Reading the gap between two retakes

Two retakes under the same conditions should differ by about your own usual retake gap: a small gap is normal, and one well outside that baseline means something in the conditions moved. The retake test gives you the pair; this is about reading the gap.

Why a gap is expected, not a failure

No two photos taken minutes apart are pixel-identical, even with a phone held as still as you can manage in the same spot. Hand position shifts by a small amount, ambient light flickers slightly, the crop lands a few pixels differently each time. On top of that, the tool itself carries its own internal noise, separate from anything in your setup, and a retake pair measures both sources at once rather than isolating either one. Metrology builds this expectation into its vocabulary: the International Vocabulary of Metrology (JCGM 200:2012) defines repeatability conditions as the same procedure, operators, measuring system, operating conditions and location, with replicate measurements "over a short period of time" - the tightest setup it recognises, and still one whose results are compared rather than assumed identical. A gap of some size is therefore the expected outcome of a working protocol, not evidence that anything went wrong.

What a small gap tells you

A small, consistent gap across repeated retake pairs tells you the setup is tight and the noise floor is low. That is useful information in itself: it means a later comparison - a different tool, a claimed change, a longer time gap - only needs to clear this small floor to be worth taking seriously, rather than clearing an unknown and possibly much larger one.

What a large gap tells you

A gap noticeably bigger than your usual pattern is a signal, and the signal points at the conditions rather than at the subject. Something moved between the two shots that the protocol was supposed to hold still - the light source, the distance, the crop, the angle - and a large gap on a retake pair, taken minutes apart with nothing intentionally changed, is close to the only clean way to catch that kind of slip. Protocol drift builds up slowly across sessions; an unusually large retake gap is the same failure caught in miniature, inside a single sitting, before it has had months to compound.

Finding what moved

Work back through the log line by line: device, distance, angle, light, crop, background, state. A large gap with everything on the log holding constant across the two shots is a different and more useful finding than a large gap where you already suspect the lamp got bumped between shots - the first says something about the tool's own noise, the second says something about your process. This is a separate question from whether the noise originated in the photo or in the tool itself, which needs its own pair of tests to isolate; a large retake gap only tells you that the total noise is bigger than usual, not which side it came from.

No invented numbers

There is no universal figure for what counts as small or large here, and a number offered without a source would be worse than no number at all. Your own baseline retake gap, measured a few times under your own conditions, is the only reference worth comparing a new pair against - a size that is normal for one protocol, one device and one tool is not automatically normal for another, because the sources of noise it is summarising are specific to your setup.

Building your own reference gap

Run the retake test a handful of times before you rely on it, not just once. A single retake pair is itself a sample of one, and treating its gap as "the" gap risks mistaking one unusually small or large draw for the typical case. Three or four pairs, taken across a few different sessions rather than all in one sitting, give you a rough sense of the range your setup normally produces - tight sessions and looser sessions both happen, and seeing a few of each is what lets you tell an outlier from the ordinary spread. Once that range exists, a new pair falling well outside it is the signal worth investigating; a new pair falling inside it is not.

What a consistently large gap means

If every retake pair comes back with a larger gap than you would like, the answer is not to distrust the tool - it is to tighten the protocol itself. A large typical gap, as opposed to one unusual pair, generally traces back to a variable that is harder to hold than it looks: a light source that is more directional than "diffuse" suggested, a crop done by eye rather than against a fixed reference, a distance judged by feel rather than marked on the floor. The full protocol is the place to look for which line is loosest, one variable at a time, rather than assuming the whole setup needs to be rebuilt.

Where the same reading applies elsewhere

A documented measuring method reports its own repeat variation for exactly this reason - a gap between two measurements of the same thing is diagnostic, not disqualifying, once you know what your own typical gap looks like. A human reviewer forming two impressions of near-identical material will show some variation too, and the same principle holds: a small difference is ordinary, a large one is worth asking what changed between the two. A tool that shows a breakdown by axis makes this diagnosis easier than a single total does, because a large gap that is concentrated on one axis narrows down which condition moved far faster than a total that only tells you the sum changed.