nearly free by comparison.
| Feature | Cost effect | Cheaper approach |
|---|---|---|
| All six faces ground | High — multiple setups | Grind only the functional faces; leave non-critical surfaces as-sintered |
| Chamfers and radii on every edge | Moderate — extra operation | Specify a break edge only where handling or assembly requires it; state a general edge condition |
| Through holes and counterbores | High — drilling brittle material has real scrap | Press the hole where geometry allows; or capture the magnet mechanically instead of fastening through it — see assembly and retention design |
| Thin walls and thin sections | High — handling breakage and yield loss | Respect practical minimums; ring wall thickness below about 1.5 mm drives scrap sharply |
| Extreme aspect ratios | High — pressing and sintering distortion | Moderate the ratio, or build from stacked segments |
| Non-standard size, marginally different from stock | Very high — full custom cost for no function | Snap to a stocked size where the design tolerates it; check available inventory before finalising |
| Oversized magnet from a conservative field estimate | High — material is the biggest cost element | Calculate field at the actual working point rather than assuming; every cubic millimetre removed is rare earth removed |
The highest-yield question in any drawing review
“Is there a stocked size within the functional envelope?” A 12.0 × 3.0 mm disc that could have been a stocked 12.7 × 3.2 mm carries custom tooling, a custom MOQ and a ten-week lead time for a dimensional difference no part of the assembly can detect. This is the most common and most expensive avoidable decision in magnet specification, and it is almost always made without anyone realising a choice was being made.
Lever 3 — Tolerance right-placement
Note the framing: right-placement, not loosening. The objective is to move precision onto the dimensions that carry function and take it off the ones that do not. Applied blindly, a tight tolerance on every dimension signals to the supplier that none of them were analysed.
| Tolerance | Process implication | Relative cost |
|---|---|---|
| As-sintered (±1–2%) | No grinding operation | 1.0× |
| ±0.10 mm | Standard grinding | 1.1–1.2× |
| ±0.05 mm | Precision grinding, tighter process control | 1.3–1.5× |
| ±0.02 mm | Lapping or specialist grinding, higher scrap | 1.8–2.5× |
Where tolerance usually matters
- The air gap dimension in a motor or sensor — it sets field at the working point directly.
- The press-fit or bond-line dimension that controls retention.
- Parallelism and squareness in stacked or arrayed assemblies, where errors accumulate.
- Outside diameter after coating where the part enters a bore — and note that coating thickness is part of that stack.
Where it usually does not
- Non-functional lengths and widths with clearance on both sides.
- Dimensions consumed by an adhesive gap — the bond line absorbs the variation by design.
- Surface finish on faces that are neither bonded nor sealing.
- Edge conditions away from handling and assembly interfaces.
State the basis, not just the number
A tolerance without a stated basis creates disputes that cost more than the tolerance saved. Specify whether dimensions are before or after coating, which datums apply, and which dimension governs if two conflict. The full treatment, including acceptance criteria and workmanship limits for chips on a brittle material, is in tolerances and acceptance criteria.
Lever 4 — Coating fit
Coating is a smaller cost element than material or machining, but it is where specifications drift upward most easily — a premium coating gets specified once for a demanding application and then propagates across a part family that does not need it.
| Coating | Relative cost | Right choice when |
|---|---|---|
| Ni-Cu-Ni (triple nickel) | 1.0× | Indoor, dry or incidental moisture — the correct default for most industrial applications |
| Zinc | 0.8–0.9× | Cost-sensitive, benign environment, cosmetic appearance not critical |
| Epoxy over nickel | 1.2–1.5× | Sustained humidity, outdoor exposure, condensing service |
| Parylene | 2.5–5× | Conformal pinhole-free barrier genuinely required — implantable, chemical, ultra-clean |
| Gold over nickel | 3–8× | Biocompatibility or a contact-resistance requirement that nothing else meets |
The one place not to economise
Under-specifying coating is the most expensive mistake on this page, because it fails in the field rather than at the dock. Uncoated or under-protected NdFeB in a humid environment corrodes from the grain boundaries outward — the magnet swells, cracks and loses coercivity, and by the time it is visible the assembly is scrap. The saving is a few percent of part cost; the failure is a warranty campaign. Match the coating to the real environment, including washdown, condensation cycling and any solvent exposure during assembly.
Two adjustments that usually are available
- Drop cosmetic coatings on hidden parts. Black nickel and decorative finishes cost more than standard nickel and add nothing to a magnet inside a housing.
- Right-size the salt spray requirement. An inherited 500-hour salt spray specification on a part that lives inside a sealed indoor enclosure eliminates the cheapest qualifying coatings for no reason. Specify against the real exposure.
Lever 5 — Magnetization pattern
Magnetization is inexpensive when the pattern is simple and a genuine cost driver when it is not, because complex patterns need dedicated fixtures that must be designed, wound and amortized.
| Pattern | Fixture requirement | Relative cost |
|---|---|---|
| Axial | Standard solenoid | 1.0× |
| Diametric | Standard, with orientation control | 1.0–1.2× |
| Multipole | Custom multi-tooth fixture per pole count and diameter | 1.3–2.0× plus tooling |
| True radial, one piece | Custom radial fixture | 1.5–2.5× plus tooling |
| Clocked to a mechanical feature | Fixturing plus verification per part | Adds inspection cost |
- Match the pole count to an existing fixture where the sensor allows it. A 16-pole ring using a fixture that already exists is materially cheaper than a 14-pole ring that needs a new one, and most encoder resolutions can be met either way.
- Do not specify clocking unless the application needs it. A pole-axis-to-feature tolerance forces per-part fixtured verification. Where the assembly can be aligned electrically or set at calibration, it costs nothing to leave the magnet unclocked.
- Understand what true radial actually buys. It is genuinely required for uniform field through 360° and for eliminating periodic signal error from segment joints. It is not required for a holding application. The distinction is in magnetization directions.
- Consider magnetizing at assembly. Shipping unmagnetized removes air-freight field restrictions, reduces handling damage and simplifies packaging — at the cost of owning a magnetizing station.
Lever 6 — Material substitution
The largest available saving and the one requiring the most engineering. This is a redesign, not a substitution — a like-for-like swap will not work, because the materials differ by an order of magnitude in energy product.
| Material | Relative cost per unit energy | Realistic substitution candidates |
|---|---|---|
| Sintered NdFeB | baseline | — |
| Ceramic / ferrite | 0.1–0.3× | Holding, latching, separation, speakers, low-duty motors — where volume and mass are available |
| Bonded NdFeB | 0.7–1.2× | Complex geometries, fine pole pitch, thin walls, over-moulded assemblies |
| Alnico | 1–3× | Very high temperature, long-term stability, instrument applications |
| SmCo | 3–7× | Not a cost reduction — a performance requirement above 200 °C |
The ferrite case, made honestly
Ferrite is roughly an order of magnitude cheaper per unit of energy, contains no rare earths, is immune to the export licensing regime, needs no coating, and classifies under a different tariff subheading. The cost is size and mass: matching an NdFeB magnet’s field requires substantially more volume, and the assembly has to accommodate it. Where the envelope allows — and in holding, latching, separation and many low-duty motor applications it does — this is the largest single saving available anywhere in magnet procurement. Where it does not, forcing it produces a worse product. Run the field calculation before assuming either way.
The trap in the other direction
Replacing a ferrite magnet with a same-size neodymium one because it is “stronger” overloads the circuit — saturating sensors, overwhelming latch mechanisms, and changing the mechanical forces the assembly was designed around. The correct substitution is smaller in NdFeB, sized by a load-line analysis of the original circuit.
Lever 7 — Commercial structure
No engineering risk, no requalification, and available on parts already in production. This is the lever to pull first while the design changes work their way through validation.
- Consolidate part numbers across programs. Three similar discs used by three product lines often collapse into one, which triples the volume on a single tooled size and eliminates two MOQs, two inventory positions and two qualification records.
- Commit volume to get volume pricing. A blanket order against an annual quantity prices at the annual tier while delivering monthly — see inventory programs.
- Buy against the market, not against the shortage. Rare earth input pricing moves independently of your demand. Buying only when stock runs out guarantees purchasing at whatever the market happens to be doing.
- Right-size the packaging. Individually blistered parts on a component that goes straight into a bowl feeder is pure cost. Specify packaging against the receiving process.
- Review the Incoterm. DDP is convenient and carries a risk premium; taking on the import yourself can be cheaper if you have the capability. Detail in tariffs and country of origin.
- Reduce documentation to what is required. Full PPAP on a non-automotive, non-safety part adds cost to every lot forever. Match the document set to the actual requirement — see certificates and compliance documents.
- Question inherited specifications on legacy parts. Requirements written for a previous application often survive into new programs unexamined. The most expensive line on many drawings is there because nobody has asked why since 2009.
A worked example
A sensor target magnet in an industrial position sensor, drawn conservatively by a designer working without cost feedback. Nothing about the original is wrong — it is simply specified as though every requirement were critical.
| Specification line | As drawn | After review | Effect |
|---|---|---|---|
| Grade | N45SH | N42H | −18% |
| Size | Ø10.0 × 4.0 mm custom | Ø9.53 × 3.18 mm stocked | −12%, tooling removed |
| Tolerance, all dimensions | ±0.02 mm | ±0.05 mm OD, ±0.10 mm thickness | −9% |
| Coating | Parylene C | Ni-Cu-Ni | −7% |
| Magnetization | Diametric, clocked ±1° | Diametric, unclocked | −4% |
| Order structure | Quarterly spot | Annual blanket, quarterly release | −11% |
| Combined | — | — | ≈−45% |
| Lead time | 12–14 weeks | Days, from stock | Export licence exposure removed |
The reasoning behind each change: the sensor runs at 95 °C maximum, so H class carries adequate margin on a load-line check and SH was precautionary. The stocked imperial size sits inside the mechanical envelope. Only the outside diameter enters a bore, so it alone needs precision. The sensor is inside a sealed enclosure, making Parylene an inherited requirement from a washdown application. The controller performs an electrical zero at commissioning, so mechanical clocking was never used.
What made this possible
Not a negotiation. A conversation between the design engineer, the buyer and the supplier’s applications engineer, held with the drawing open and the real operating conditions on the table. That conversation is available on any part — it just has to be scheduled before the tooling is cut, not after.
What not to cut
Each of these reliably costs more than it saves.
| False economy | What it looks like | What it actually costs |
|---|---|---|
| Under-specifying coating | Zinc where epoxy is needed; saves a few percent | Field corrosion, warranty exposure, assembly scrap |
| Cutting temperature class without a load-line check | Dropping SH to N on the rating alone | Irreversible flux loss in service, drifting calibration, no recovery |
| Removing acceptance testing | Dropping the helmholtz moment limit to save inspection cost | Lot-to-lot variation reaches your line undetected — see how magnets are tested |
| Buying on unit price from an unqualified source | A quote materially below the market | Grade substitution, missing certificates, origin exposure, no recourse |
| Eliminating safety stock on a long-lead part | Inventory reduction target met | Air freight, expedite premiums, line stoppages — typically several times the carrying cost saved |
| Skipping first-article verification | Saves two weeks on launch | The failure is discovered in production instead of on a sample |
| Single-sourcing to maximise volume leverage | Best possible unit price | No alternative when the single mill has an interruption |
The test
A legitimate cost reduction removes something the application does not need. A false economy removes something it does need and defers the cost to a stage where it is larger and harder to fix. If the change cannot be justified against the actual operating requirement in one sentence, it belongs in the second category.