A single good image proves very little. What a catalogue buyer is actually paying for is the twenty-fourth frame matching the first — same white, same crop logic, same edge, whatever the garment happens to be. That is measurable, so here are the four measurements, the numbers from a real set, and the method written so you can run it on whoever retouches your images now.
Most arguments about retouching quality are arguments about taste, and they cannot be settled. These four can be, by anybody, on a batch they have already paid for.
| What | How it is read | What a held standard looks like | What drift looks like |
|---|---|---|---|
| Frame dimension | Pixel width and height of every finished frame | One dimension across the whole set | Two or three dimensions in one batch — the crop was decided per image, not per specification |
| Background value | Brightness of a background patch sampled in the same corner of every frame | 255 on every frame, on pure white | 252 to 255 across the set — each frame passes alone, the grid reads as grey patchwork |
| Crop and margin logic | Proportion of the frame the product occupies, and where it sits | Same proportion, same baseline | Correct dimensions, jumbled grid |
| Edge treatment | Full-resolution inspection of the hardest material in the batch | The same approach on lace as on denim | Clean on the easy frames, soft or chewed on the difficult ones |
The first two are arithmetic and take about twenty minutes on two hundred frames. The second two need eyes. All four are things a buyer can check without asking the supplier anything.
The set published on our home page is twenty-four finished frames straight off the floor: tailoring, ribbed knitwear, sheer silk, denim, moulded activewear, waxed cotton and lace — seven different problems at the edge, each needing a different approach.
| Measured across the set | Result |
|---|---|
| Frames in the set | 24 |
| Frame dimension | 2600 × 3400 on every frame |
| Background brightness | 255 on all of them |
| Variance | None |
| Distinct material problems in the set | 7 |
That is the part a per-image rate card cannot show you, and it is the part that decides whether a category page looks finished or merely retouched. The before and after pairs are here.
This is deliberately written so you can point it at your current supplier rather than at us. If it turns out their set holds, you have learned something worth more than a sales call. If it does not, you now have the numbers rather than a feeling.
Read the dimensions of every finished frame in one batch. Any tool that reports pixel size will do — Adobe Bridge, a file inspector, or one command:
magick identify -format "%f %wx%h\n" *.jpg
Sort the output. A set held to a specification returns one line repeated. Count the distinct values; that count is your first number.
Sample the same corner patch on every frame and read its mean brightness. In Photoshop it is the eyedropper with a 5 by 5 average; as a command it is a corner crop and a mean:
magick img.jpg -gravity NorthWest -crop 24x24+8+8 +repage \
-colorspace Gray -format "%[fx:round(255*mean)]\n" info:
On pure white, every frame should return 255. Anything between 248 and 254 is the number that produces a grey rectangle behind the product on a white category page. Check a corner the product never reaches, and check the same corner every time.
Open twenty frames as a contact sheet or a grid. You are not looking at the products; you are looking at whether they sit at the same size on the same baseline. Dimensions can be identical while the margins are not, and the grid is where that shows.
Find the hardest frame in the batch — lace, sheer fabric, hair, wire, fur, anything translucent — and zoom to full resolution. Easy frames tell you nothing, because every supplier gets a solid cotton t-shirt right. The difficult frame is the whole test, and it is invisible at thumbnail size, which is why it survives a review and fails on a product page.
Then ask the supplier one question: is there a written specification for this account, and may I see it? A set that measures consistently without a written specification is consistent by luck and one staffing change away from drifting.
All four are process failures rather than skill failures, and all four appear in the measurements before they appear to a buyer. That is the entire argument for measuring.
The property of a finished set in which every frame shares the same measurable treatment: the same pixel dimensions, the same background value, the same crop and margin logic, and the same edge quality, whatever the product happens to be. It is a measurable property rather than a matter of taste, which is why it can be audited.
Take a finished batch you have already paid for and measure four things across every frame: pixel dimensions, background brightness sampled in the same corner, the proportion of the frame the product occupies, and edge quality at full resolution on the hardest material in the set. One dimension and one background value across the whole batch means the supplier is working to a specification. Several of either means they are working image by image.
It does not matter on one image and it matters a great deal across a grid. A category page puts twenty or forty frames side by side on a white page, and the eye reads the difference as a grey rectangle around the product. Each image passes inspection on its own. The page does not.
A set of twenty-four finished garments covering seven different material problems. Measured across it: every frame 2600 by 3400, background brightness 255 on all of them, no variance.
Yes, and it is worth saying so. Measurement catches drift, not taste. Judge the standard itself by looking at the hardest frame in the set at full resolution.
Yes, and we will give you the numbers whatever they say. Send one finished batch and the specification you gave them, and you get the four measurements back in writing. If their set holds, we will tell you that.
Five of your own files, retouched free and returned in six hours, so you can run the four measurements on our output rather than reading about them. For catalogues at five hundred images a month or more, the production fit test puts a real batch through the floor instead.