Introducing a New Product Without Losing Control

There is a specific moment in a hardware business where everything changes, and most founders go through it without noticing until afterwards. It is the point where you stop making one of something to prove it works, and start making them because people are paying for them.

Up to that point, undocumented knowledge is an advantage. Two or three people hold the design in their heads, change it daily, and move fast. After that point, the same thing becomes the central risk in the business. The part that worked was built by someone who is now on another project, from a drawing that has been superseded twice, using a supplier chosen because they answered the phone.

Nothing dramatic goes wrong at first. It goes wrong on the fourth batch, when a customer asks why this one is different.

Control the design before you scale the volume

The first thing to put in place is not a quality manual. It is knowing, with certainty, what you are making.

That means a drawing or model set with issue numbers that actually change when the design changes, a bill of materials that is maintained rather than rebuilt from memory each time, and a record of which configuration each unit was built to. If you ship serial numbers 1 to 20 and they are not all the same, you need to know how they differ and why.

This sounds like bureaucracy to a twelve-person company. It is the opposite. The expensive version is reconstructing the build standard of a unit that has failed at a customer site eight months later, from emails and somebody's memory. I have watched businesses spend weeks on that.

Change control: light, but real

Early-stage businesses change designs constantly and should. The discipline is not to stop changing — it is to know what changed, when it took effect, and what it affects.

A workable version for a small team is a single log: what changed, why, who approved it, which serial numbers it applies from, and what it affects downstream — tooling, test, documentation, spares, anything already in the field. One page per change. Fifteen minutes.

The test of whether your change control is real is simple. Pick a unit you shipped three months ago and ask what build standard it was. If the answer takes more than a few minutes to establish, it is not real yet.

Suppliers, before you are dependent on them

Early suppliers get chosen on responsiveness and willingness to take a small order, which is reasonable. The problem is that this choice silently becomes permanent, and by the time it matters you have no leverage and no alternative.

Before you scale, decide which parts are critical — the ones where a failure reaches the customer, or where a single source could stop your line — and do something deliberate about them. That means knowing whether the supplier has a management system, whether they can provide material certification and traceability, whether their process is actually repeatable, and whether a second source exists if you need one.

You do not need a procurement department. You need to know which three suppliers could stop your business, and what you would do.

Prove the first one properly

Even if no customer requires a formal first article inspection, do the work. Build one unit, measure it against the drawing in full, record the results, and keep the evidence with the build record.

It achieves three things. It finds the features your process cannot actually hold before you have made two hundred. It gives you a baseline to argue from when a customer queries conformance. And when a prime or an investor asks how you control quality, you have an answer made of evidence rather than intention.

Customer audits arrive earlier than founders expect

This is the one that catches people out. A startup selling into aerospace, defence, rail or medical will be audited by a customer far sooner than it expects — often before there is any revenue worth defending.

The audit is rarely about sophistication. It is about whether you can show what you did: who built it, to what standard, from what material, checked by whom, and what happened when something was wrong. A company that can produce that from a handful of controlled records passes comfortably. A company with excellent engineering and no records does not, regardless of how good the product is.

Our post on investor-ready governance and controls covers the funding side of the same question.

Do not import a corporate stage gate

The overcorrection is as damaging as the gap. I have seen twelve-person companies adopt a stage-gate process borrowed from a prime, with phase reviews, sign-off matrices and documentation sets that nobody has the hours to complete. It gets abandoned within two quarters, and the abandonment teaches everyone that process is theatre.

Pick the smallest version that answers the questions that actually matter: is the design defined and frozen enough to build, can we make it repeatably, do we know what it costs, can we prove what we shipped, and what happens when it comes back. Four or five decision points, each with a short written answer. Grow it only when something goes wrong that a heavier process would have caught.

Know the cost before you price it

The prototype cost is not the product cost, and founders routinely price off the former. The real number includes scrap and rework at the yield you are actually achieving, inspection and test time, the setup cost spread over a realistic batch size, packaging, and the engineering time still being spent on each unit.

Work it out honestly before the price is set with a customer, because resetting a price afterwards is a conversation nobody enjoys and it is usually the point where a growing hardware business discovers it has been buying revenue.

Frequently asked questions

When should we put proper controls in?

Before the first repeat order, or before the first customer audit, whichever comes first. Retrofitting configuration control and traceability onto units already in the field is far more expensive than doing it from the start.

Do we need ISO 9001 or AS9100 at this stage?

Usually not immediately — but the controls underneath them, yes. Many startups build the practices first and certify when a customer requires it. Building the habits early makes certification later a matter of formalising what you already do rather than inventing it under time pressure.

How much of this can two people realistically maintain?

More than you would think, if it is proportionate. A controlled drawing set, a maintained bill of materials, a one-page change log and a build record per unit is perhaps an hour a week at low volume. The same information reconstructed after a failure is weeks.

What if the design is still changing every week?

Then you are not yet at the point of scaling, and that is fine. The controls matter from the moment you are shipping units someone else depends on. What you must avoid is crossing that line without noticing.

We help early-stage engineering and hardware businesses put in the governance and controls that investors and primes expect, at a weight the business can carry. See our startup structures work or get in touch.