The Brief
Azar operates in 230 countries. Vividi, its sister app, runs in 29, spanning Asia, Europe, the Middle East, North America, and Latin America. The challenge wasn't just making ads. It was building a system to figure out what message, visual, and format actually moves people to register, and more importantly, which ones bring in users who will actually use the app.
Visual Testing - What visuals work before we even test messaging?
Message Testing - Which message angle actually converts?
With visual direction confirmed, the next question was, what do we say?
Insight : What the data actually said
People don't open a video chat app to build relationships. They open it to enjoy a moment, kill some time, say something they can't say to their friends, have a conversation worth having. The winning messages weren't about what Azar could become for someone. They were about what it is, right now, in that moment.
Light topics. Casual energy.
Entertainment first.
That's the creative insight. And it's now the foundation for everything that comes next.
Intro CTA Testing In Progress
With winning messages confirmed, the next variable is the intro CTA, the first 3–5 seconds that determine everything.
Tooling - Built to Move Faster
A web based ad review tool that automatically translates video scripts, scores expected performance, and flags legal risks by country law, all in one pass. Exports directly to PDF or Notion.
Before: translation, performance review, and legal check were three separate manual steps, repeated across 80 - 100 ads per month.
After: one upload, one click, full report.
Script Translator & Formatter
Paste a script, select target markets, get localized versions with subtitle line breaks and Arabic phonetics, ready to paste directly into the edit file.
Before: 3 minutes per country × 10+ markets, frequent copy paste errors in right-to-left scripts.
After: 10 seconds. Zero errors.
Both tools built through vibe coding, a problem worth solving, not a spec worth waiting for.
The Work
A selection of assets across markets. Each piece was designed with a specific audience, message angle, and platform behavior in mind, not just made to fill a brief.
Gym mirror selfie - Brazil & Algeria. Male AI model outperformed female.
Product UI mockup - UK/IT/DE/FR. Shows the card swiping UX in the first 3 seconds.
"Every app is too serious" - Vividi. Positions the app as the lightest option in the category.
Dog content - Morocco & Philippines. Low-pressure, warm, consistently high CTR.
Late night loneliness - Algeria. Localized script, AI-generated talent, Arabic subtitles formatted by custom tool.
Results
Across markets tested, top-performing creatives drove measurable improvement across the full funnel, from registration rate to in app engagement. The three winning messages (Killing Time alone, Killing Time in group, Talk Unfiltered) consistently outperformed on the metrics that actually matter: not just who clicked, but who stayed and used the app.
The broader campaign also validated a channel strategy: acquiring female users through Rest of World markets (Morocco, Algeria, Philippines, Brazil) proved significantly more cost efficient than European markets, delivering better results across all six priority metrics simultaneously, from acquisition cost to retention to in app match quality.
The testing framework and tooling built during this project now form the foundation for ongoing creative iteration across both Azar and Vividi.
190+ assets produced across 8 markets
7 message angles tested → 3 winners identified
Killing Time & Talk Unfiltered confirmed as core message framework
RoW strategy validated: female acquisition significantly cheaper than EU markets
2 custom AI tools built
Female ratio adopted as primary creative KPI
What's Next
The gym selfie worked. The dog content worked. But that just opened a bigger question, what else works and why? The next phase expands in two directions at once. First, the visual universe. If aspirational and approachable are the two modes that drive performance, the hypothesis is that other interest-based visuals can tap into the same logic, just with different entry points. Currently in exploration: foodies, travelers, festival goers, campers, photo enthusiasts, fashionistas. Each a potential "wanna be" hook for a different kind of user.
Second, the ad structure itself. So far, intro has been the primary variable, and that was intentional. But intro alone doesn't tell the whole story. The next rounds of testing will move into body and outro as independent variables, while simultaneously running visual and message tests within the intro. The goal is to eventually have a formula, not just "this message works," but "this message, delivered this way, to this audience, in this structure."
