A team we know flew into the Gulf with a neat integration plan: pick the dominant super app, ship one mini-program, expand later. On day two the plan was dead. There wasn't one host to pick — there were several, none of them interchangeable, each with its own wallet, its own login, its own review queue and its own idea of what a third-party service is allowed to do.

Calling it 'the Middle East market' is the first mistake. It's a handful of medium-sized markets that happen to share a language and a climate, and the platforms inside them behave like separate countries. None is open in the way a web browser is open. You don't publish to the region; you negotiate with hosts, one at a time.

Several walled gardens, one map

What surprises people isn't the number of apps, it's how complete each one tries to be. A regional super app here typically bundles messaging, payments, a national-ID-backed login, ride or food verticals, and a mini-program container of its own. It has little incentive to make that container portable, because the whole point is to keep the user — and the transaction — inside. So you get parallel stacks: one identity system, one wallet, one review policy, one catalog, repeated across each host.

Arabic makes the gap wider than a translation task. Right-to-left layout isn't flipping a stylesheet; it's mirroring navigation, rethinking which side a back arrow belongs on, handling numbers and Latin brand names that stubbornly stay left-to-right, and re-testing every gesture. Then there's settlement. Local payment schemes and bank-linked wallets dominate, and hooking into them means a commercial relationship, not an API key. Add data-residency rules that say certain records can't leave the country, and 'just host it in one region' stops being an option.

The honest read: these aren't emerging versions of a global platform waiting to converge. They're mature, defended ecosystems, and the walls are a feature, not an accident.

The honest tradeoff

Here's the part the pitch decks skip.

ConstraintWhat it looks like on the groundWhat a shared runtime absorbsWhat still needs local work
Arabic RTLLayout, icons and gestures mirror; numbers stay LTRBidirectional layout, logical propertiesCopy, tone, and a native reader's final pass
Local payment railsWallets and bank-linked schemes per marketOne checkout interface, several adaptersEach provider contract, settlement, refunds
IdentityNational ID or telco login behind each hostA single session abstractionConsent wording and KYC per country
Data residencySome data must stay inside the borderConfigurable storage regionLegal review and local hosting deals
Host reviewEach super app runs its own queueOne build, multiple submission targetsSeparate appeals and category rules
DiscoveryClosed catalogs, no shared searchA portable deep-link schemeRelationships with each host's team

Now the cost. Each market on its own is small, and taken together they're still fragmented — different regulators, different payment mixes, different dominant hosts. A studio that adapts its mini-program host by host is paying the same engineering bill four or five times for a fraction of the addressable users each time. We've watched teams spend a whole build budget on the second integration and then quietly shelve the third. On paper the region looks big; line by line, arithmetic is what kills it.

And the hosts don't especially want you. The stronger ones are building the same services in-house, because owning a vertical is worth more than collecting a revenue share from someone else's. That flips the usual platform logic: you're not filling a gap they can't serve, you're asking for shelf space next to their own product. Some will say no, some will say yes and then deprioritise you, and a policy change can close the door without a roadmap you can plan against.

There's a compliance tail too. Data residency, KYC wording, local licensing for anything touching payments — these are per-country legal questions, not engineering ones. Get one wrong and it isn't a bug, it's a blocked launch.

What to share and what to keep local

So split the work deliberately. The stuff that genuinely repeats belongs in one layer: bidirectional layout, a session abstraction that doesn't care which ID provider answered, one build that can be submitted to several catalogs, a deep-link scheme that survives being opened from different hosts. That's where a shared runtime earns its keep, and it's the layer Cross Mini App is aimed at — not at erasing the differences, but at stopping you from re-solving the same six problems in every country.

The rest stays local, and pretending otherwise is expensive. Payment contracts are negotiated market by market. Arabic copy needs a human who writes it, not a mirrored string. Consent screens, refund policies, support hours and the relationship with each host's review team are all local jobs. Hire for those, budget for those, and don't let a platform abstraction promise them away.

The teams that do well here tend to be unglamorous about it: they pick one host and one service, get the residency and payment questions answered in writing, and only then talk about a second market. The region rewards patience far more than it rewards a big launch. There's no single door to kick open — just several, each with someone on the other side you have to convince.

Sources