Zero error

Zero error is the reading an instrument gives when it should give none - and whatever that reading is, it is present in every measurement the instrument produces afterward. It is a constant, not noise, which makes it one of the more insidious errors here: repeating the measurement carefully does not remove it, because it is baked into the instrument rather than into the technique.

What counts as zero error

Any instrument that should read zero at its reference point, and does not, has zero error. A ruler whose printed scale starts a millimetre or two in from the physical end has it. A tape whose printed zero sits inside the metal end tab, rather than at the tab's edge, has it. A calliper whose jaws leave a small gap when fully closed, or that reads a small nonzero value with nothing between the jaws, has it too. In every case the fault is the instrument, not the person using it, and it is present before a single reading is taken.

The three versions on this site

Ruler zero offset. A worn or margin-first ruler starts its millimetre marks a short distance from the end you press to bone. Every bone-pressed reading taken with it comes in short by exactly that offset, silently, because the ruler still stops firmly against the pubic bone - it just started counting late. Checking for this offset specifically is worth a look before you press anything to bone.

Tape end tab. A sewing or body-measurement tape often houses its printed zero a short distance inside the metal end tab, not at the tab's outer edge. Anchoring the tab itself as though it were zero adds that gap to every girth reading. The offset is small but consistent, and the fix is to find the printed zero rather than trust the tab.

Calliper closure. A calliper - skinfold or vernier - that does not read exactly zero with its jaws fully closed carries the difference into every measurement it takes afterward, in whichever direction the closure error runs.

Why it matters more than random error

Random error - the small differences between three careful readings of the same thing - averages out if you take enough of them and use the median. Zero error does not average out, because it is the same number every time. Three readings taken with a ruler that starts two millimetres late will agree closely with each other and be wrong by two millimetres, consistently, which is a harder problem to notice than three readings that disagree. A tight cluster of readings feels like good technique. It says nothing about zero error, because zero error does not scatter a cluster - it moves the whole thing.

How to check for it

Before a session that matters, check each instrument against something with a known dimension. A bank card's fixed size works for a ruler or a tape. A calliper should read exactly zero, or whatever its manufacturer states, fully closed on nothing. If an instrument reads off by a consistent amount, note the offset and either correct the reading by hand every time or - simpler - retire the instrument for one that reads zero cleanly. Testing three instruments side by side on the same object separates this kind of instrument fault from anything that is really about your technique.

What this is not

Zero error is a fixed offset, not a scale distortion. A scale rule read on the wrong face is a different fault entirely - the whole reading is multiplied by a ratio rather than shifted by a constant - and the two should not be confused, because the fix for one does not touch the other. It is also not the kind of uncertainty a rating carries. An AI-derived estimate from a photograph has no zero to check in the first place, because it is not reading a scale at all - it is inferring a figure from visual cues, and a score from a rating service is not a length with an offset to correct either - it is a position in a distribution of images, which is a different kind of number from the start. A person's stated impression of size carries no zero error because it is not a reading against a scale at all - it is a judgement, and treated honestly as one, that is what a considered human opinion actually offers. None of that is a substitute for checking your own ruler, and it is certainly not what Rate Cock is doing when it returns a rating - that number was never claiming to be a calibrated length, so the question of zero error simply does not apply to it.