What to round to

Round to the nearest millimetre, or the nearest eighth of an inch if that is the scale you are reading. Not finer than that, and not coarser either - both directions throw away real information about what your reading actually supports.

Why a millimetre, specifically

A millimetre is roughly the practical resolution of careful technique with a rigid ruler and a tailor's tape: finer than that and you are reading between marks that a small variation in pressure or angle would move anyway, and the extra digit reports precision your technique does not have. Coarser than a millimetre - rounding to the nearest half-centimetre, say - discards resolution the method genuinely supports, and it flattens differences between readings that a repeat session would actually show you. If your readings themselves spread by more than a millimetre or two, which they often do, round to whatever unit your spread justifies rather than clinging to the finest mark on the ruler.

Half-up rounding

Where a reading sits exactly between two marks, round up rather than down. This is a convention, not a law of arithmetic, but it needs to be a convention you apply consistently rather than a judgement call made fresh each time - inconsistent rounding at the midpoint is a small, avoidable source of drift across a set of readings taken on different days.

The alternative conventions - round half to even, round half down - are not wrong in principle, and a statistician would point out that always rounding up at the midpoint introduces a tiny upward bias over many readings. For a handful of readings taken by hand, that bias is smaller than every other source of error already discussed on this site, and it is not worth the inconsistency of switching rules mid-log. Pick one convention, most simply half-up, and use it every time you write a number down. That fixed convention has a subtler cousin worth watching for in yourself - rounding bias in your own readings covers the unconscious pull toward a favoured number that no rounding rule alone fixes.

Round after averaging, not before

Take your raw readings to the resolution the instrument actually shows, average or take the median of those raw numbers, and round only the final result. Rounding each individual reading first and then averaging the rounded numbers compounds the rounding error across every reading instead of applying it once. Three readings of 14.3, 14.4 and 14.4 cm, rounded before averaging to a half-centimetre each, could easily collapse to three identical rounded values and erase a real half-millimetre of spread that was worth keeping until the last step. Whether that final central value should be a mean or a median in the first place is a separate decision that comes before rounding, not after it.

Reporting in inches instead

The same logic sets the resolution when you report in inches rather than centimetres: an eighth of an inch, not a sixteenth, matches what careful technique typically supports, and going finer just displays a false sense of certainty in a different unit. Whichever unit you settle on, pick the resolution before you round, not after - deciding on the fly tends to drift toward whichever mark the reading happens to sit closest to.

What rounding is not

Rounding to the resolution your technique supports is different from quoting a converted figure to more decimal places than the original measurement had - that specific trap gets its own coverage, because it is a common enough mistake to deserve one, and this page is about the rounding step itself rather than what happens when you convert units afterward. The general method a reading is built from treats rounding as the last step in the chain, after pressure, landmark and repeats are already settled.

What a rounded number does not carry

However carefully you round it, a length in millimetres or eighths of an inch is still only a length. It carries no information about how the result is perceived, which is a separate and subjective judgement - the kind a human reviewer gives rather than the kind a rounding rule produces. It is also unrelated to what a photograph can support: an image has no resolution in this sense at all, and an AI estimate drawn from a picture is not a rounded reading, it is an inference with its own separate uncertainty. A photo-based composite score, as compiled by a site like Penis Rater, answers a different kind of question and was never reporting to the nearest millimetre in this sense. And the number a service like Rate Cock produces is a different kind of figure again - rounding rules for a tape measurement have nothing to say about it.