Repair tools · torque
One wrench does not cover the car.
A click-type torque wrench is one of the few things in a garage with a certified tolerance rather than a marketing figure. It is also sold with a range printed on the handle that is wider than the range it is actually specified over — and the gap falls exactly where fasteners strip.
What ±4 per cent actually is
The international standard permits a click wrench to deviate by ±4 per cent. That is not a defect; it is the specification, and a wrench that holds it is working correctly. What it means in practice:
| Set to | Actually applied | Spread |
|---|---|---|
| 25 N·m | 24.0–26.0 N·m | 2.0 N·m |
| 30 N·m | 28.8–31.2 N·m | 2.4 N·m |
| 110 N·m | 105.6–114.4 N·m | 8.8 N·m |
| 120 N·m | 115.2–124.8 N·m | 9.6 N·m |
Nine newton-metres of spread on a wheel nut is fine — the fastener and the stud are sized for it. Two on a spark plug in an aluminium head is also fine. Neither is the problem. The problem is the next section.
The rule nobody is told
A torque wrench is specified only over the upper part of its scale — conventionally from about 20 per cent of maximum upwards. Below that the ±4 per cent no longer applies and the reading is not covered by anything. The handle is still marked down there.
So a 1/2 in drive wrench marked 40–200 N·m is specified from 40 N·m up. A 3/8 in marked 20–100 is specified from 20 N·m up. The number where the guarantee starts is not printed anywhere on the tool.
Which wrench for which job
| Job | N·m | 1/4 in drive5–25 | 3/8 in drive20–100 | 1/2 in drive40–200 |
|---|---|---|---|---|
| Spark plug, alloy head | 25 | in spec | in spec | — |
| Oil drain plug | 30 | — | in spec | — |
| Brake caliper bolt | 110 | — | — | in spec |
| Wheel nut | 120 | — | — | in spec |
Computed from the standard's rule rather than typed: a job is in specification when it falls at or above 20 per cent of the wrench's maximum and at or below the maximum.
Read the two right-hand columns. The 1/2 in wrench most people own — bought for wheel nuts, which it does perfectly — is out of specification for both the spark plug and the drain plug. Those are the two fasteners that most commonly strip, in aluminium, on a thread that costs a cylinder head or a sump to repair.
That is the argument for owning two: a 1/2 in for suspension, brakes and wheels, and a 3/8 in for everything on top of the engine. Not an upsell — the standard simply does not cover one tool doing both.
It goes out of calibration and does not say so
The standard calls for re-verification annually or after about 5,000 cycles, whichever comes first. A click wrench gives no indication when it has drifted: it still clicks, at the wrong number.
Two things follow that are free. Wind it back to the bottom of its range before storing it — leaving the spring compressed is what puts a wrench out of calibration fastest. And treat a wrench that has been dropped as unknown until it is checked, because nothing about it will look different.
Where the standards people disagree
Worth stating because we cannot settle it. A calibration laboratory argues that ASME B107.300 is a product standard and not a calibration standard, and that ISO 6789 is the correct reference; a trade publication treats both as relevant. We are not merging them.
It does not change the number: the ±4 per cent tolerance and the upper-part-of-scale rule are the same whichever document you cite, which is why this page is built on those rather than on the argument.
What nobody has measured
- How far a cheap wrench actually drifts. No published test takes a sample of budget wrenches, cycles them 5,000 times and re-measures. Every claim about cheap tools going off is anecdote.
- Out-of-specification error below 20 per cent. The standard declines to cover it; nobody has measured how bad it gets. It may be 5 per cent or 30 and no buyer can find out.
- Click versus digital in a real hand. A digital wrench beeps and a click wrench breaks, and no test measures how much a human overshoots each one under a car.
use it ISO 6789, Assembly tools for screws and nuts — hand torque tools, current edition 2017. The international standard hand torque tools are built and calibrated to. It sets the permitted deviation for a click-type wrench at ±4 per cent, specifies accuracy only over the upper part of the scale rather than all of it, and calls for re-verification annually or after about 5,000 cycles. It is a certified tolerance rather than a manufacturer's claim, which is why a wrench can be held to it.
marked Torque Lab and Quality Magazine, Which standard actually covers torque wrench calibration, February 2026. Sources disagree about which document governs. A calibration laboratory argues ASME B107.300 is a product standard rather than a calibration standard and that ISO 6789 is the correct reference; a trade publication treats both as supplying relevant information. Both are shown here because the dispute is real and unresolved, and because the tolerance figure itself is the same either way.
arithmetic The torque bands and the in-specification table are computed when this page is built, from the tolerance and the range rule above, so the prose and the table cannot drift apart. Wrench ranges are the common nominal ones; check the figures printed on your own tool, since the specified floor is 20 per cent of its maximum.