I have watched several hardware companies fail from close enough to see why, and the pattern is consistent enough to describe. It is rarely the product idea.
The cash flow shape is brutal
The single largest structural difference from software.
You pay for tooling, components and manufacturing months before you receive any revenue. Minimum order quantities mean you buy far more than you need for the first run.
Then you hold inventory, which is capital sitting in a warehouse, until it sells.
A software business with growing demand needs more servers. A hardware business with growing demand needs more working capital, and the faster it grows the more capital it consumes.
Which is why hardware companies fail while succeeding, and why the phrase about growing yourself into bankruptcy applies here more than anywhere.
Mistakes cannot be patched
The second structural difference and the one that catches software people entering the sector.
A bug in shipped hardware is in the hands of customers permanently. Fixing it means a revision, new tooling, and a decision about what to do for existing owners.
Which means the testing burden before shipping is enormous, and the temptation to ship and iterate — which is correct practice in software — is catastrophic here.
The companies I have seen do this well spend a long time on prototypes and pilot runs. The ones that fail treat the first production run as a test.
The certification burden
Consistently underestimated in every plan I have seen.
Anything with a radio, a mains connection, a battery, or an intended use with children faces regulatory approval, and the requirements differ by market.
Testing is expensive, takes months, and failure means a redesign followed by re-testing.
Battery shipping regulations alone catch people out, since lithium cells are restricted cargo with documentation requirements.
The practical effect is that certification frequently determines the launch date rather than engineering, and a plan that treats it as a formality at the end is a plan that will slip by a quarter.
Manufacturing is a relationship, not a transaction
The thing that separates the companies that ship from the ones that do not.
A small order from an unknown company is not a priority for a factory. Quality, timelines and communication all reflect that.
Companies that succeed generally have somebody who has done it before, or somebody physically present, or both.
Design for manufacture is a genuine discipline. A design that works as a prototype may be expensive or impossible to produce at volume, and finding that out after tooling has been cut is a catastrophic and common failure.
Support becomes the business
The part nobody plans for.
Physical products break, arrive damaged, and are used in ways nobody anticipated. Returns, warranty claims, spare parts and repairs become a substantial operation.
The cost of handling a return frequently exceeds the margin on the sale, which means a return rate that seems small can eliminate profitability entirely.
And customers reasonably expect support for the life of a physical object, which is much longer than the life of a typical startup.
The crowdfunding trap
Worth addressing since it is how many of these begin.
Crowdfunding provides capital before manufacturing, which addresses the cash flow problem, and it does so at a fixed price set before costs are known.
Component prices move, tooling costs more than estimated, certification finds problems, and shipping costs rise. The price was fixed a year earlier.
Which is why a large proportion of successfully funded hardware campaigns deliver late, deliver something different, or do not deliver.
Usually not fraud. Usually an underestimate of every cost, locked in before any of them were known.
What the successful ones did
From the small number I have seen work.
Started with a smaller, simpler product than they wanted to make.
Sold at a price with enough margin to absorb being wrong about costs, which generally meant a higher price than felt comfortable.
Had somebody with manufacturing experience involved from the beginning rather than hired later.
Planned certification into the schedule from the start.
And raised more money than the plan required, on the assumption that the plan was optimistic, which it always was.
The distribution question
One further obstacle that surprises people.
Selling directly online gives you the margin and requires you to build the demand yourself.
Retail gives access to customers and takes a substantial share of the price, demands terms on returns and payment timing, and frequently requires packaging and compliance work specific to that channel.
Many hardware companies discover that retail economics do not work at their price point only after committing to it, because the margin structure was calculated on direct sales.
Working the retail arithmetic backwards from the shelf price, early, changes what the product needs to cost to make.
Prototyping cost versus production cost
The gap that catches almost everybody.
A prototype built with rapid manufacturing techniques and off-the-shelf parts can be produced for a modest sum, which produces a misleading sense of the unit cost.
Production at volume requires tooling, which is a large fixed cost, and delivers a much lower unit cost across a run.
Which means the prototype is expensive per unit and cheap in total, and production is the reverse, and plans built on prototype economics are wrong in both directions simultaneously.
Getting a manufacturing quote for realistic volumes early, before the design is fixed, is the single most useful thing a hardware founder can do.