How to test if your food photo works at mobile thumbnail size

The photo that wins the order on a delivery app is not the photo on your hard drive. It is a 200-pixel-wide square on a phone screen, sandwiched between five other dishes, scrolled past in under two seconds. We keep finding restaurants who shot beautiful 4K studio plates and lost to the burger joint next door because that burger joint's photo still reads at thumbnail size.

This piece is about that gap. Specifically, how to test whether your food photo survives the trip from full size to the actual mobile thumbnail, before you upload it to Thuisbezorgd, Uber Eats, or Deliveroo.

Why the mobile thumbnail is the entire purchase decision

Delivery apps render food photos at thumbnail size in every list view: search results, category grids, "popular near you", reorder shortcuts. The customer scrolls. They are not reading. They are pattern-matching: yes, no, yes, no, tap.

DoorDash sets a thumbnail minimum of 230 by 230 pixels and crops everything to a 1:1 square from the centre of your upload, per the DoorDash and Uber Eats photo specs comparison. Uber Eats applies a similar 1:1 crop on menu thumbnails even though the upload spec is 5:4 at 1200 by 800. Thuisbezorgd, run by Just Eat Takeaway, ingests landscape 3:2 at 1200 by 800 minimum and crops a 1:1 square from the centre for search cards and mobile lists, per Just Eat Takeaway's partner guidelines.

In every case the photo your customer sees first is a square crop, displayed roughly between 150 and 250 pixels wide on a typical phone. Anything you painted into the corners is gone. Anything subtle is gone. What remains is a colour, a shape, and maybe two recognisable details.

The 1.5-second mobile thumbnail test

Eye-tracking work on online menus shows users scan a page in roughly six seconds before focusing, with the first two to three seconds spent on the highest-contrast items, per the first impressions eye-tracking study summarised by ScienceDaily. On mobile delivery apps the scrolling cuts that even shorter. The window where a thumbnail has to land is closer to a second and a half.

Here is the test we run on every photo before it goes live:

1. Export the image at 1200 by 800 (or whatever the platform asks for).
2. Resize a copy to exactly 200 by 200 pixels, square-cropped from the centre.
3. Hold a phone at arm's length.
4. Glance at the 200-pixel version for 1.5 seconds.
5. Look away.

If you cannot say what the dish is, in one short sentence, the photo fails. Not because it is a bad photo. Because it is a bad thumbnail. Those are different jobs.

The dish recognition has to be near-instant. "Burger with cheese." "Bowl of green curry." "Margherita with basil." If your answer is "some kind of pasta thing" or "a plate with sauce", a customer scrolling at 19:00 on a Tuesday will skip it.

Thumbnail eye-tracking patterns on a mobile menu

Restaurant menu eye-tracking research is mostly done on printed and large-screen menus, where users read in roughly an F-pattern, summarised in the Interaction Design Foundation overview of visual hierarchy. On a single-column mobile delivery list the F-pattern collapses into a vertical strip, and the eye does something simpler: it lands on the photo first, jumps to the dish name, jumps back to the photo, and then either taps or scrolls.

Two practical implications:

The photo decides whether the name even gets read. A weak thumbnail means the title underneath never gets a fixation. Customers who scroll past your dish are not making a price decision or a menu-copy decision. They are making a thumbnail decision.

Position matters less than people think. The widely cited "top-left gets the most attention" finding applies to two-column print menus. On a phone, every dish gets the same amount of attention if its thumbnail earns it. The first three dishes in a category have a slight advantage. Not enough to save a photo that does not work at 200 pixels.

Thumbnail contrast at 200 by 200 pixels

The single biggest reason photos fail at thumbnail size is low contrast between the food and the background. A dark sauce on a black plate disappears. A pale risotto on a white tablecloth disappears. A green salad on a wooden table that has gone slightly green in the white balance disappears.

We run a quick contrast check on every photo:

Open the 200 by 200 version. Squint at it, hard enough that the image goes slightly out of focus. If you can still see a clear shape of the dish against the background, the contrast is good. If the whole thumbnail blurs into one tone, the contrast is failing.

A second check: convert the thumbnail to greyscale. Food photos that work as thumbnails almost always survive greyscale conversion. The shape and the lighting carry the image, not the colour. If your photo only reads in colour, it is leaning on saturation that the platform's compression will eat.

Concrete examples from our customer accounts. A noodle bowl shot top-down on a dark grey ceramic plate with a pale broth and bright green herbs reads at 200 pixels because there are three contrast levels: light broth, dark plate, bright herb. A flat-lit pasta in a white bowl on a white tablecloth has one tone and disappears. The food is the same. The thumbnail is not.

A fast A/B mental test for mobile thumbnails

Before uploading, we run two photos against each other in a quick mental A/B. This works whether you have one option or several:

Open both photos at thumbnail size, side by side on your phone. Look at them in turn for one second each. The one that names itself first wins. Not the prettier one. The faster one.

If you only have one photo, run it against a competitor's thumbnail in the same category on the actual app. Open the customer-facing app, search the dish category, find a competitor near the top of the list, and put your photo next to theirs at the same size. If your photo loses the glance test, the algorithm and the customer will agree.

This is uncomfortable. Most photos that look gorgeous on a laptop lose to a slightly worse photo with stronger contrast and a more obvious dish silhouette. We have killed beautiful images this way and replaced them with photos that felt less impressive but converted better. Conversion does not care about beauty.

Five steps to evaluate your own mobile menu thumbnails

A practical workflow we walk new customers through:

Open your live menu on the customer-facing app, on a phone, in the lighting you would actually use it in (so, dim, indoors, evening). Screenshot the category page. Each dish is now a thumbnail next to every competitor in the same neighbourhood.

Sort the screenshot mentally into three piles: dishes you can identify in 1.5 seconds, dishes you have to think about, and dishes you cannot identify without reading the title. Anything in the second or third pile is the rework list.

For each rework, run the contrast check and the greyscale check. Most failures are contrast failures, not composition failures. A photo with the right contrast but a slightly awkward angle still wins over a perfectly composed photo that goes flat at 200 pixels.

Reshoot, recrop, or regenerate. If the issue is the original photo, a food photo generator can rebuild the dish on a high-contrast background without you needing to reshoot. If the issue is the crop, sometimes pulling the plate slightly off-centre and zooming in fixes the thumbnail without touching the source.

Reupload, then check the live thumbnails on the app two days later. Both Thuisbezorgd and Uber Eats cache aggressively, and what you see in the merchant portal is not always what the customer sees. Always verify on the customer app.

What thumbnail-first thinking changes about how we shoot food

Once you start grading photos by their thumbnail performance instead of their full-size beauty, the priorities flip. Strong silhouettes beat soft styling. High-contrast surfaces beat tasteful neutrals. Tight crops with the dish almost filling the frame beat wide compositions with breathing room. Bright, identifiable colours beat muted palettes that look elegant on a desktop screen.

None of this means full-size photos do not matter. The dish detail page still gets the landscape shot, and on tablets and laptops the breathing room helps. But the order is decided at thumbnail size, and the dish detail page only loads after the customer has tapped. If the thumbnail does not earn the tap, the rest of the photography never gets seen.

If you want to test the workflow without reshooting your menu, three free credits are enough to regenerate one dish, drop the new thumbnail next to the old one on your phone, and run the 1.5-second test yourself. That is the only test that matches what your customers are actually doing.