We supply the photography — what do we actually need, and when?
You need fewer images than you have pages, you need them specified by template slot and aspect ratio before anyone shoots, and the biggest waste is a production shoot booked before the templates exist.
Count slots, not pages
A redesign does not consume images by the page. It consumes them by the template slot. Every redesign built on a design system has a fixed set of page designs, and each one has a fixed number of image positions with fixed aspect ratios. Six templates can drive hundreds of URLs, and those hundreds of URLs do not need hundreds of unique photographs.
So build your shot list from the templates, not the sitemap. Ask the studio for the template list and work down it.
A worked case. A twelve page marketing site turns out to have five templates: home, a product page, a solutions page, an about page and a blog post. The home hero takes one wide image. The product template has one hero and three supporting images, and it repeats across three products, so that is twelve. The solutions template reuses the same three supporting images across all four solutions, so that is three, not twelve. About needs six portraits and one interior. Blog posts pull from a general library and can share. Total: roughly twenty-five images, of which two are hero-scale. Not one hundred. Twenty-five is a list a photographer can quote against in an email and a stock search you can finish in an afternoon.
- List the templates in the redesign, not the pages. There are usually five to eight.
- For each template, count the distinct image slots. Write down the aspect ratio of each one.
- For each slot, decide whether every instance of the template needs a unique image, or whether one image can serve all of them. Most can be shared.
- Sort the resulting list into three groups: hero-scale images, small supporting images, and portraits or object shots.
- That is your shot list. Count it before you brief anyone.
Art direction is cheap as a design decision and expensive as a rescue
The sequencing rule is simple. Anything decided before layouts are approved costs nothing. The same decision made after a shoot costs a reshoot or a redesign.
Four decisions are free while the design is still being made. Whether the site leans on people, product, place or abstraction. What crops and ratios the design system locks in, so any future image drops into the slot without a layout change. Whether text ever sits over an image, which decides overlays, contrast and how tightly the photographer can frame. Where the negative space goes, so headlines have somewhere to live.
The same four decisions made late look like this. A shoot delivered before templates exist, where nothing crops to the shapes the site actually uses. A landscape library bought for a site whose hero is a tall mobile-first crop, so every image loses its subject on a phone. Gradient scrims bolted on afterwards because the photos are busy exactly where the copy lands. A subject swapped out after brand sign-off, which means the shoot again.
The cheapest version of this is one sentence in an email before you book anyone: send me the slot dimensions and safe areas. In writing.
What not to pay for
Five image costs are usually waste, and one of them is a line item a studio can bill you for.
On that last one: handing you a shot list you can execute yourself earns us less than sourcing and art directing it ourselves. It is also faster for you and cheaper. A client-supplied library genuinely slows a studio down, because it arrives in the wrong shapes at the wrong time and someone has to reconcile it. The fix is not to take the work off you. It is to give you the slot list early enough that what arrives already fits.
- A full production shoot booked before design. Wait until slots and ratios exist, then shoot to the list.
- Unique imagery for every page. Repetition across a template family reads as consistency, not laziness.
- A stock subscription with credits you will not spend, when a handful of single licences costs less.
- Retouching an image that ends up cropped to a 280 pixel supporting slot.
- A studio billing hourly to source stock, when the shot list turns it into an afternoon.
When spending less is the right answer, not a compromise
An image-light design is a legitimate direction. A site with confident type, generous spacing and a real colour system needs fewer photographs than a weak one, because the layout is already doing the work photography would otherwise be hired to do.
Software and product companies are the clearest case. Interface screenshots, diagrams and typographic layouts usually outperform photography, and they cost nothing to shoot. If the product is the thing being bought, show the product.
B2B services with no premises a customer will ever visit are the second case. Real portraits of the people who answer the phone, plus abstraction or texture for the rest, beat staged office scenes every time. You do not need a photograph of a meeting.
The counterexample, so this is not a blanket rule: if the thing being sold is a place, a physical product or a person’s craft, photography is the product page and cutting it is false economy. A restaurant, a manufacturer, a clinic. Spend there.
Licensing, file weight and AI images
Check the licence covers paid ads and social, not just the website. This is the practical failure we see most often with stock. The image is licensed for the site, marketing runs it in an ad set six weeks later, and nobody notices until someone else does. Sort it at purchase, when it costs a small upgrade rather than a takedown.
File weight and format matter more than resolution. An oversized original is the usual cause of a hero that takes a second to appear. Modern build stacks resize and convert images automatically, so send the largest clean version you have and let the site generate what it serves. Do not send a 40MB TIFF for a slot that renders at 1600 pixels wide.
AI-generated imagery is useful for texture, backgrounds and abstraction, where nobody is checking it against reality. It is unreliable for people, hands, products and anything a customer will compare to the real thing. Use it where the image is decoration, avoid it where the image is evidence.
Three questions to send the studio this week
These three answers turn imagery from a vague obligation into a countable task. Ask for them in writing before you brief a photographer or open a stock site.
The third question matters most. If the answer is that the layout collapses without every image, the design is too dependent on photography and it is better to know now than in the week before launch.
- What are the image slots and aspect ratios for each template, and can I have them before I brief anyone?
- Which images repeat across a template family, and which have to be unique?
- What happens to the layout if I only supply half the list?
Frequently asked questions (FAQ)
How many images does a five page site actually need?
Usually fifteen to thirty, and only two or three of those at hero scale. The count comes from templates rather than pages: five pages typically means four or five templates, each with one to four image slots, and many of those slots can reuse the same image across instances. Build the list from the template slots and the number is nearly always smaller than the page count suggested.
Can we launch with placeholder images and swap them later?
Yes, if the crops and ratios are fixed in the design system, because a correctly shaped image drops into the slot without touching the layout. It is worth agreeing which slots are placeholder and which are final before launch, so nobody signs off a page thinking the picture is the real one.
Should we book a photographer before or after the design is approved?
After. Booking before design means shooting without knowing the aspect ratios, the safe areas for headline text, or how many unique images the site needs. In our projects, shoots commissioned ahead of the templates are the ones that get half used, because the frames do not crop to the shapes the site uses.
Is stock photography a problem?
Not inherently. Stock fails when it is generic in the exact place a visitor is deciding whether to trust you, such as a homepage hero or a team page. It works fine for supporting slots, blog headers and texture. The useful test is whether a visitor could tell the image was not taken for you, and whether it would matter at that point on the page.
Related articles
- Our copy isn't written yet — should we start the redesign anyway?Start — but only if someone on your side has a name, a deadline and time cleared for writing, because the thing that delays redesigns is almost never design, it is the copy nobody was assigned to write.
- Why did our new design look great in Figma and worse once it was built with our real content?Because the mockups were dressed with content that behaves better than yours does, and a layout that only works with ideal content is not a finished design.
- We have 400 pages and a redesign coming — how do we decide what to keep, merge, rewrite or delete?Give every page one of four verdicts before design starts: keep as is, merge, rewrite, or delete. Make delete the default, and sort by what a page earns you rather than what it costs to move.
- We were told our change is out of scope — is that fair or padding?Judge the request against what the studio agreed to produce, not against how difficult it sounds: restyling inside existing components is cheap and should rarely be refused, while new page types, components or content models are genuinely new work.