Madar AI is here
Today we’re opening Madar AI to the first wave of MENA mobile subscription apps. After a long stretch of building in the open with a handful of design partners, we’re confident enough in the foundation to put a real signup form behind it.
This is the post I wanted to write for two years.
What it actually is
Madar AI is an AI Growth Operating System for mobile subscription apps in the MENA region. That’s a long phrase, so let me break it down the way I’d explain it to a founder over coffee in Istanbul or Riyadh.
Your stack already produces every signal you need to grow. AppsFlyer or Singular knows your installs. RevenueCat knows your subscriptions. Meta, Google, TikTok, and Snapchat know what you spent. App Store Connect and Play Console know what Apple and Google saw. Your product analytics tool knows what users actually did.
The problem isn’t data. The problem is that all five of those tools tell you a different version of yesterday, none of them agree on what an “install” is, and none of them tell you what to do next.
Madar AI sits on top of that mess and does five things:
- Reconciles the contradictions between sources, with the math written down so you can audit it.
- Forecasts subscription LTV by cohort, market, and acquisition channel — with a v2 GBM model now serving 65% of traffic.
- Recommends the next action, ranked by projected revenue impact, in plain language.
- Localises every prediction to MENA realities — Ramadan distortion, Iyzico vs Stripe vs IAP, Arabic creative performance, regional pricing power.
- Speaks your language — English, Arabic, and Turkish in the dashboard, with the RTL flip handled properly instead of as an afterthought.
Why we built this
I spent six years at Zyrix building growth infrastructure for apps that scaled inside MENA. The pattern I kept seeing was the same: founders with great products getting their growth advice from playbooks written for North American B2B SaaS, then quietly losing 30-40% of their potential revenue to assumptions that don’t survive contact with a Friday afternoon in Jeddah.
The big growth platforms aren’t bad. They’re just built for a different problem. Their benchmarks come from US apps. Their LTV models don’t know about Ramadan engagement spikes. Their attribution doesn’t reconcile with how iyzico settles. Their recommendation engines treat “spend more on Meta” as a universal answer.
MENA apps need a tool that starts from MENA realities. Madar AI is that tool.
What’s live today
- Forecasting — 30/60/90-day subscription LTV with model-version-aware A/B routing between a dict-lookup v1 baseline and the GBM v2 production model. Drift monitoring with auto-rollback if the population stability index crosses critical thresholds.
- Attribution reconciliation — pulls AppsFlyer / Singular / AppsFlyer Conversion Studio side-by-side, surfaces the deltas, and explains which source you should trust for which decision.
- Benchmarks — by category, by country, by monetisation model. No more guessing whether 1.4% trial-to-paid conversion is good for a meditation app in Saudi Arabia in March (it’s slightly above the median).
- Copilot — an LLM that has read your data, your benchmarks, and the playbooks. Ask it “why did my retention drop on Tuesday” and it’ll tell you in two sentences with the SQL it ran behind it.
What’s next
The roadmap that gets a lot of internal attention right now:
- Track-tier billing with quotas (Free / Pro at $499 / Enterprise at $1,999), shipping next sprint.
- ASO competitor tracking with KSA App Store compliance built in.
- Bayesian sequential testing for app store experiments, so you stop running them for two weeks “to be safe” when the answer was clear on day 4.
- Fraud signals — install fraud detection tuned for MENA install farms.
If any of that sounds like a tool you’ve been waiting for, I’d love to hear from you. There’s a signup form on the marketing site, but honestly, send me an email — I read all of them.
— Mehmet