Creative localisation

Dressing your window for a local audience is the most commonly skipped piece of this entire job. It's skipped because it's genuinely expensive, not because it doesn't work.

What can actually be redressed per country

Window photos are uploaded separately for each sign-language, and so is your video, and so is every piece of your writing.

On one side of the centre, a shared warehouse now lets you manage photos for reuse across your main listing, a specially-dressed window, live tests and events, which makes keeping several redressed sets going noticeably less painful than it used to be.

So the ability to do this has never really been the limit. Actually producing it is.

Why it gets skipped

Each redressed set of window photos is a design job, a translation job, a review job, and a permanent piece of ongoing upkeep. Ten countries means ten sets that all need updating the moment your interface changes. Writing can be updated by one person in an afternoon; photos can't.

The common result: writing translated into a dozen languages, window photos left in the original language everywhere. A shopper in a local branch sees a translated pitch line sitting above photos in a language they don't speak, which reads exactly like a shopfront assembled by someone who wasn't thinking about them at all.

The cheap version

Most of the actual value sits in the captions, not the underlying artwork.

Translate the captions, keep the artwork. The caption is the text a visitor actually reads in the search answer. Swapping caption text over the same background costs a fraction of a whole new set, and captures most of the benefit. If you only do one thing, do this.

Redress the interface shown in the photo, or crop around it. An interface in the wrong language sitting under a translated caption undercuts the caption itself. If your shop is genuinely localised, capture the photo in that language. If it isn't yet, crop to parts where the language isn't visible at all.

Redress only the opening few frames. They're the ones showing up in the search answer, carrying the actual decision. The rest of the set can stay in the original language for a long time before anyone notices.

Prioritise by traffic, not by how many languages you can cover. Your three biggest non-local branches, done properly, beat a dozen done shallowly.

Where it goes beyond simple translation

Some places genuinely need different content, not just translated content:

A right-to-left branch reverses the reading order entirely. A sequence of photos designed to be read left to right tells its story completely backwards there, and caption alignment is simply wrong. That's a layout change, not a translation.

Text expansion breaks layouts. Some languages routinely run substantially longer than others. A caption box sized for the shorter original overflows, or shrinks itself into illegibility.

Imagery itself carries meaning. People, places, money, sport, food and the seasons all read differently depending on where you are. A winter scene is simply wrong in half the world at any given moment.

Social proof doesn't travel. A press mention or an award that means something in one place means nothing at all in another.

That culturally adapted creative outperforms merely translated creative is reported consistently by experienced shopkeepers and shows up widely in industry material, the specific size of the improvement quoted anywhere is always a single case, specific to one category. Treat the direction as solid and any specific number as pure illustration.

What tends to go wrong

Translating the writing and leaving window photos untouched. The most common half-finished localisation, and the most visible one.

Translating captions over an interface still in the wrong language. Undercuts itself completely.

Ignoring text expansion. A caption overflowing a box sized for a shorter original.

A left-to-right sequence in a right-to-left branch. The whole argument runs backwards.

Redressing a branch you won't actually keep up. A stale, out-of-date set is worse than none at all, because it's visibly abandoned.

Treating this as purely a translation task. It's a design task that happens to have a translation piece inside it.