Manufacturing Tolerances: How Precise Does It Need to Be?

Last Tuesday, a buyer in Frankfurt lost $47,000.

His plastic housings arrived. Looked fine. Until assembly.

Nothing fit.

The screw holes were 0.3mm too small. The snap-fit tabs were 0.5mm too thick. His German assembly line stopped cold for three weeks while he scrambled for replacements.

The factory? They shrugged. “Sample was good, no?”

Yeah. The sample was good. Because someone hand-filed it for two hours before shipping.

Welcome to the tolerance trap. Where the difference between “close enough” and “garbage” is thinner than a human hair. And where most buyers don’t even know what numbers to put on their drawings.

What Your Factory Says vs What They Mean

Let’s get this straight first. Here’s the translation guide nobody gives you:

I’ve watched this game for six years. The factory always says yes during the quote. Then production starts and suddenly physics exists again.

The Anatomy of a Tolerance Disaster

Walk into any mid-tier injection molding shop in Shenzhen.

You’ll see shiny machines. Clean floors. Maybe even a CMM in the corner collecting dust.

Now ask to see the mold. Really look at it.

See those wear marks on the core? That’s where your tolerance went. Every cycle, the steel erodes a tiny bit. After 50,000 shots, your “±0.1mm” part is now ±0.3mm. But nobody’s measuring. Nobody’s adjusting.

I once took a caliper to a “good” batch.

First piece: 24.98mm. Perfect.

Middle piece: 25.14mm. Drifting.

Last piece: 25.31mm. Trash.

The factory wanted to ship all of it. “Still works, right?”

No. It doesn’t work. It jams. It breaks. It costs you a customer.

We sent them back. The factory boss got mad. Said we were too picky. Six months later, he called asking for more orders. I told him to get his molds serviced first.

How Tight Is Too Tight?

Here’s where buyers kill themselves.

They see blueprints from their German or Japanese competitors. Everything’s ±0.02mm. So they copy it.

Bad move.

Those tolerances need expensive machines. Skilled operators. Climate-controlled workshops. You want that in China? Fine. Pay triple.

Most products don’t need aerospace precision. A plastic toy? ±0.2mm is plenty. A phone case? ±0.15mm works. A simple bracket? ±0.3mm is fine.

But.

If parts need to fit together? If there’s an assembly? That’s where you tighten up. Mating surfaces, screw bosses, alignment pins—these need real control.

The trick is knowing where to spend your precision budget.

• Critical dimensions: Anything that touches another part. ±0.05 to ±0.1mm depending on complexity.

• Functional features: Clips, latches, hinges. ±0.1 to ±0.15mm or they’ll snap or rattle.

• Cosmetic surfaces: ±0.2mm is fine. Nobody’s measuring your logo depth with calipers.

• Hidden internal stuff: ±0.3mm. Save your money.

I learned this the hard way on a speaker grill order.

The buyer wanted every hole ±0.05mm. I told him it was pointless. He insisted. The quote tripled. Lead time doubled. And you know what? The end user never noticed.

Six months later, he came back. “Can we relax the tolerance?” Yeah. Now you’re learning.

The Material Matters More Than You Think

Plastic shrinks.

Not a lot. But enough to ruin your day.

ABS shrinks about 0.5%. PC shrinks 0.6%. PP shrinks up to 2% depending on the grade. Your factory molds the part. It cools. It shrinks. Now your 100mm part is 99mm.

Good factories know this. They compensate in the mold design. Bad factories just shrug and ship it anyway.

Metal’s worse in some ways. Aluminum expands in heat. Steel warps if the heat treatment is rushed. I’ve seen CNC parts come out perfect, then twist like a pretzel in the anodizing tank.

And don’t get me started on die-cast zinc.

It looks great. Feels solid. Then six months later, it’s cracking because some genius used recycled scrap with lead contamination. Your tight ±0.08mm tolerance? Gone. The part’s now deformed junk.

This is why we push for material verification before production. Not just a datasheet. Real testing. You want virgin ABS? Prove it. Show me the raw pellet supplier invoice. Show me the burn test. Show me the melt flow index.

Sounds paranoid? Last year, a factory swapped out the agreed PC material for cheap PC/ABS blend. Saved them $0.40 per kilo. Cost the buyer $80,000 in returns because the parts warped in summer heat.

The Inspection Nobody Does

You know what I bring to every factory audit?

A digital caliper. A micrometer. A pin gauge set. And a really bad attitude.

Because I’m going to measure stuff. Not the golden sample sitting on the manager’s desk. Real production pieces. Random ones. From the middle of a box.

Most buyers never do this.

They show up. Factory gives them tea. Shows them a nice sample. Everyone smiles. Deal done.

Then the shipment arrives and it’s all out of spec.

“But the sample was perfect!”

Yeah. That one was. What about the other 9,999?

A good QC inspection doesn’t just check appearance. It checks dimensions. At least five critical ones. With calibrated tools. Against the drawing.

We caught a batch last month where the wall thickness was 1.8mm instead of 2.5mm. Factory saved 30% on material. Parts broke during drop tests. If we hadn’t measured, the buyer would’ve shipped them to Target and had a recall.

The factory? Acted shocked. “Our QC checked!”

Their QC was a 19-year-old kid with a plastic ruler.

What You Actually Need to Specify

Stop sending drawings that just say “tolerance ±0.1mm” at the bottom.

That’s lazy. And it leads to fights.

Instead, mark your critical dimensions. Use geometric dimensioning and tolerancing if you know how. If you don’t, at least call out what matters:

• Hole diameter for your M3 screws? ±0.05mm.

• Overall length that nobody cares about? ±0.3mm.

• Wall thickness for strength? ±0.2mm max, or it’ll fail testing.

• Surface finish? Ra 1.6 for visible parts, Ra 3.2 for hidden stuff.

And put it in writing. In the purchase order. In the contract. Because when the factory ships junk, they’ll claim “you never said.”

Yeah. I did. It’s on page three. In red. With your signature next to it.

The Tool Wear Trap

Here’s something nobody tells you.

That first article inspection? Perfect. Every dimension on spec.

Production run number one? Great.

Production run number five? Starting to drift.

Production run number ten? Garbage.

Because the mold is wearing out. The cutting tools are dulling. The machine’s spindle has play in it.

And the factory isn’t maintaining anything.

You need to build in periodic checks. First article, mid-run, last article. At minimum. Better yet, random sampling every 500 or 1,000 pieces.

Costs you a few hundred bucks in inspection fees. Saves you tens of thousands in rework.

I’ve got a client who learned this lesson hard. Ordered 50,000 brackets. Approved the first samples. Skipped all the in-process checks to save money. Shipment arrived. Half were out of tolerance. Couldn’t use them. Had to scrap them. Cost? $35,000.

Now he pays for inspections.

Your Move

Tighter than ±0.05mm on plastic? You’re wasting money.

Looser than ±0.3mm on mating parts? You’re asking for trouble.

Send your drawings. Mark what matters. And for the love of efficiency, measure the actual production pieces before you wire the final payment.

One caliper. Five minutes. That’s all it takes.