Standardising a submission, not flattering it

To make a submission comparable, fix everything that is not the subject - distance, angle, light and framing - and write those settings down. Most photo advice is about scoring higher; this is about scoring repeatably, because a number you can reproduce is a number you can compare.

The four variables worth fixing

Distance. Phone lenses are wide-angle, so anything near the lens is enlarged relative to anything further away, within the same frame. Change your distance and you change the proportions the tool reads. Ward and colleagues (2018) put a number on this for faces: a photo taken at 12 inches enlarges the nose by 30% in men and 29% in women compared with an undistorted projection, while at 5 feet there is essentially no difference. Pick a distance, note it, use it again. Further back with a crop afterwards is more stable than close in, because small errors at close range are proportionally large. Why the score moves when you move the camera is the geometry underneath all four of these.

Angle. Level with the subject, not looking down. Downward angles foreshorten, and how much depends on exactly how far down you tilted - which you cannot reproduce by feel. Level is the one angle you can hit again.

Light. Consistent, diffuse, from the side. Not flash - direct flash blows the surface flat and takes the texture judgement with it. Window light on an overcast day is the most reproducible light most people have access to, because it does not depend on the time of day the way direct sun does.

Background and framing. Plain, and cropped the same way each time. Both affect the subjective axes without touching the physical ones, which means leaving them uncontrolled adds variance to exactly the axes that are already noisiest.

Why this beats optimising

Optimise for the highest score and you get one number you cannot reproduce. Standardise and you get a number that means something the second time you take it - which is the only way to tell whether anything actually changed.

It also makes the tool's own behaviour visible. Hold everything constant, change one variable, rescore: now you are measuring the tool's response rather than guessing at it. That is the same procedure that tells you whether two tools genuinely disagree or merely use different scales.

The same standardisation pays off if you send the material to a person rather than a model - a human judge responds to what you send, and two clean photos beat six inconsistent ones there too.

Write it down

Distance, angle, light source, crop. Four lines in a note. Anyone who has measured properly will recognise the discipline: the method goes on the same line as the number, or the number is not data.

This sounds fussy for about a week and then it is the entire reason your results are worth anything. Without it, a comparison between March and June is a comparison between two unknown setups.

What standardising does not fix

The scale is still arbitrary, the ends are still compressed, and the middle is still crowded. Standardising gives you a reproducible position within one tool's distribution, not an absolute anything.

And the tool still has to show you enough to interpret the result. A total alone tells you the number moved; it does not tell you which judgement moved, so it cannot tell you whether the change you made did what you meant it to. That is why a rubric is the property worth selecting a tool on - services like Rate Cock that report per-axis scores let you attribute a change to a cause, and services that report a single figure structurally cannot.