Takeaway.com, Deliveroo and Uber Eats, managed in both languages

Because dish names are search terms. Someone typing boulettes will not be matched by an item called balletjes, and the reverse holds in Dutch-speaking postcodes. A listing written in one language simply has nothing there to match a whole set of queries it could have answered, which is a menu problem rather than a platform problem.

Brussels makes the point sharpest: one storefront, one delivery zone, two languages of customer. The usual shortcut, running the menu through a translation tool, produces item names nobody searches for and descriptions that read like an instruction manual. Menu copy is sales copy, so it gets written twice and properly, and you approve both versions, because you know your own dishes better than any translator does.

With one master menu and a controlled rollout, rather than piecemeal edits made on each platform in turn. Categories, item order, naming, combos and modifiers are decided once. Each platform then receives that structure in its own format, because description lengths, image requirements and promotion types are not identical across Takeaway.com, Deliveroo and Uber Eats.

The split runs deeper than the item list. An image is language-neutral; the name and description beside it are not. Platform-reported benchmarks put real dish photos at up to 20% more clicks and up to 32% more orders, and that lift depends on the words around the photo being written for the person reading them. The same applies to combos and modifiers: across the stores we manage in the Netherlands, rebuilding that structure has lifted basket size by 38%, and a full menu rebuild has roughly doubled menu conversion.

That is the Belgian version of the budget question, and it is invisible on a dashboard that reports one number per listing. A storefront serving a mixed catchment is really two demand pools sharing one page: people searching in French and people searching in Dutch. Spend applied to the listing as a whole treats them as one audience, and the half that converts poorly quietly funds the half that converts well.

So we measure per language version and per delivery zone, not per platform alone. A brand can rank respectably in French across a Brussels zone and be almost invisible for the Dutch terms covering the same streets, or the reverse in a Dutch-speaking town where the French version was added as an afterthought. Ranking still rewards the storefront that converts, and a version whose item names do not match the words people type will not convert, so budget behind it buys impressions that were never going to land.

Money then follows the version and the zone with evidence behind it, measured against your own baseline from before we started, rather than being divided three ways because there are three platforms. On the strongest storefronts we manage, that has meant return on ad spend of up to 14x. Offers are margin-checked first: platform-reported benchmarks of around 39% more order volume and 25% more order value on buy-one-get-one are only good news if the arithmetic works on your dishes, and only reach everyone if the promotion copy exists in both languages. The plan arrives weekly naming the platform, the language version, the zone and the offer, and it stays a recommendation until you approve it.