How to Reduce Product Development Costs

Development budgets are rarely blown by expensive engineering. They are blown by rework — requirements that moved, prototypes built to answer questions a simulation could have answered, and design decisions locked in before anyone asked a manufacturer what they cost. Every durable saving below comes from removing a loop, not from buying cheaper hours.

This guide covers where money actually goes by phase, the six levers that reduce it, the false economies that increase total cost, and how to tell the difference before you commit.

A typical funded hardware program spends far less on the phase people worry about — engineering — than on the consequences of decisions made before it.

Shares vary by product class — connected devices carry more certification cost, moulded consumer goods more tooling. The pattern holds: the earliest and cheapest phases determine the largest ones.

The single largest avoidable cost in hardware is engineering against a target that keeps moving. Before modelling begins, write down the numbers a design either meets or fails: target landed cost, size and weight envelope, battery life or duty cycle, environmental and drop requirements, applicable certifications, and target launch volume. Anything not on that list is a preference and must be labelled as one.

Custom parts cost money three times: to design, to tool, and to qualify. Off-the-shelf and previously validated parts arrive with datasheets, supply history and known failure behaviour.

Reserve custom work for the parts that carry the differentiation. Everything else should be boring on purpose.

Every part carries a drawing, a supplier, a tolerance stack, an inspection step and an assembly operation. Removing one removes all of them. Consolidate brackets into a single moulding, use snap fits where fasteners are not required for service, integrate features into existing surfaces, and delete parts whose only job is to correct another part's tolerance. Fewer parts also shorten assembly time, which is where labour cost lives.

Prototypes are for questions that require physical reality — fit in the hand, thermal behaviour in a real enclosure, user error, drop damage. Questions about stress, deflection, mould flow, tolerance stack-up and basic thermals should be answered analytically first, and the prototype used to confirm the result.

Sequence the builds so each one answers several open questions at once. Our rapid prototyping guide covers matching prototype fidelity to the question being asked.

Blanket tight tolerances are one of the most expensive habits in mechanical design. Tightening a dimension by one class can raise part cost noticeably and cut the number of shops willing to quote. Identify the few critical-to-function dimensions, tolerance those properly, and open everything else to standard shop capability. The same applies to surface finish and cosmetic specifications, which quietly drive scrap rates.

A DFM review with the shop that will actually build the part is usually free and routinely returns changes worth more than the entire engineering fee: wall thickness adjustments, draft angles, a different fastening approach, a material substitution with the same properties and a shorter lead time. Ask for it at concept freeze and again before tooling release, not after. Our checklist for vetting manufacturing partners covers what a serious DFM response looks like.

For a phase-by-phase view of what development actually costs, see our mass production cost guide and the product discovery process that sets the requirements in the first place.

Rework caused by unstable requirements. Engineering hours are visible on an invoice, but the cost that dominates a program is redesigning work already done because a target moved or a manufacturing constraint surfaced late. Locking numeric requirements before CAD and getting DFM input before design freeze removes most of it.

Write numeric requirements before modelling, reuse off-the-shelf enclosures, modules and reference designs for everything that is not the differentiator, answer analytical questions with simulation instead of prototypes, batch prototype rounds, tolerance only the critical dimensions, and request a DFM review before the design is frozen.

Usually yes for the total, even when the hourly rate is higher. Split vendors create handoffs where drawings are reinterpreted and manufacturing feedback arrives after the design is committed. A single accountable team catches those constraints during design, when changing them is still cheap.

It depends on complexity, but the proportions are more useful than a single number: roughly 5–10% on discovery, 10–15% on concept and industrial design, 25–35% on engineering, 15–25% on prototyping and testing, 20–30% on design for manufacture and tooling, and 5–15% on certification. Connected devices skew toward certification, moulded goods toward tooling.

They save money when each build answers questions that require physical reality — fit, feel, thermals in a real enclosure, drop performance, user behaviour — and waste it when they answer questions simulation could settle for free. Batch several open questions into each build rather than printing a new revision for every small change.

Practically, once production tooling is cut. Around 70–80% of a product's unit cost is committed by the end of detailed design, so cost reduction after tooling release is limited to negotiation and small process changes. Serious cost work belongs in requirements, part-count reduction and DFM.