Crop before you annotate, not after
There is a reflex, when you need to point at something in a screenshot, to capture the whole window and then draw a large red arrow. It feels helpful. It usually isn't, and the reason is worth understanding because it changes how you capture in the first place.
An arrow is a suggestion; a crop is a decision
When you draw an arrow on a full-window screenshot, you have added a signal — but every other element in the frame is still there, still legible, still competing. The reader's eye has to find the arrow among the noise, follow it, and then filter out everything the arrow is not pointing at. You have given them a hint, not an answer.
Cropping removes the competition entirely. There is nothing else to look at, so there is nothing to filter. The reader cannot look at the wrong thing because the wrong thing is not in the image.
This is why the order matters. If you crop first, you often discover the arrow was never necessary. If you annotate first, you tend to keep the wide frame, because the arrow feels like it has already solved the problem.
The practical sequence
- Capture generously. Grab more than you need. Recapturing is more annoying than cropping.
- Crop to the claim. Cut down to the smallest region that still makes sense on its own. This is the step that does most of the work.
- Add context back deliberately — if the crop is now ambiguous, widen it slightly or add a caption. Don't undo the crop wholesale.
- Annotate only what the image cannot say. Sequence, causality, the value that's wrong, the thing that should be there and isn't.
That fourth step is the test for whether an annotation earns its place. An arrow
pointing at the only button in frame says nothing the crop hasn't already said.
A number 1 on that button, next to a 2 on another, says something the image
genuinely cannot.
What annotations are actually good at
Cropping cannot express order, and it cannot express absence. Those are the two jobs worth reaching for a tool over.
Order. Numbered step markers turn a static image into a sequence. "Click here, then here, then confirm" is three sentences of prose replaced by three small circles. This is the single highest-value annotation type and it is consistently underused.
Absence and comparison. A box drawn around empty space where a field should appear communicates something no crop can. So does a side-by-side where one half is marked correct and the other incorrect.
Values that are wrong. Highlighting 2,847 in a table of numbers tells the
reader which figure you're disputing. Cropping to just that cell would lose the
context that makes it wrong.
Everything else — arrows at the only interactive element, boxes around the obvious focal point, circles around a lone error message — is usually a crop that didn't happen.
When you genuinely can't crop
Some screenshots need the full frame: layout bugs, spacing problems, anything where the relationship between elements is the point. A crop would destroy the evidence.
In those cases the annotation is carrying real weight, so make it carry it well. Use one colour consistently. Keep line weight heavy enough to survive being viewed on a phone. And add a caption — when the image can't be reduced, prose is what narrows attention instead.
The habit worth building
Before you annotate anything, ask: could I delete part of this image and lose nothing? If yes, do that first and see whether the annotation is still needed.
More often than you'd expect, it isn't.