Product image too small? Check the rest of the gallery

For the primary comparison, we chose a portrait canvas measuring 1,080 pixels wide by 1,350 high. This is a design example, not a universal platform requirement. An image met the condition only if its recorded width was at least 1,080 and its height at least 1,350, in the stored orientation.

Of the 19,542 eligible amazon.com listings, 5,985 had a main image that missed one or both dimensions. Within that group, 3,073 had a secondary image that cleared both. That is 51.3% of the listings where the main file failed the size check, not 51.3% of every listing in the archive.

The complete table also shows the other side of the finding. There were 2,912 listings where neither the main image nor any recorded secondary image met the chosen condition. Looking through a gallery is a useful check, not a promise that a suitable file will appear.

Across all 19,542 eligible listings, those 3,073 cases represent 15.7%. The difference matters when interpreting the result: the 51.3% share describes the group that already has a main-image size problem. Within that group, checking another file can reveal a candidate the first image did not provide.

To fill a portrait canvas, a photo must cover its full width and height. If the photo has a different shape, part of it will extend beyond the canvas and be cropped. The scale needed is the larger of two ratios: canvas width divided by image width, and canvas height divided by image height.

For our chosen canvas, a 1,350 × 1,080 landscape image fails the condition. Its width is sufficient, but its height is not. Having the same total number of pixels as the canvas does not solve that mismatch. The calculation keeps the recorded orientation; it does not rotate the photo, add borders or generate missing background.

Keeping the whole photo visible asks a different question. A designer could shrink an image to fit inside the canvas and leave space around it, instead of filling every edge. That may suit a layout with a background and a caption, but our results do not count that approach. Decide whether the brief requires a photo that fills the canvas or allows space around it before using a size check to reject a file.

For example, a 1,800 × 2,400 image can fill the canvas after shrinking to 60% of its original dimensions. The result is 1,080 × 1,440, leaving 90 pixels of height to crop in the resized image. That arithmetic establishes a candidate, not a safe composition. A handle, label or product edge could sit in the area that would disappear.

This is why “large enough” and “ready to publish” should be separate decisions. We analyzed dimension records, not the image pixels. A secondary image could be a close-up, packaging view or instruction panel. Someone still needs to open the actual file and decide whether it communicates the intended product story.

We also specified a square canvas of 1,080 × 1,080 before examining the results. With that condition, 5,107 main images fell short and 2,544 of those listings had a qualifying secondary image: 49.8%. The portrait result was therefore not the only chosen shape for which the paired comparison found a substantial group of alternatives.

Some listings refer to exactly the same main image and set of other images. Counting each such combination once reduced the portrait comparison to 5,204 combinations with a main-image failure. Of those, 2,641 had a qualifying secondary image, or 50.7%. Repeated image combinations did not account for the headline pattern.

Expanding the analysis to every eligible marketplace listing in the archive produced a lower share: 14,890 of 37,934 main-image failures, or 39.3%. That comparison covers a different population. Products can recur across marketplaces, and their mix differs. The variation is a reason to check your own files, rather than adopt 51.3% as an expected result for your catalog.

Each row answers its own question. The square test changes the chosen canvas while keeping the eligible listings. The image-combination check changes how repeated references are counted. The all-marketplace check changes the source population and can count a product in several markets. Reading the rows as a single ranking would lose those differences.

For someone managing a catalog, the important point is that many listings still had an alternative after repeated image combinations were counted only once. The exact proportion does not necessarily carry over to another catalog. A team’s products, photography and filing practices may differ from this archive, so the practical recommendation is an inspection step rather than a forecast of how many usable replacements the team will find.

For a business briefing a creator, the finding suggests a small change to the asset handoff. Provide access to the approved originals for the relevant product, rather than assuming the main website photo defines the available options. Ask the creator to shortlist files for the intended layout and explain what each one would show.

For the creator, make the request specific enough to answer in one handoff: the exact product variant, the approved original files, any restrictions on their use, and the person who can approve a changed crop. For the business, retain the file names when sharing the gallery so the selected view can be traced back to its source. Use those file names when discussing a different crop or requesting approval.

Start with the message. A post introducing the whole product needs a different view from a post explaining a fastening, texture or assembly step. A large detail shot can be useful for the second job and unsuitable for the first. The size check should help the creative decision, not choose the subject of the post.

Next, check the actual file dimensions against the planned canvas. Then open the candidate and test the crop with the intended text or graphic elements in place. Look for the product edges, identifying details and any information the viewer needs. Review the exported result as well as the source file, because the layout is what customers will encounter.

Keep a short reason alongside the selection. “Enough pixels; label remains readable after crop” tells a teammate more than “use photo two.” Where the crop fails, record the missing view or framing. That gives a photographer or brand contact a concrete request instead of a vague demand for something bigger.

The handoff can also return a useful rejection. If a secondary file is large enough but shows the wrong color, keep that reason. If a detail shot works for one post but cannot introduce the whole product, save the distinction. It helps the next brief start with what the asset actually shows.

This workflow applies to your own approved assets and the permissions for the campaign. A photo appearing on a public product page does not by itself establish permission to reuse it. The dataset’s research license is separate from the rights needed for an individual business’s marketing materials.^[4][6]

Consider a fictional refillable-bottle brand preparing a portrait product introduction. Its chosen canvas is 1,080 × 1,350 pixels. The designer has a 900 × 1,200 main image and a 1,800 × 2,400 alternative. These invented files illustrate the calculation; they are not images we inspected in the archive.

The main image needs a scale of 1.2 to cover the width. At that scale it would become 1,080 × 1,440, so it would require enlargement and a 90-pixel vertical crop. It fails this study’s no-enlargement condition. That result does not decide whether enlargement would be acceptable for another creative brief; it answers the rule chosen here.

The alternative needs a scale of 0.6. It also becomes 1,080 × 1,440, but by shrinking. It passes the pixel condition and still requires the same total vertical crop. If centered, 45 pixels would be removed from the top and 45 from the bottom of the resized image.

Suppose the folder also contains a 2,000 × 1,000 landscape image. Its wider edge does not make it a better answer to this portrait brief: reaching 1,350 pixels high would require enlargement. This third fictional file illustrates why sorting candidates by their longest edge alone would miss the relevant condition.

The designer now checks whether the cap, bottle base and label remain visible. Suppose the lower edge of the label would be cut off. The numerical pass does not override that observation. The designer can try a different crop position and inspect it, choose another approved view, or ask for a file with more useful space around the bottle.

If the revised composition works, the team records the selected file, crop position and final approval. If it does not, the request for new material becomes precise: a full-bottle view with enough room above and below for this portrait layout. The selected file must work for that message as well as that canvas.

Use the following fields for one product and one intended layout. Duplicate the candidate entry for each file you review. The downloadable package contains the complete blank sheet and the worked example, with no sign-up required.^[5]

Brief: product and version ______; campaign or post ______; intended message ______; person reviewing ______; date ______.

Layout: width ______ pixels; height ______ pixels; intended placement ______; product features that must remain visible ______; planned text or graphic elements ______.

Candidate file: name or link ______; source and permission reference ______; original width ______ pixels; original height ______ pixels; main image or alternative ______; view or detail shown ______.

Size calculation: layout width ÷ file width = ______; layout height ÷ file height = ______; larger ratio = ______. A ratio of 1 or less meets the no-enlargement condition for filling this canvas. A larger ratio needs enlargement. Unknown dimensions stay “unknown” until checked.

Crop check: resized width ______; resized height ______; horizontal pixels to remove ______; vertical pixels to remove ______; crop position ______. Open the file and mark whether the product, version, necessary details and intended message remain correct ______.

Export review: final file ______; visible quality and legibility checked ______; permission for this use confirmed ______; approved by ______; approved date ______.

Decision: use this candidate / inspect another file / change the layout / request new material. Reason ______. If requesting material, describe the missing view, framing or detail ______.

Keep the finished sheet with the selected asset so another person can understand the choice. A recorded size pass is useful evidence for the handoff, but the final decision belongs to the inspected composition and approved use.

We read all 16 listing-metadata files and the original-dimension image table in Amazon Berkeley Objects. The producer describes the archive as 147,702 listings and 398,212 unique catalog images.^[1][2] We used the listing’s recorded main and other image references, not the archive’s smaller display-image versions.

The primary unit was a unique item ID and marketplace-domain pair with the exact domain amazon.com. Of 19,616 such listings, 74 lacked a main-image reference and were excluded. The remaining 19,542 had the required dimensions, giving 99.6% coverage. Missing data were not counted as small images.

Within each listing, we removed repeated secondary references and kept the main image in its stated role. The paired table asks whether the main meets the chosen dimensions and whether any distinct recorded secondary does. Three repeated secondary references were removed in the primary group. Conflicting image or listing identifiers would stop the calculation rather than be resolved by choosing a convenient row.

The portrait and square dimensions, gallery-count groups and supporting comparisons were fixed before the outcomes were examined. The listing files carry June 16, 2021 last-modified dates and the image table June 17; those dates identify the archived objects, not the precise date each photograph was taken. These listing counts are not a random sample of today’s businesses, so we report exact counts and shares rather than population confidence intervals. The accompanying package provides the calculation and source details.^[3][5]