Why Your Injection Molding Tolerances Are Off (and It’s Probably Not the Molder)

I remember the exact moment I started caring about tolerance callouts. We had approved a tool for a plastic housing, then received 2,000 parts with every mounting boss sitting about 0.4 mm off—or maybe 0.5 mm, I don’t have the inspection report in front of me. Either way, the assembly line couldn’t seat inserts. About thirty percent of the housings cracked. My first reaction was to call the molder and demand they fix it immediately. But after five years of managing prototype and production orders, I know the question is almost never that simple.

Why the real problem is buried in vague specs

Here’s what I wish I’d learned earlier: manufacturers can only meet tolerances you actually specify. A CAD file without dimensions is like a grocery list without quantities. It tells the shop what shape you need, but not what level of precision matters. For injection molding specifically, plastic shrinks as it cools. Polyamide (nylon) data sheets typically list mold shrinkage around 1.5% to 2.5%, depending on the grade and wall thickness. So if the critical hole spacing doesn’t have a tolerance callout, the tool designer will pick a target based on average shrinkage and hope for the best. That’s not malicious—it’s just human nature. We assume the other side is thinking about the same features we are.

And “standard tolerances” are anything but universal. Some machine shops work to ISO 2768-m, which gives general ±0.2 mm for dimensions up to 100 mm. Others have internal guidelines that are tighter or looser. I’ve seen shops claim “standard tolerance” and deliver parts that mathematically meet their given standard but don’t work for the application. The same problem happens with processes. A friend of mine spent three weeks arguing with a supplier about the surface finish on an aluminum part that was machined in Vietnam. He’d assumed “anodized-ready” meant the same thing to everyone. The part met dimension, but the visible grain wasn’t acceptable. If you don’t define exactly what you need, you’re leaving it to chance.

Take draft angle, for example. It’s easy to design a vertical wall with zero draft in CAD, but in injection molding you need at least a little taper to eject the part cleanly. If you ignore that, the mold maker will add a draft that changes the wall thickness slightly. That tiny change can throw off a mating component if you haven’t accounted for it. I’ve seen designers complain about a supplier deviating from the model, when the supplier was actually preventing a production disaster. That’s the kind of knowledge you only get after working with the same process multiple times.

There’s also a deeper issue around basic manufacturing process selection. People confuse casting with additive manufacturing all the time. Is casting additive manufacturing? No. Casting means pouring molten metal into a mold—it’s been around for centuries. Additive manufacturing (3D printing) builds parts layer by layer from a digital model. Each process leaves a different fingerprint: casting has shrinkage and porosity, additive has layer lines and ironing effects. Using the wrong process for your prototype or production part can mess up your tolerance assumptions before you even get to material selection.

The real price of tolerance confusion

Now, the real cost. Back to that housing. The part price was $1.80 each. Great deal, right? But we didn’t specify which features were critical. After the first run failed, the molder quoted $6,000—actually it ended up being $6,400 after we added a couple of tweaks—to rework the tool. Then we had to pay $2,400 for rapid freight to make an upcoming trade show. Toss in engineering hours spent reviewing the tool revisions and coordinating with assembly, and we easily lost $10,000. All to save maybe $500 by not asking our engineer to mark up the drawing with tolerance callouts. It’s the definition of penny wise, pound foolish.

That same mistake shows up in less dramatic ways. Someone orders aluminum CNC machined parts from a new shop in one country to save money. In my case, a contact in Vietnam produced an accurate bracket but didn’t match the specified surface texture. Another time, I met an industrial designer in Burbank who was proud of his CO2 laser cutter—it made perfect acrylic prototypes, he said. But when the prototypes were tested, the laser-cut edges were too brittle for the application. He hadn’t realized that CO2 laser cutting creates a heat-affected zone, changing the edge properties. With a clearer spec, both of those problems could have been avoided.

What actually works

So here’s my hard-earned advice, trimmed down to the essentials.

First, mark critical tolerances on your drawing. Even if it’s just a few, identify which dimensions actually affect fit, function, or assembly. If it’s not critical, leave it loose. Second, design for the specific process. A tolerance that’s easy in CNC machining might be impossible or expensive in injection molding. Third, work with a supplier that gives you design feedback up front, instead of waiting until tool cutting.

That’s where ProtoLabs comes into my story. Their online platform does an automated DFM check before you commit. It flags risky geometry, points out likely tolerance issues, and shows alternatives. They also publish their manufacturing capabilities very openly: for injection molding, you can see how they define standard tolerances based on material and geometry. For CNC machining, they give you clear guidance on achievable surface finishes. As a buyer who isn’t an engineer, having those guardrails in place is worth a lot. It’s not just a supplier—it’s a second set of eyes on your design.

For example, their tolerance guidelines for injection molding are structured by material class. A general-purpose ABS part will have a different standard tolerance than a glass-filled nylon part. They break that down in a table-like format in their resource center. Same for CNC machining—they suggest what cutter radii and flatness you can reasonably hold. That’s not marketing fluff; it’s the difference between getting a quote that works and a part that doesn’t.

If you’re not sure where to start, try this: pick one product you’re about to prototype, spell out the three or four features that absolutely must fit, and send that drawing to a ProtoLabs test quote. Compare the DFM feedback with your own assumptions. It’s a quick way to educate yourself without a big commitment.

I won’t say ProtoLabs is perfect for every job. No vendor is. But they do one thing extremely well: they keep the manufacturing process transparent, so you don’t go into production with unrealistic expectations. And when your procurement process includes that kind of clarity, your internal stakeholders will thank you—your finance team sees fewer emergency invoices, your engineers get parts that actually match the CAD, and your line assemblers stop needing to file down bosses to fit.

That’s the real value of customer education. It’s not about dumbing things down. It’s about making sure both sides understand the same spec before signing off. An informed buyer is my favorite kind of customer, and I try to be that buyer myself.