Direct imagery: subject, crop and grade

The section says bread, baked every morning, for the people who queue for it. The photo beside it is a stock picture of a pastry counter, and nobody knows whose counter it is. There's no bread from this bakery in it, and nobody is queuing. That gap between what the copy claims and what the picture shows is where most imagery goes wrong.

So direct the subject first, and the subject is the claim. If the copy says people queue for it, the picture shows the queue, at the door, in the early light. If it says today's bread, the picture shows today's bread: the loaves and their prices, on the board. My test for any image is dull and it works: cover the caption and ask what the picture proves. If the answer is 'that somebody sells pastries', it's in the wrong place.

The Quiet Luxury record names the failure precisely: expensive-looking stock photography that creates atmosphere and gives nothing to inspect. The same record lists useful subjects instead, including material detail, construction, hands and process, and objects in context. That list works for almost any business that makes something.

Then crop. A crop decides what the viewer looks at, so cut to the one thing the image is there to prove. A tight crop that lets the subject press against the frame adds tension. A centered, comfortable crop rarely adds anything. And size images by importance. Swiss Style's record documents it: pictures cut to a few sizes that are whole multiples of the grid field, each size reflecting the picture's importance. You can borrow the logic without being Swiss.

Then check every crop at phone width, because that's where crops die. A wide image squeezed into a narrow column either shrinks the subject, here to a strip a hundred and sixty-four pixels tall, or gets cropped by the browser, and the browser doesn't know what the picture is for. Decide the phone crop yourself, and write it down: on a phone, the front of the queue and the door, nothing else. Your developer, or your agent, can set that crop for phone widths.

Grade is where a set of images becomes a set. Pick one color temperature and one contrast level, and apply them to every image on the site. Warm and soft or cool and hard, it doesn't matter which, as long as there's one. A golden-hour photo next to a blue fluorescent one looks assembled from different shoots, because it was. Keep the light coming from the same side, too. Write the grade down so it survives the person who made it. Here's the grade note from another fictional client, a bookbinder. Warm daylight, soft shadows, saturation slightly under so the oxblood leather stays true to the brand color. No filters. And the check that catches most problems: every image, side by side on one screen, before launch.

All of that goes into an asset plan, one row per image slot: the claim it supports, the subject, the crop, and where the image comes from. The source line matters. If you don't have the image yet, the row says so, with a placeholder that can't ship. If it's generated by AI, the row says that too, and the image is never used as evidence of the work: it can set a mood, but it can't stand in for the bindery, the books or the hands.

So the order is claim, subject, crop, grade, then a review of the whole set at once. Direct images one at a time and you end up with twelve good photographs that don't belong on the same site. Direct the set and even ordinary photographs start to look like a decision.

Every image has to show the thing the words claim.