# BOM cost estimation before your first prototype

> A practical approach to BOM cost estimation for connected devices: grouping parts, price breaks, hidden costs and contingency that hold up.

- Author: Milo Perović, Founder & lead engineer
- Published: 2026-11-21
- Topic: Process

Most founders ask what their device will cost to build long before there is a schematic to price. That is a fair question, and a rough answer is worth having early, but BOM cost estimation done badly gives a number that is wrong in a predictable direction: too low. This post is for founders and product managers who need a figure they can put in a business plan without it falling apart at the first production quote.

We build these estimates during the architecture stage of nearly every connected-device project. The method below is how we keep them useful from the first block diagram through to the pilot batch.

## Why early estimates come out too low

The first estimate is usually a list of the obvious parts: microcontroller, radio module, sensor, battery, enclosure. Someone looks up single prices on a distributor site, adds them up and calls it the BOM. The result often misses a third or more of the real unit cost.

The reasons are consistent from project to project:

- **The long tail is forgotten.** A small board can easily carry 60–150 placements. Passives, connectors, ESD diodes, regulators, crystals and test points are cheap individually and meaningful together.

- **The wrong quantity is priced.** Prices at 10,000 units get quoted for a run of 500, or single-unit prices get used for a volume plan.

- **Non-component costs are left off.** Bare PCB, assembly, programming, testing, packaging and freight are not parts, but they are all part of what each unit costs.

- **Nothing is reserved for change.** The design will move. A sensor gets swapped, a second regulator appears, the battery grows.

## Structure the BOM before you price it

Before looking up a single price, we split the bill of materials into groups that behave differently. Each group has its own cost drivers and its own uncertainty, so lumping them together hides where the risk is.

- **Key silicon:** MCU or SoC, radio module, main sensors, power management ICs. Few lines, most of the electronics cost, and the most exposed to lead times and allocation.

- **Supporting electronics:** regulators, protection, crystals, connectors, LEDs, buttons. Moderate cost, mostly multi-sourced.

- **Passives:** resistors, capacitors, ferrites. Fractions of a cent to a few cents each at volume, but numerous.

- **Electromechanical and power:** battery, antenna if external, cables, display, motors or buzzers.

- **Mechanical:** enclosure parts, gaskets, light pipes, screws, labels.

- **Packaging and accessories:** box, insert, quick-start card, charging cable.

At the architecture stage the key silicon and the mechanical parts can usually be priced with reasonable confidence. The supporting electronics and passives are estimated from experience of similar boards. That is fine, as long as the estimate says which lines are quoted and which are guessed.

## BOM cost estimation at the right quantity

Component pricing is stepped. Distributors publish price breaks, and the same part can cost several times more at quantity one than on a full reel. Passives show the steepest curve; a module or a large MCU flattens out sooner. The quantity you price at should match the batch you actually intend to build, not the volume you hope to reach in year three.

We normally price three scenarios side by side, because the decisions they drive are different:

ScenarioTypical quantityWhat it is used for

Prototype batch5–50 unitsBudgeting the engineering phase; unit cost here is not a product cost

Pilot run100–1,000 unitsFirst customers, early pricing, cash-flow planning

Production target1,000–10,000+ unitsGross margin and retail price decisions

For the production column, distributor list prices are a starting point rather than an answer. At real volumes a contract manufacturer or a franchised distributor quote will often come in lower, but you should not count on that saving until you have it in writing. We keep the published price in the estimate and note the expected reduction separately.

## The costs that never appear on the parts list

A connected device is not a bag of components. The unit cost a business plan needs is the landed cost of a finished, tested, boxed unit. The lines most often missing are these:

- **Bare PCB:** driven by layer count, board size, panel utilisation and finish. A [4-layer stackup](https://miloperovic.com/journal/choosing-a-4-layer-pcb-stackup-for-connected-devices/) costs more per board than 2 layers, and the gap narrows with volume.

- **Assembly:** per-placement cost, plus setup and stencil charges that dominate small runs. Through-hole parts and double-sided placement add cost.

- **Programming and test:** flashing firmware, calibration and end-of-line checks take fixture time on every unit.

- **Yield loss:** some units fail and get reworked or scrapped. Early runs need a larger allowance than mature ones.

- **Final assembly and packaging:** fitting the board into the enclosure, labelling, boxing.

- **Freight, duty and handling:** often several percent of unit cost, more for heavy or battery-containing shipments.

One-off costs belong in a separate column: injection mould tooling, test fixtures, certification and the engineering itself. Amortise them over a stated quantity so it is obvious how unit economics change if that quantity is not reached.

## Contingency and confidence levels

Every line in our estimate carries a confidence tag: quoted, catalogue price, or estimated by analogy. The contingency then depends on the mix rather than being a single flat figure added at the end.

As a rough guide, an architecture-stage estimate with most lines estimated deserves a contingency in the region of 20–30%. Once the schematic is complete and every line has a catalogue price, 10–15% is more reasonable. After a contract manufacturer has quoted the full build, a few percent covers price movement. These are working ranges, not rules; a design built around one scarce part deserves more caution than one using common, multi-sourced components.

We also flag single-source parts in the estimate. A part with one supplier is a price risk as well as an availability risk, and it is cheaper to address while the schematic is still open. Our notes on [component sourcing for small production runs](https://miloperovic.com/journal/component-sourcing-for-small-hardware-production-runs/) go into that side in more detail.

## Keeping the estimate alive as the design changes

An estimate that is produced once and filed is out of date within weeks. We treat it as a living document tied to the design, updated at each milestone: architecture, schematic freeze, layout complete, first build and pilot quote.

A few habits make this cheap to maintain:

- Keep the costed BOM generated from the same source as the schematic BOM, so parts cannot drift apart.

- Record the price, quantity break, supplier and date for every line.

- Log every change that moves unit cost by more than a small threshold, with the reason. "Added load switch for sensor power gating" explains a cost increase better than a revised total.

- Compare the target cost against the estimate at every design review, not only at the end.

That last point matters most. If the product needs to retail at a certain price, the target unit cost should be written into the requirements from day one, alongside battery life and size. It then becomes a design constraint engineers can trade against, rather than a disappointment discovered at the pilot quote.

## What we would do before committing to a number

If you need a cost figure for a pitch deck or a pricing decision, this is the short version of our process:

- Split the BOM into the groups above and price key silicon and mechanics properly first.

- Price at the quantity of the batch you will actually build next, and show the other scenarios alongside.

- Add PCB, assembly, test, yield, packaging and freight as explicit lines.

- Keep one-off costs separate and amortise them over a stated volume.

- Tag every line by confidence and size the contingency to match.

- Set a target unit cost in the requirements and review against it at each milestone.

Done this way, the first estimate is still uncertain, but it is honest about where the uncertainty sits. That is what lets a founder make pricing and funding decisions early without being caught out when the first real quote arrives.
