Almost always the photography. Layout, type, and colour can be excellent, and one stock image of a diverse team laughing at a laptop will undo all of it. Images carry more of the credibility signal than any other element on a page.
This is the note we give most often on design reviews, and it is rarely what people expect to hear. They came asking about the hero layout. The problem is the picture inside it.
Photography is also where budgets get cut first, which is understandable and usually a mistake. It is the one element visitors read as evidence rather than decoration.
Specificity. Real places, real people, real products, real work. Credible images look like they could only belong to your business. Generic images look like they could belong to anyone, and visitors read that as a signal about the business itself.
The test we use is whether an image could be swapped onto a competitor's site without anyone noticing. If it could, it is doing nothing for you beyond filling space.
Lighting and consistency do the rest. A set of photos shot in the same conditions reads as deliberate. The same photos gathered from different sources, with different colour temperatures and crops, reads as improvised no matter how good each one is individually.
Sometimes, but not for anything that claims to represent you. Stock works for abstract or supporting imagery, textures, backgrounds, and illustrative concepts. It fails for team photos, office shots, customer stories, and anything a visitor might reasonably assume is real.
The specific failure is trust. If a prospect recognises your team photo as stock, they now wonder what else is not real. That doubt is expensive and it lands right where you were trying to build confidence.
Where stock genuinely helps is at the edges. A blog header, an abstract background behind a section, a conceptual illustration. Nobody is deceived, and nothing important rests on it.
If you must use stock for a while, choose quieter images. Understated photographs age better and draw less attention to their own origin than the heavily staged ones.
When your business sells trust, a physical product, or a place. Professional services, hospitality, property, manufacturing, and anything with a team the client will actually meet all justify a shoot. Software companies with no physical product often do not.
The economics are usually straightforward. A single day of shooting produces enough material for a homepage, an about page, a careers page, and a year of social content. Compared to the cost of a rebuild, it is a small line item with a long life.
Plan the shot list against the pages you are building rather than shooting first and choosing later. Knowing you need one wide hero image, four portraits, and six process shots makes for a far more productive day than turning up hoping for good material.
When a shoot is genuinely out of reach, our advice is to use fewer images rather than filling gaps with stock. An honest, restrained page beats a padded one.
Fewer than most sites use. Every image should either prove something, explain something, or give the eye somewhere to rest. Images added because a section looked empty are the ones we remove first, and pages almost always improve.
The weight cost is real. The HTTP Archive's 2025 Web Almanac found images to be the largest resource type on a median page, at 1,058 KB on desktop and 911 KB on mobile. That is more than every other resource type carries.
There is a compositional cost too. When everything is illustrated, nothing stands out. A page with two strong images gives them both room to work, and the reader notices them.
Our default on hero sections is one image doing one job, which we go into in our guide to designing a hero section that converts. The hero is not the place to hedge.
Modern formats for photographs, with appropriately sized versions for each screen. Serving one large file to every device is the most common cause of a slow, image-heavy page, and it is also one of the easiest things to fix on an existing site.
The principle is to never send more pixels than the layout will display. An image shown at 600 pixels wide does not need a 3000 pixel file, and the difference in bytes is enormous.
Compression settings matter more than format arguments. Most photographs survive fairly aggressive compression without visible loss, and the saving is far larger than anything gained by agonising over which modern format to choose.
We covered the mechanics properly in our guide to optimising images for web performance. On most sites this work alone moves the performance numbers more than anything else available.
Heavily, because the largest image on screen is usually the element Largest Contentful Paint measures. Google's web.dev documentation defines a good LCP as 2.5 seconds or less at the 75th percentile, and a slow hero image is the most common reason sites miss it.
One mistake dominates. Google's documentation is blunt about it, stating that you should never lazy-load your LCP image, because it will always lead to unnecessary resource load delay and have a negative impact on LCP.
The opposite move helps. Google's guidance recommends setting a high fetch priority on an image element you believe is likely to be the page's LCP element, so the browser starts fetching it early rather than treating it as one request among many.
Lazy loading is still right for everything below the fold. The rule is simply that the first big image gets priority and everything after it waits its turn.
A short description of what the image conveys in context. Not a caption, not a keyword list, and not the filename. If the image is purely decorative, an empty alt attribute is the correct answer rather than a description nobody needs.
The web does this badly at scale. The 2025 Web Almanac's accessibility chapter reports that 69% of images pass the Lighthouse audit for having alt text, meaning roughly a third still fail a basic check.
Quality is a separate problem from presence. The same chapter found that approximately 8.5% of image alt texts end with common file extensions such as .jpg or .png, which tells a screen reader user nothing at all.
Context decides the wording. The same photograph might be described differently on an about page than in a case study, because what it is doing on the page has changed. We looked at where automation helps and where it does not in our piece on whether you should let AI write your image alt text.
Decide the rules before you need them. Aspect ratios, treatment, colour grading, and cropping conventions should all be written down while the site is being built, rather than negotiated from scratch every time somebody on the team adds a new case study or team member two years later.
Aspect ratio is the one that causes the most damage. When a CMS expects a consistent shape and receives whatever the marketing team had to hand, cards break and layouts shift. Fixing that later means recropping everything.
Naming and organisation matter almost as much. A library where nobody can find the original file ends up with someone downloading a compressed version from the live site and re-uploading it, which quietly degrades quality over time.
We hand over these conventions as part of every build, because the site only stays coherent if the people adding to it know the rules.
Audit the images on your three most important pages and ask, for each one, whether it proves something a visitor needs. Remove the ones that do not. Then check that the largest image on each page is properly sized and not lazy loaded.
Those two passes cost an afternoon and usually improve both credibility and speed. It is rare that a site needs more images. It is common that it needs better ones and fewer of them.
If a shoot is warranted, plan it against the pages you are building rather than the other way round. Across our own projects the average PageSpeed score is 98 and the average conversion lift is 3.5x, and honest, well-sized photography contributes to both.
If you want a second opinion on whether your imagery is helping or quietly undermining the site, we are happy to take a look. Get in touch through phoenix.studio and we will come back to you within 48 hours.
Tell us where you want to go. We'll tell you how we'd get you there.