BOM Cost Reduction for EMS Companies and Hardware Startups: Counting Net Savings

The cheapest quote does not always make the cheapest product. A lower-priced part can add inspection, programming, failures or a redesign. A volume discount can leave a startup holding stock after the next board revision. A smaller package can save board area and then cost more in assembly and rework. A useful cost-reduction programme measures the net effect across design, sourcing and production, and counts a saving only once it shows up in conforming builds.

The best time to do most of this is before the design, suppliers and validation evidence are frozen for mass production. At that stage architecture, part choices and PCB rules can still change cheaply. After launch, the same change may need requalification, field-compatibility work and a plan for obsolete stock. This guide covers both windows, and it is written for the two parties who usually share the work: the OEM or startup that owns the product, and the EMS company that builds it.

Define the cost you are trying to reduce

Direct material is one layer. A product cost model that supports good decisions separates component price; PCB fabrication and assembly; programming, test and inspection; tooling and non-recurring engineering; freight, insurance and customs; MOQ and pack exposure; scrap, rework and yield loss; financing and inventory holding; warranty and field service; and the cost of obsolescence or redesign.

NIST's manufacturing extension programme points out that purchasing teams judged on quoted price can push total cost of ownership up when the operational effects are ignored. The comparison that matters is the cost of a conforming, delivered and supportable product.

Agree one financial convention for every proposal: currency and exchange-rate date, annual volume, amortisation period, tax treatment, and whether a saving will be reported as price variance, cost avoidance or margin improvement. Without it, two teams can claim the same rupee twice.

Freeze a baseline you can defend

Fix the product revision, approved parts, suppliers, volume scenarios and quote dates. Reconcile quantities across fitted variants. If the baseline mixes old and new revisions, every later savings claim is open to argument.

For each line, work out:

• Annual usage = quantity per product × annual product volume
• Annual material cost = annual usage × current landed unit cost
• Excess exposure = the larger of zero and (order quantity − requirement), × unit cost

Use production-intent volumes. Prototype catalogue prices exaggerate high-volume opportunities, while an optimistic forecast can justify a price break that turns into surplus stock. For the few largest items, build a should-cost range (material, process, tooling, test, packaging, logistics and supplier margin) rather than a single falsely precise target. A should-cost model is good at raising design and quotation questions. It should not be used to force a capable supplier below a price it can sustain.

Rank before brainstorming

Sort lines and process steps by annual extended cost. A minority of items usually carries most of the spend. Run three views side by side: highest material spend, highest supply or obsolescence risk, and highest manufacturing or yield burden. Add a column for "cost to change" and another for "validation impact".

This keeps engineers from spending a week replacing cheap passives while the display, battery, enclosure or processor sets the product's economics. It also catches the low-value line that keeps stopping the line, which can deserve attention for reasons a spend chart never shows.

Challenge the requirement before the part

Expensive parts often trace back to assumptions nobody has tested: interfaces the firmware never uses, memory sized for a roadmap rather than the shipping code, sensors with more precision than the application needs, connectors rated for far more mating cycles than the product will see, display brightness beyond real use, power margin stacked at several design levels, or tolerances tighter than the mechanism requires.

Remove margin with evidence, not opinion: worst-case electrical analysis, tolerance stacks, thermal models, load profiles and test data. NASA's EEE-INST-002 sets baseline criteria for the selection, screening, qualification and derating of parts on its space-flight projects. Those rules are far stricter than a commercial product needs, but the principle carries over. A cheaper, lower-rated part that runs close to its limits is not a saving.

Simplify the architecture

Architecture changes often beat line-by-line negotiation. Ask whether you can combine processing, connectivity or power functions; remove a redundant regulator or level shifter; replace a duplicate sensor with firmware; merge memory devices; cut connectors with board-to-board integration; standardise voltage rails; or share a module across a product family. Would an integrated circuit beat a pre-certified module at your volume, or would certification and RF work make the module cheaper overall?

Every consolidation concentrates risk. A highly integrated IC may save placements and area but increase single-source exposure. Record the requirement that justified the original part and the evidence that allows its removal.

Standardise parts and qualify alternates

Reusing proven resistor and capacitor values, packages, connectors, regulators and protection devices across the portfolio pools volume, shortens the feeder list and widens the choice of alternates. Build equivalence groups from controlled parameters, not from catalogue descriptions. For a capacitor that might mean capacitance, tolerance, voltage, dielectric, package, temperature range, ageing and DC-bias behaviour. Engineering approves each exact alternate; the guide to cross-referencing alternate components covers the checks.

Lifecycle status belongs in the same review. TI's lifecycle categories run from Preview and Active through Not Recommended for New Designs and Last Time Buy to Obsolete, and a discounted part that TI already marks as not recommended for new designs can turn into a redesign by the next build. IEC 62402:2019 treats obsolescence management as a cost-effective process across the whole life of an item, including strategies to minimise obsolescence during design.

Work with the fabricator and the EMS on DFM

PCB and assembly cost often reflects design choices more than supplier margin. Review with a production-intent fabricator and your EMS before layout release:

• Can board dimensions improve panel utilisation?
• Are blind, buried or microvias really needed?
• Can track, space and drill sizes move into standard capability?
• Is the chosen laminate needed for temperature, frequency or reliability?
• Can controlled-impedance scope be reduced safely?
• Is the surface finish right for the shelf life and assembly process?
• Can hand-soldered parts become SMT, and can package orientations be made consistent?
• Are there test points for in-circuit or functional test, and can programming run in parallel?

IPC describes DFM as a necessary step that is often neglected or skipped, and its DFM profiles combine the IPC-A-610 performance classes with the producibility levels of IPC-2221. Check which revisions apply: IPC's revision table lists IPC-2221C (December 2023) and the IPC-2231A DFX guidelines (October 2021) as current (checked October 2026).

Ask suppliers to explain what drives the cost before you ask them to lower the quote. A stack-up or spacing change repeats its saving on every build; a price concession may last one contract. Log each DFM finding with an owner and a validation impact, so it does not become an uncontrolled design change.

Bring sourcing in, without importing risk

Requote high-spend lines on one scope and one delivery basis (see comparing component quotes on equal terms), then work the cost drivers: annual commitment, forecast visibility, pack, schedule, payment and NCNR terms. Negotiating with distributors and managing MOQ go into those levers.

An EMS provider can pool common parts across customers for better pack use; a startup can pool releases across variants. Pooling needs approved MPNs, allocation rules, traceability and clear ownership of excess, and no substitution across programmes without engineering sign-off.

Keep authorised sources as the normal baseline. ECIA's view is that buying only from trusted sources is the best guard against counterfeit parts. A grey-market saving has to carry authentication, lost warranty, uncertain storage and failure risk; for many active parts it will not survive the sum.

Localise by capability, not by invoice

KPMG's June 2026 note on India's EMS opportunity says 60 to 90 per cent of the bill of materials is sourced from outside the country across semiconductors, displays, PCBs and passive components. That is a real reason to develop Indian suppliers, but a local invoice is not the same as local manufacturing, and neither guarantees a lower landed cost. For each proposal, check where the part is actually made, which subcomponents are imported, capacity, process capability, quality history and continuity plans, and include qualification, yield ramp and the cost of running two sources during the switch.

Count yield and test

A material saving disappears fast if first-pass yield drops. Rank defects by cost as well as frequency, and measure cost per good unit (total build cost ÷ accepted units).

An illustrative case: a change adds ₹6 of material per board but cuts the failure rate needing rework from 4% to 1%. If each rework costs ₹300, expected rework per board falls from ₹12 (4% × ₹300) to ₹3 (1% × ₹300). That is ₹9 saved for ₹6 spent, a net ₹3 per board. The reverse also holds: a cheaper part that raises placement or test failures is not a saving.

Map each test to the defect it catches before removing anything. Boundary scan, programming during panel test, parallel fixtures and data-driven limits can cut seconds per unit, but a faster test that lets more faults escape costs more in the field.

Write the business case on net, not gross

For every idea, record:

• Gross annual saving = (baseline recurring cost − proposed recurring cost) × annual volume
• Net saving in a period = recurring saving on units actually built in that period − qualification − tooling − scrap − stock written off − transition cost

Here is why the distinction matters, with illustrative numbers. A change saves ₹14 per unit on a product selling 40,000 units a year, a gross annual saving of ₹5.6 lakh. Qualification costs ₹1.2 lakh, a new stencil and fixture ₹40,000, and 3,000 old parts worth ₹55 each (₹1.65 lakh) must be written off: ₹3.25 lakh of one-time cost. If the change only takes effect for the last six months of the year, year one sees 20,000 units × ₹14 = ₹2.8 lakh of saving, and a net result of −₹45,000. Year two brings the full ₹5.6 lakh. The idea is worth doing if the product will sell for well over a year, but it is not a ₹5.6 lakh saving this year.

The approval record should name the affected item and revision, the change and its reason, baseline and proposed cost, one-time costs and old-stock disposition, validation and compliance impact, lifecycle and source risk, owner and effectivity, and the measured result after release.

Run it through gates

Keep ideas, approved changes and realised savings in separate columns. Most programmes run a sequence like this:

Validation should match the risk of the change: bench and tolerance checks, thermal review, EMC pre-compliance, reliability or regulatory testing, then a pilot build on production-intent tooling, firmware and materials. The ECO should define the first affected lot, what happens to old stock and when to roll back. The guide to moving from prototype to production covers what changes at that hand-over.

Who owns what between OEM and EMS

The OEM or startup owns requirements, technical approval, compliance and the business trade-offs. The EMS owns transparent manufacturing assumptions, DFM feedback, supplier management within the contract and accurate yield and cost data. Agree in advance how savings and qualification costs will be shared, require written approval for every substitution, and reward realised savings on conforming output rather than quote reductions. A regular value-engineering review between the two keeps the list moving.

Suppliers who can help with DFM or a second source can be found in forUdyog's PCB and PCBA directory.