Your delivery radius decides this, not nationality
Start with the map, not the flag. About two out of every three people living in Kuwait are not Kuwaiti citizens, and in the blocks around Salmiya, Hawally, Farwaniya, Mahboula and Jleeb that share is far higher. If your delivery radius is fifteen minutes wide, a Kuwaiti-only audience deletes most of the households inside it. Very few restaurants can afford that. The honest answer for almost every operator is both, weighted by what your neighbourhood actually looks like rather than by the customer you imagine.
Weighting is where the money is. A cafe on the Gulf Road or inside a Mishref compound sells to a different street than a shawarma counter in Jleeb, and both are real businesses. Pull your last ninety days of Talabat and Deliveroo orders, sort them by area, and you have your real audience map for free. If seventy percent of your tickets land in Hawally and Salmiya, that is your budget split. Spend against the streets that already order, then test one adjacent area at a time with 50–80 KD.
There is also a spending difference worth naming plainly, without stereotyping anyone. Settled family households order bigger trays, especially Thursday and Friday night. Single-worker households order smaller, far more often, and are much more price-sensitive. Neither is a better customer. They just need different offers: a 12–18 KD family bundle for one, a 1.5–2.5 KD daily meal deal for the other. Run both and let the offer sort people, instead of trying to sort them yourself before they ever see the ad.
Ready to start your Shopify store?
Start a free trial and try the platform for yourself.
You cannot target nationality, so target language, area and behaviour
Meta, Snapchat, TikTok and Google do not let you target by nationality, ethnicity or religion, and anything close to it is restricted and routinely rejected in review. Stop looking for that checkbox. What you can target is location down to a radius, language, device and operating-system language, interests, and behaviour such as people who watched your last video or ordered from you before. Those signals get you ninety percent of the way there and they survive a policy change. Anyone selling you a nationality-targeted campaign in Kuwait is selling you something that does not exist.
Language is the strongest lever you actually have. Run every campaign as two ad sets: one Arabic, one English, same offer, same budget, and let delivery find who responds. Within a week the cost per order tells you the split, and it is often not what the owner guessed. A Kuwaiti-owned burger brand in Salmiya can easily find its cheapest orders come from the English creative, because the phone's language, not the passport, decides which ad feels readable at eleven at night when someone is hungry.
Watch the traps. Do not write one ad that mixes Arabic and English in the same sentence to save production money, because it reads as amateur in both. Do not machine-translate either; a stiff Modern Standard Arabic caption feels foreign in a Kuwait feed, and Kuwaiti phrasing beats it almost every time. And keep the landing step in the same language as the ad. An Arabic ad that drops someone onto an English-only menu with a broken KNET checkout loses the order you already paid for.
Your menu picks your audience before any ad does
Cuisine self-selects, and it does it more accurately than any targeting panel. A Filipino silog counter, a Keralan fish thali, a Levantine mashawi grill and a Kuwaiti machboos house each pull their own crowd without a single demographic setting being touched. That is your advantage, not your limit. Cook and photograph the food honestly, write the menu in the language the dish belongs to plus English, and the right people find it. The mistake is diluting the concept so it sounds universal, which is how a restaurant becomes invisible to everyone at once.
Then keep the data. Talabat and Deliveroo take 20–30% and never hand you the customer's name, language or area, so every lesson about who your audience is stays with them. Your own ordering channel does the opposite: a bilingual storefront on a platform that ships Arabic and English with KNET checkout out of the box tells you which language each person chose, which blocks they order from, and who comes back. After ninety days that file is a better audience than anything you can buy from an ad platform.