A Sentinel-2 scene is not a photograph. It is a stack of measurements in 13 wavelength bands, from visible blue through near-infrared to shortwave infrared. Your monitor can only show three bands at once, one each for red, green and blue. Which three you pick changes the image completely, and experienced analysts switch combinations the way a photographer switches lenses: each one answers a different question.

For a natural-looking image, assign band 4 (red) to red, band 3 (green) to green, and band 2 (blue) to blue. The result looks like an aerial photograph: blue water, green fields, grey cities, white clouds. True colour is the best starting point for orientation, for checking cloud cover, and for any map meant for a non-specialist audience, because nothing needs explaining.

Its weakness is that it hides information. Healthy crops and stressed crops can look almost identical in true colour, and shallow water over bright sand can look like dry land. True colour shows what the landscape looks like, not how it is behaving, which is why analysts rarely stop there.

Once you find combinations you like, save them as QGIS layer styles so you never retype band numbers. Right-click the layer, choose Styles, Save as QGIS Layer Style File, and you get a small file you can reuse on any scene from the same sensor or share with a colleague. Name the files clearly, for example sentinel2-false-colour-infrared, so a teammate can apply the right view without asking you what the numbers mean.

Saved styles also make your work reproducible. When you hand over a project, the style files travel with it, and anyone opening the project sees exactly the view you used to make your interpretation. That matters when a map is questioned months later: you can show not just the data but the exact visualisation that led to the conclusion, which turns a judgement call into a documented step.