I own right-to-left design across five brands. Not a checklist at the end of a sprint — a system that decides, for every screen, what mirrors, what holds still, and why.
Most products treat RTL as a post-production step: design in English, flip at the end, translate what's left. The result is a product that technically reads right-to-left and feels wrong in a hundred small ways — reversed Latin buttons, scrambled artwork, phone numbers broken in half.
Half the people our brands serve read Arabic first. In Saudi Arabia, Arabic isn't the localization — English is. So the mirror can't be an afterthought. It has to be a system with rules, exceptions, and taste.
Every element on a screen gets resolved down a ladder. The design system answers first; geometry is the last resort, never the default. Flipping a component that already has an Arabic version doesn't give you Arabic — it gives you a backwards Latin component.
The craft of RTL isn't in the flipping — it's in knowing what must not flip. Every rule below exists because getting it wrong shipped a bug somewhere first.
Bidirectional text has failure modes most designers never see. In an Arabic sentence, a phone number is a left-to-right island in right-to-left water. Let it wrap mid-value across a line break and it reads as two different numbers.
So the rule, documented and enforced in every handoff: numbers never wrap mid-value. If a number can't fit on the line with its preceding text, the line breaks before the number — never inside it. It's one line in a spec document. It's the difference between an OTP screen someone trusts and one they re-read three times.
A Saudi customer opening Virgin Mobile in Arabic should never sense the product was designed in English and flipped for them. The typography is chosen for Arabic. The components were built with an Arabic self. The numbers hold their shape.
That's what owning RTL across five brands actually means: not flipping screens — building the system that makes every screen already know its second language.