Ask a founder why their hardware is late and you will usually hear a number: six months, nine months, a year. Ask what the factory said, and the answer is almost always the same — they’re slow.
Factories are not fast. But they are rarely the reason a program loses half a year. The time usually disappears before the purchase order, in a loop that looks like work and produces very little:
- A spec is sent to five suppliers, four of whom are trading companies.
- Quotes come back that are not comparable, because each one assumes a different BOM.
- One quote is chosen on price. The BOM turns out to contain parts nobody can source.
- Samples arrive. The enclosure is fine; the board is not; the firmware team assumed a different connector.
- A redesign happens at the worst possible moment — after tooling, before certification.
None of those steps is a manufacturing failure. They are decisions made without enough information, sequenced badly.
Latency is the product
A good factory can turn a clean package in weeks. A clean package is the hard part: a BOM with sourced and alternated parts, a drawing set that reflects the actual design, a test plan that matches the failure modes, and a person on the founder’s side who knows which questions are load-bearing.
The industry calls this NPI. It is less a phase than a translation layer. Every hour spent in that layer is an hour not spent waiting for a line to free up.
The uncomfortable implication: most of the schedule is under the founder’s control, and most founders do not know it. They optimise the thing they can see — the factory — and leave the invisible loop unmanaged.
What closes the loop
Three things, in order of leverage.
A map of who is real. The difference between a factory and a trading company is not visible on a website. It is visible in whether the person quoting you can tell you which line the board will run on, and what happens when a part goes on allocation.
The vocabulary to review the package. If a founder cannot read a DFM report, the factory’s engineering team learns to stop sending them. Silence is the most expensive thing a supplier can give you.
Physical context. There is no substitute for standing on a line and watching a board go through reflow. It recalibrates every estimate in the founder’s head, permanently, in about four minutes.
That last one is why I run a factory tour. Ten-plus visits in six days does not make anyone a manufacturing engineer. It does something narrower and more useful: it replaces a founder’s imagination of a factory with an actual one, and it makes the next six months of correspondence dramatically shorter.
The part that does not scale
This is a relationship business wearing an industrial costume. The founders who move fastest are not the ones with the best deck or the lowest quote. They are the ones whose factory owner will pick up the phone on a Saturday because they have stood in the same room.
That is not romanticism. It is the mechanism by which exceptions get made — and hardware is a business of exceptions.