You Didn't Fall in Love With Your Product. You Fell in Love With One Lucky Unit.

If you've read the Boundaries vs. Targets post on this blog, you already know that a spec is a window, not a number. That idea comes up constantly in cost and schedule conversations. But the same logic applies somewhere most founders don't think to look: the physical feel of their product.

And when it bites you there, it shows up in user reviews and return rates, not engineering reports.

The lab is lying to you

Not on purpose. But when you build prototypes by hand, you tune them. Parts get selected and fitted. Tolerances get massaged. The result is a unit where everything lands near the middle of its acceptable range. It feels great. It should. You built it to feel great.

That unit becomes the benchmark. Leadership signs off. The bar is set.

What you've actually done is validate one point in a very large design space. Your factory doesn't ship that point. It ships everything inside the tolerance windows you specified, distributed across thousands of units according to your supplier's process capability. Some customers get the middle. Some get the edges. A few get the corners.

If you've never asked what the product feels like at the corners, you haven't finished designing it.

The stack compounds

Every part in an assembly has its own tolerance window. When those parts come together, the variation compounds. A button that feels perfect when every part lands at nominal dimensions can feel cheap or stiff or loose when parts all land at the “legal” limits of their specs simultaneously.

Both assemblies are in spec. Only one matches what you intended.

This isn't a quality failure. Your CM isn't dropping the ball. The design simply never accounted for what it would feel like across the full range it was legally allowed to produce.

Model it before you commit

Before any mechanical interface a customer will feel is locked for production, ask one question: what does this feel like at worst-case compliant?

Your supplier's process capability data gives you what you need. It tells you how their parts distribute across the tolerance window. With that, you can estimate what percentage of assembled units will fall inside your intended experience range. That's not a mechanical number. It's a customer satisfaction number.

If 5% of your units produce an experience that's noticeably below your intent, and you're planning to ship 50,000 units, you already know how many disappointed customers you're building toward. You can model that against the cost of tightening the design now.

The ROI is not subtle

Tightening a tolerance at design freeze costs time and maybe some tooling iteration. Discovering the same problem after production tooling is cut costs returns, warranty claims, customer service load, and in some cases a full engineering change under pressure.

The question is simple: what does closing this gap cost now, versus what does leaving it open cost later?

Run the numbers early and it's rarely a close call. Founders who skip this step find out the hard way, from their customers, not from their engineering team.