There is a point in every shop where the printing stops being the interesting part. The parts you sell, the brackets you install, the jigs you promised a colleague — they all share the same quiet requirement: the second one has to be as good as the first. That is not a printer setting. That is tooling.
Most makers treat jigs, gauges and fixtures as things you buy from a machinist, not things you print. The arithmetic rarely supports that instinct. A printed gauge costs a few grams of filament and forty minutes; a machined one costs a quote, a lead time and a minimum order. When a job needs only a handful of pieces, printing the tooling is not a shortcut — it is the only version of the job that fits the budget.
This article is about the tooling layer: the small, unglamorous objects that make everything else repeatable. What each kind of tool is actually for, which decisions decide whether it works, where printed tooling genuinely fails and where it holds up for years, and what to check on the tool itself before you trust it enough to trust the parts it makes.
Four jobs, four different objects
The word “jig” gets used for everything, which is why printed tooling so often disappoints. There are at least four distinct jobs, and the design requirements barely overlap. A tool built for one job and applied to another will be wrong in a way that looks like a print quality problem.
| Tool type | Job it does | Primary requirement | Typical failure |
|---|---|---|---|
| Gauge | Answers a yes/no question about a part | Correctness of the dimension itself | The gauge was never verified, so it approves bad parts |
| Jig | Puts a feature in a fixed place during an operation | Locating geometry, not clamping force | Drill walk, wander, or an oversize hole |
| Fixture | Holds a part still while the load arrives | Stiffness and load path | Flexes under load, so the operation drifts |
| Holder | Keeps something in place between uses | Repeatability of position | Everything ends up in the same heap anyway |
The distinction matters because the failure modes are so different. A gauge that is 0.3 mm wrong silently passes bad parts for months. A fixture that flexes is obvious within minutes. The quiet one is the gauge, and quiet failures are the ones worth designing against.
If tooling is new territory, the general case for printed workshop aids — and the categories that pay back fastest — is covered in 3D printing jigs and fixtures. What follows is the layer underneath: why each one works or does not.
A gauge is only as good as its own calibration
A go/no-go gauge is the most useful thing in this article and the most dangerous. It compresses a measurement into a yes/no decision, which is exactly why it earns its keep on a bench where nobody wants to read a caliper forty times an hour. It is also why a wrong gauge is worse than no gauge at all: no gauge makes you measure, and measuring invites doubt. A wrong gauge makes you confident.
Slicers compensate for extrusion width, and printers are tuned by eye more often than by instrument. Two printers can be half a percent apart and both look fine. If the gauge is printed on the machine that is off, every part it approves inherits that error. The discipline is simple: print the gauge, then measure the gauge, with a caliper that has itself been checked against something you trust. Write the number on the tool in permanent marker. A printed gauge with a verified dimension and a date on it is a real instrument. One without either is a plastic opinion.
The same logic explains why gauge design is almost ascetic. Every feature that does not contribute to the decision is a place for error to enter, and a place for the operator to misread. The best ones look like they were designed in ten minutes, because they were.
Size it at the dimension that matters, never at nominal
The single most common reason printed tooling fails is that the critical dimension was drawn at the nominal size. Printed holes come out small by roughly 0.1 to 0.2 mm on a well-tuned machine, and printed bosses come out generous by a similar amount. A jig designed at nominal will fight every part it touches.
The fix is to design at the dimension that decides success. For a locating bore, that means subtracting the measured error from the model, not from the part after printing. For a clearance slot, it means adding material to the slot rather than sanding the part. Tooling is the one place in 3D printing where you should deliberately over-correct, because the tool's accuracy is what the whole batch inherits.
There is a useful second-order trick here. Printed tooling does not have to be dimensionally perfect everywhere — only at the surfaces that locate. Three small, accurate contact pads beat a large, approximately accurate face every time, because the contact pads can be adjusted, shimmed or reprinted without rebuilding the tool. This is the same discipline as modelling a part so it fits on the first attempt: decide which feature carries the constraint, and let everything else be generous. The general rules for turning tolerances into geometry are in manufacturing tolerances and design rules; the specific numbers for a given material and printer come from tuning dimensional accuracy, and the modelling habits from draw it once. The tooling layer simply applies all three, with less margin for error than a product part would get.
Stiffness, not strength, decides whether a fixture works
People sizing a printed fixture tend to ask what load it will take. That is the wrong question. A fixture almost never breaks — it deflects. And because the operation it guides is measured in tenths of a millimetre, a deflection that would be invisible on a shelf bracket is a functional failure on a fixture.
This is where the lessons from load-bearing part design transfer directly. Short load paths, closed sections and generous fillets do more for a fixture than extra wall thickness does, and the reasoning is the same one that governs ribs, gussets and bosses. A printed fixture that spans a gap should be treated as a beam and shaped accordingly: put the material where the bending moment is, keep the neutral axis far from the load, and avoid the temptation to add a thick flat plate because it feels solid.
Layer orientation matters here more than it does on decorative parts. A fixture that is loaded across layer lines is a fixture whose stiffness depends on adhesion, and adhesion is the one property you cannot set numerically. Where possible, print fixtures flat and load them in-plane.
Clamping force and locating force are not the same force
Hand-tightened printed clamps and screw-down printed hold-downs share a failure mode: the clamp provides so much force that it deforms the part and defeats the locating feature it was supposed to protect. Thin-walled parts and flexible materials are especially vulnerable. The part bulges, the hole shifts, and the operation that follows is accurate with respect to a distorted part.
The practical rule is to separate the two functions in the design. Locating should be done by hard stops that the clamp presses the part against, not by the clamp itself. A clamp's only job is to keep the part against those stops. Once that separation exists, clamping force can be as high as the tool tolerates, because it no longer controls position. The geometry of those stops is the same problem as any mount that has to survive a screw: the rules in brackets, holders and adapters apply directly.
Printed clamps also creep. A clamp left tightened overnight loses preload as the plastic relaxes, and the part shifts. For anything held for hours, either back the clamp off between operations or use a spring element in a material that recovers — the difference between PETG and ABS in sustained load is real, and neither is a metal spring.
Friction belongs in the tool, not in the part
A holder's job is to hold. Most printed holders fail not because they are the wrong shape but because they rely on friction between two printed surfaces, and printed surfaces have a friction coefficient that varies with layer height, material and how much the part has been handled.
Three changes fix most of it. First, add a mechanical stop so the part cannot slide past its intended position, and let friction only prevent lift-off. Second, use a soft or high-friction material at the contact — a strip of TPU 95A or a printed insert in TPU 85A costs almost nothing and converts a vague grip into a defined one. Third, give the contact surface a shape rather than a finish: a shallow V or a pair of ribs locates better than a flat face with more clamping force. The same reasoning about surface and grip applies to any part a hand or a tool touches.
Clearance, draft and the geometry of getting the part out
A tool that grips too well is as useless as one that does not grip at all. Printed tooling with tight clearances has a habit of becoming a permanent part of the job, especially in materials with any tack to them. Two design habits prevent it.
Add a deliberate ejection feature: a hole through the base of a pocket so a fingertip or a hex key can push the part out, or a slot pair for a screwdriver. And add a small draft angle to any deep pocket, even half a degree, because a parallel wall plus 0.1 mm of elephant's foot is a press fit you did not intend. If the tool must be tight, make it tight in one direction only and generous in the other.
Layer height interacts with this more than people expect. A taller layer produces a coarser wall texture, which grips harder; a finer layer produces a smoother wall, which releases more predictably. The trade is described in layer height explained, and for tooling the usual answer is finer than you would use for the part.
Soft jaws are the cheapest accuracy upgrade in the shop
A one-time concession to the material solves a whole category of problems. Soft jaws — printed inserts that sit between a metal vise or clamp and the workpiece — do three things at once: they conform to the part, they protect a finished surface, and they distribute load so a single point of clamping does not mark or distort the workpiece.
Because they are sacrificial, they can be made in a material chosen for the contact rather than the load. TPU soft jaws conform, PETG ones resist oil and coolant better than PLA, and both are reprintable in under an hour. The economic argument is the same as for any tooling: the moment a soft jaw saves one scrapped part, it has paid for itself several times over.
Where printed tooling genuinely loses
Honesty about limits is more useful than enthusiasm. Printed tooling is the wrong answer in four situations.
High side load over long durations. Plastic creeps. A printed fixture holding a heavy part sideways for a week will move, and no amount of wall thickness fixes a viscoelastic material under sustained load.
Precision better than about ±0.1 mm, repeatedly. Printing can be that accurate, but not reliably across a fleet of printers, materials and operators. If the operation demands it, use the printed tool for locating and a metal feature for the final datum.
Contact with sharp tooling at high speed. A printed drill guide will be reamed out by a spinning drill bit eventually. Insert a metal bushing, size the bore for a standard bushing, or accept a limited life. The same applies to a machine nozzle or a high-flow nozzle upgrade that has to be cleaned with a prick: the tool for that job is a hardened one, not a printed one.
Anything that will be sterilised, solvent-washed or cooked. The tool will not survive the process the part is going through.
None of these are arguments against printed tooling. They are arguments for using it where the load is light and the geometry is the hard part — which is most of the time.
Choose materials for the tool, not for the part
The material question for tooling is different from the material question for parts. Tooling is usually loaded lightly, handled roughly and asked to keep a dimension for a long time. That reorders the priorities.
| Material | Why it fits tooling | Watch out for |
|---|---|---|
| PLA+ | Stiffest per gram, most dimensionally predictable, cheapest to reprint | Creeps and softens in a warm shop; not for sustained load |
| PETG | Tough, resistant to oils and coolant, forgiving of abuse | Stringing on fine features; slightly gummy contact surfaces |
| PETG-CF | Higher stiffness for fixtures that span a gap, with less creep | Nozzle wear; less impact toughness than plain PETG |
| ABS / ASA | Stable at higher shop temperatures, resists solvents | Warping on large flat tooling; needs an enclosure |
| TPU 95A | Soft jaws, grips, anything that contacts a finished surface | Too flexible to locate; slow to print |
| TPU 85A | Conforming contact pads where the part shape varies | Very slow; poor dimensional control |
The most common wrong choice is PLA for a fixture that lives in a hot car or a warm workshop. The second most common is PETG-CF for a gauge that only needs to be stiff and accurate, where plain PLA+ would have been better, faster and cheaper. Stiffness matters for fixtures; stability and predictability matter for gauges. They are not the same requirement, and they rarely point to the same filament.
If the shop prints in more than one material, keep a dedicated spool for tooling that stays dry — a gauge printed from damp filament is a gauge with a dimension problem baked into it before the first layer goes down. A cabinet like the FilaBox drying box is cheaper than a scrapped batch.
The surface the tool sits on is part of the tool
A printed gauge is only meaningful if it sits flat on a known surface. A gauge placed on a bench with a 0.4 mm offcut under one corner is measuring the bench, not the part. This is why shop-built tooling usually comes with a base plate — a flat, printed plate that spends its life on the bench and defines the datum for everything else.
Base plates are the one place where sacrificing a build plate surface is worth considering. A smooth surface gives a flatter first layer than a textured one, and flatness is exactly what a datum needs. The trade-offs between smooth PEI and cold PEI matter far more for a base plate than for a decorative print, because the whole point is that the surface does not move.
Check the tool before you trust the parts
A tool earns trust the same way a part does: by being measured. The inspection discipline that applies to finished parts applies to tooling first, because tooling errors multiply. The full method is in measured, not guessed; for the tooling layer, four checks catch nearly everything.
| Check | What it catches | How often |
|---|---|---|
| Critical dimension with a caliper | A gauge that is wrong at the source | Once, on the day it is printed |
| Test part from the tool | Clearance, draft and ejection problems | Once, before the real batch |
| Repeatability: five parts, measured | Slop in a locating feature | First batch, and after any change |
| Wear and witness marks | Where the tool is being loaded or rubbed | Monthly in regular use |
The repeatability check is the one people skip and the one that matters most. Print five parts from the tool, measure the same feature on all five, and look at the spread rather than the average. A tool that produces parts with a 0.4 mm spread is not a precision tool, whatever the individual parts measure. The spread is the tool's actual contribution to error, and it is nearly always larger than the printer's.
Label the tool or lose the discipline
Printed tooling has a lifecycle problem: it looks like a part. It ends up in the bin of spare prints, the drawer of "that thing from a while ago", and eventually a reprint of something that was already correct. Two lines of permanent marker solve it — what the tool is for, and the date it was verified. Ten seconds, and it converts a lump of plastic into an instrument with provenance.
The same habit gives you a version history. When a part design changes, the tool's date tells you immediately whether it still applies. Tools are cheap enough to reprint that there is no reason to keep using one you are not sure about. If marking tiny features is difficult, the rules in stringing and oozing matter more than usual: legible tooling markings are read at arm's length, not through a loupe.
What a shop that prints its own tooling looks like after a year
The changes are not dramatic; they are cumulative. Setup time drops because the position of a part is defined by geometry instead of by eye. Scrap drops because bad parts are caught by a gauge before they reach the next operation. Typing a new part number stops being an event, because the first thing printed for a new part is the tool that makes the rest of it easy.
None of that comes from better printing. It comes from treating the tooling layer as a legitimate use of the machine — with its own design rules, its own material choice, and its own verification. The printer is already there. The tooling is just the part of the job that most people never got around to printing.
Start with one gauge. Verify it, write the number on it, and use it on the next batch you make. If it catches a single bad part, the argument for the rest of the tooling makes itself.