Most founders find out their electronic product costs too much only after the first real quote comes back. The instinct is to attack the biggest line on the bill of materials — usually the microcontroller or the display — and negotiate it down by a dollar. That almost never closes the gap, because the expensive parts of an electronic product are rarely the parts you think about. The money hides in connectors nobody specified carefully, in test minutes per unit, in yield loss, and in decisions made in the first two weeks of design that quietly set the floor for everything after.
Start by seeing the whole cost, not the BOM
Before cutting anything, build an honest picture of the loaded cost per unit. A useful breakdown for a mid-complexity electronic product looks roughly like this:
| Cost bucket | Typical share | Where it hides |
|---|---|---|
| Semiconductors and passives | 30–50% | Over-specified MCU, redundant regulators |
| Bare PCB | 5–15% | Layer count, board size, via technology |
| Assembly and rework | 10–20% | Unique part count, mixed technology, hand work |
| Electromechanical and cables | 10–25% | Connectors, harnesses, hardware |
| Test, programming, packaging | 5–15% | Seconds per unit multiplied by volume |
| Scrap, yield loss, freight | 3–10% | Nobody's line item until it appears |
These proportions shift with product type, but the lesson holds: better than half the cost sits outside the semiconductor lines. If you have never built the full picture, start with a proper bill of materials that includes every consumable and every process step, then reconcile it against a full COGS breakdown including freight and duty.
The five hiding places
1. The processor you picked in week one
Engineers pick a microcontroller early, usually the one on the dev kit that was on the shelf. That choice sets the memory footprint, the peripheral set, the support components, and often the RTOS. Products routinely ship on a part with four times the flash and RAM they use, at three times the price of an adequate one. Doing the sizing work — how much flash and RAM the firmware actually needs, and which peripherals are genuinely used — before committing frequently frees $1.50–$4 per unit, and there is a second, larger version of the same decision in whether you need a module at all: a system-on-module makes sense at low volume and becomes expensive at high volume, where a custom board pays back its NRE.
2. Connectors, cables, and the things between boards
A board-to-board connector pair at $0.80, a $3 wire harness, and four screws sound trivial next to a processor. They are not: connectors are usually the highest-cost-per-function items on the BOM, and every harness is a manual operation with a defect rate. Eliminating one interconnect by combining two boards, or replacing a harness with a flexible circuit section, often saves more than any component substitution — a reason to look hard at how the harness is designed and assembled.
3. Layer count and board geometry
Going from four layers to six can add 30–50% to the bare board price, and going from a standard panel-friendly outline to an awkward shape wastes panel area you pay for anyway. Fine-pitch parts that force laser-drilled microvias raise the price again. Board area is real money: a 20% shrink is roughly a 20% board saving and often a smaller enclosure too. Understanding what each additional layer costs and buys lets you decide whether the extra plane is a requirement or a convenience.
4. Test and programming seconds
Every unit spends time on a fixture being programmed, calibrated, and functionally tested. At a loaded line rate, 90 seconds per unit is real money at volume, and a test that fails intermittently costs far more because operators retest and eventually pass marginal units. A well-designed production test fixture with proper test points and a fast programming path is one of the highest-return investments in the whole program. The same is true of matching the electrical test strategy to your volume — flying probe is cheap to set up and slow per board; in-circuit test is the opposite.
5. Yield and rework
A 3% first-pass failure rate that requires manual rework can cost more per unit than several BOM lines combined, because rework is skilled labor plus retest plus the risk of a field failure. Most yield problems trace back to designs that were never reviewed for manufacturability: tombstoning-prone footprints, parts too close to the board edge, insufficient thermal relief. Running a real DFM review before the first production build is not a formality.
A cost-reduction sequence that works
- Get real quotes at real volumes. Cost conversations based on distributor single-piece pricing are fiction. Quote at your actual annual usage.
- Rank by total line cost, not unit price. Forty passives at $0.01 with four different values is a bigger target than one $2 part, because consolidating values reduces feeder setups and inventory.
- Consolidate part numbers. Standardizing on fewer resistor and capacitor values across the board reduces line changeovers and improves purchasing leverage.
- Question every subsystem's existence. The cheapest circuit is the one you deleted. Does the product need both a display and an app? Does it need a rechargeable pack, or would primary cells and a longer duty cycle serve the user better?
- Attack the enclosure and the box together. Electronics cost reduction that ignores the plastic and the packaging leaves half the money on the table.
- Re-quote after each change. Cost reduction that is never verified against a new quote is a spreadsheet exercise.
This is the same discipline as a formal value engineering pass, and it works best as a scheduled activity between the engineering build and the production build, not as a panic response to a bad quote.
Costs that are not worth cutting
Some savings buy a much larger bill later. Undersized protection components — the TVS diodes and series resistance that keep a product alive through an electrostatic discharge event — save cents and cost returns. So does a marginal power supply, a connector rated below its actual mating cycles, or a shortcut that risks failing certification and forcing a respin. Certification failures are the most expensive kind of cheap: a second round of testing plus a board revision plus schedule is typically a five-figure event. Design margin, protection, and testability are the three places to spend; everything else is negotiable.
Projects House takes electronic products apart line by line — BOM, layout, test time, and assembly steps — and rebuilds the cost model with the savings that survive production. If your quote came back higher than your price point, send us the design through the contact form.