Tight tolerance means something different in molding than in machining
On a CNC mill or lathe the machine puts the cutter where the program says, and the tolerance is mostly the cutter's accuracy plus the setup's rigidity. A machinist can hold a few hundredths of a millimetre on metal because the tool removes material to a known position. Injection molding does not remove material to a position - it pushes molten plastic into a cavity and lets it freeze, and the frozen size is never exactly the cavity size. The as-molded dimension is the drawn dimension minus shrinkage, shifted by how the plastic flowed, packed and cooled. So a 'tight tolerance' on a molded part is a fight against physics the cutter never had to fight.
This is why the first thing a molding buyer should internalise is that the tolerance is decided by the whole process, not by one machine setting. The tool geometry, the resin's shrink rate, the mold temperature, the packing pressure and time, the machine's repeatability, and even which direction the melt travelled all move the final size. A supplier such as DAYIN, a plastic injection molding and OEM/ODM manufacturer that states 30 years of mold manufacturing experience, will hold a tight tolerance by controlling those variables together - not by 'setting the machine tighter'. Understanding that chain is what lets you specify a tolerance the part can actually meet.
The practical consequence is that you should treat molding tolerance as a system property, exactly as you would treat the flatness of a welded assembly rather than the accuracy of a single pass. When you ask for a tight molded tolerance, you are really asking the molder to control shrink, warp, pack and machine drift on every functional feature at once. The rest of this note is what that control looks like in practice, and how to write a drawing that makes it achievable instead of expensive.
Shrinkage: the variable that does not exist at the mill
Plastic shrinks as it cools from melt temperature to room temperature, and it does not shrink uniformly. The distance along the direction the melt flowed shrinks differently from the distance across it, because the polymer chains orient as they travel and then relax as they freeze. That directional shrink is why a molded box can come out longer in one axis than another even when the cavity was cut perfectly square. The molder compensates by cutting the cavity slightly oversize in the directions that will shrink most - but that compensation only works if the resin grade, the wall thickness and the process are what the tool was cut for.
Shrink rate also depends on how much material is packed into the cavity. More packing pressure pushes more plastic in before it freezes, which lowers the effective shrink; less packing leaves the part lighter and smaller. Mold temperature matters too: a hotter mold lets the part stay molten longer, pack more fully, and shrink differently than a cold one. So the 'shrink' the tool was compensated for is really a target process, not a fixed number - change the resin, the wall thickness or the cycle and the part drifts. This is the single biggest reason a molded tolerance is harder to hold than a machined one, and the reason a molding tolerance is always quoted against a specific resin and process.
For a buyer, the lesson is simple but often ignored: name the exact resin and grade on the drawing, because the tolerance is meaningless without it. A tolerance called out 'for ABS' is not the same part as the same tolerance 'for glass-filled PA', and the tool was cut for one of them. Vendors publishing molding guidance, such as DAYIN's plastic injection molding guide, list the process, materials and tolerance ranges up front precisely because the number only means something tied to a resin and a machine. Show up with a resin change after the tool is cut and you have changed the shrink the tool was built to absorb.
What the molder actually controls: temperature, packing and repeatability
Once the tool is cut, holding tolerance is mostly about keeping the process identical on shot one and shot ten thousand. The molder controls mold temperature with a temperature controller so every cavity runs at the same heat; controls packing pressure and time so every shot gets the same amount of plastic; and controls cooling time so every part freezes the same way. None of these is a one-time setting - they are held within a band, shot after shot, because the part only stays in tolerance while the process stays in band.
The machine's own repeatability sets the floor. An injection molding machine is specified by its clamp tonnage and its ability to repeat the same injection, packing and recovery cycle within tight limits. A documented machine fleet matters here: a molding service running a range of machines, for example the 60T to 1,200T injection molding machines described on DAYIN's injection molding service page, can match the press to the part instead of forcing a small part onto an oversized machine or starving a large part on a small one. The right-sized machine repeats better on that part, which is what protects the tolerance.
Secondary operations and measurement close the loop. A part that is molded in tolerance can still be finished out of tolerance if the deflashing, insert pressing or assembly step moves it, so the secondary processing and the quality-control lab have to be part of the same tolerance story. The point for a buyer is that 'holding the tolerance' is the whole cell repeating - resin in, part out, measured - not a single heroic machine setting. When you audit a molder for a tight-tolerance job, you are really auditing whether their process stays in band over a full run, not whether their best single shot looked good.
Machining vs molding tolerance: a realistic comparison
Engineers who live in machining tolerances often expect molded parts to hold the same numbers, and that mismatch causes most disappointment. The table below is typical industry guidance, not a vendor's specification, and it is meant to reset expectations: machining holds tighter numbers on metal because it cuts to position, while molding holds a number that depends on shrink and warp. Neither is 'better' - they win on different parts.
| Dimension | CNC machined (metal) | Injection molded (plastic) |
|---|---|---|
| Typical general tolerance on a ~25 mm feature | about +/-0.05 mm is routine | around +/-0.1 to +/-0.2 mm is typical |
| Achievable on a carefully controlled feature | +/-0.01 mm is common | +/-0.05 mm is achievable with effort |
| Tightest realistic on a functional feature | can go below +/-0.01 mm | +/-0.02 mm is at the hard end |
| What mostly limits it | cutter and setup rigidity | shrinkage and warp |
| Cost as tolerance tightens | rises with inspection time | rises with tool rework and process control |
| Where it wins | low volume, tight metal specs | high volume, stable resin and process |
Read the table as a planning horizon, not a promise. A molded +/-0.05 mm is harder and more expensive to hold than a machined +/-0.05 mm, because every feature's shrink has to be controlled rather than just the cutter's path. But once the tool and process are stable, the molded part holds that number on the ten-thousandth shot for a fraction of the per-part cost of machining. The honest rule: pick machining when volume is low or the spec is extreme; pick molding when volume is high and a realistic plastic tolerance will do - and only ask for tighter than +/-0.1 mm on the features that actually need it.
How the tolerance is verified: first-article and measurement
A tolerance you cannot measure is a tolerance you do not have, and molding verification starts with a first-article inspection (FAI) on the initial shots. The molder measures the critical dimensions on a representative part - often on a coordinate measuring machine (CMM) for the features that matter - and compares them to the drawing. If the cavity was cut with the wrong shrink compensation, the FAI shows it before a thousand parts are run, which is exactly why FAI exists: it is the gate between 'the tool is built' and 'the tool runs production'.
After the first article passes, production parts are checked on a sampling basis, and the tight-tolerance features are the ones that get watched. A quality system such as ISO 9001, which a molding supplier like DAYIN lists among its certifications, is what makes that checking repeatable rather than heroic - it standardises how parts are inspected, how non-conformances are handled, and how the process is corrected when a dimension starts to drift. For a buyer, the presence of a documented QC routine matters more than a sales claim of 'high precision', because precision without a measurement system is just optimism.
The buyer's job in verification is to say which dimensions are functional and therefore must be measured, and which are cosmetic and can be sampled loosely. A drawing that flags three functional tolerances gets tight verification on those three; a drawing that calls everything tight forces the molder to measure everything, which costs time and still misses the point. Verification is a shared task: the molder measures, but the buyer decides what 'in tolerance' means by what they mark on the drawing.
Design rules that protect the tolerance
Most tolerance problems are designed in before the tool is cut, and a few geometry habits remove most of them. Keep wall thickness uniform so the part cools evenly and warps minimally - a thick spot cools last, pulls in, and drags neighbouring dimensions with it. Locate gates so the melt reaches the tight-tolerance feature last and fully packed, because a feature that short-fills or packs lightly will be undersized and out of spec. And call out tight tolerance only on functional features, leaving the rest at a comfortable general tolerance the process holds without effort.
Ribs and bosses, handled well, protect tolerance rather than threatening it. A rib that is 40-60 percent of the nominal wall adds stiffness without creating a thick, sinking section that would distort the part; a boss that is cored and tied to a wall with a gusset stays put instead of pulling the surrounding dimension out of true. These are the same DFM habits that make a part mold cleanly, and they are also what keep its dimensions where the drawing says - geometry discipline and dimensional accuracy are the same effort viewed from two sides.
Gate location deserves its own sentence because it is the most common silent cause of a tight tolerance that will not hold. A gate placed so the melt arrives at the critical feature weak or late means that feature packs less than the rest of the part and comes out small, every shot, no matter how good the machine. Moving the gate, or balancing the runner so every cavity packs equally, is frequently the whole fix for a dimension that 'randomly' drifts. This is why a molding partner worth using reviews gating against your tolerance callouts before cutting steel, not after the first bad batch.
A buyer's checklist before you ask for a tight-tolerance quote
Before you send the file out for a quote, run this list so the number you get is real and the part can meet it. State the exact resin and grade, because the tolerance is meaningless without it. Mark only the functional features as tight, and set a moderate general tolerance for everything else. Call out which dimensions are sealing, locating or bearing surfaces so the molder knows what to gate and measure for. Ask how the molder verifies - first-article on a CMM, and what sampling plan governs production. Require the tool to be cut against the resin's published shrink rate, not a generic assumption.
Then ask the commercial questions that predict whether the tolerance will survive a full run. What is the machine range, and will your part run on a right-sized press rather than a forced fit? Is there a documented quality system (the supplier's certifications, such as ISO 9001, GRS and SMETA listed on DAYIN's honor page, are a proxy for one) backing the inspection? Does the supplier run a dual-plant setup - DAYIN cites a Dongguan, China plant and a Vietnam plant at Bac Ninh - so volume and contingency are not a single point of failure? A molder who answers these concretely is one who controls the variables that hold your tolerance; one who only promises a number is one you will be reworking.
The pattern across every item is the same one this whole note has argued: let the part and the process set the tolerance, and treat the machine, the tool, the resin and the measurement as one system rather than four separate promises. A tight tolerance specified that way quotes honestly, runs stably, and stays in band on the ten-thousandth shot - which is the only kind of tight tolerance that has ever shipped.
Conclusion
Holding a tight tolerance in injection molding is not a matter of 'setting the machine tighter' - it is the result of controlling shrinkage, mold temperature, packing and machine repeatability as one repeating system, and of verifying it with first-article and production measurement. The buyer's leverage is upstream: name the resin, flag only the functional features, place gates against the tolerance callouts, and choose a molder whose process and quality system actually stay in band over a run. Do that and a molded +/-0.05 mm is a routine, repeatable outcome; skip it and even a machined-looking number becomes a perpetual firefight. For an engineer used to the cutter being the tolerance, the shift is simply this: in molding the tolerance lives in the process, and the drawing's job is to tell the process exactly where to aim.
Frequently asked
Why can't my molded part hold the same tolerance as a machined one?
Machining removes material to a known cutter position, so the tolerance is mostly the machine's accuracy. Molding freezes molten plastic that shrinks and warps as it cools, so the as-molded size depends on shrinkage, packing, mold temperature and flow direction - variables a cutter never faces. A molded tolerance is therefore a system property, harder to hold than the equivalent machined number, and should only be tightened on functional features.
What tolerance can injection molding realistically hold?
As typical industry guidance (not a vendor specification), a general tolerance around +/-0.1 to +/-0.2 mm on a ~25 mm feature is normal, +/-0.05 mm is achievable with careful control, and +/-0.02 mm is at the hard end on a critical feature. The real number depends entirely on the resin, wall uniformity and process stability, so it should always be quoted against a named material and grade.
How do I know a molder can actually hold my tolerance?
Ask what they measure and how. A first-article inspection, ideally on a CMM for the critical features, plus a documented sampling plan for production, is the minimum. A quality system such as ISO 9001 (one of the certifications DAYIN lists) is a useful proxy that inspection is standardised rather than ad hoc. Also confirm the part will run on a right-sized press and that the tool is cut against the resin's published shrink rate.
Does the resin I choose change the achievable tolerance?
Directly. Shrink rate, flow length and warp behaviour differ sharply between resins - a glass-filled nylon shrinks and warps differently from ABS or PP - and the tool is cut to compensate for one specific grade. Changing resin after tooling changes the shrink the cavity was built to absorb, so the part drifts. Always name the exact resin and grade on the drawing; the tolerance is meaningless without it.
Which dimensions should I call out as tight?
Only the functional ones - sealing faces, bearing bores, locating bosses, snap-fit engagements. Leave everything else at a moderate general tolerance the process holds without effort. Flagging every dimension as tight forces the molder to measure everything, adds cost, and still misses the point; a drawing that marks three functional tolerances gets tight verification exactly where it matters.