Annalytical Design | Anna Eunbyeol's Portfolio

The Brief

Azar operates across 230 countries. Vividi, its sister app, runs in 29. Together, they reach audiences across Asia, Europe, the Middle East, North America, and Latin America, each market with its own language, cultural context, and platform behavior.

The brief wasn't just "make ads." It was: figure out what message, visual, and format actually moves people to register, and build a system that could find that answer faster than intuition alone.

Creative Approach: Different markets. Different creative logic.

Rather than producing one size fits all assets, I developed a market segmented creative strategy based on platform behavior and cultural context.

Europe (UK, IT, DE, FR)

Product UI Mockup European audiences respond to clarity. Showing the actual app experience (the card interface, the invite mechanic) reduced friction and set accurate expectations from the first frame.

MENA & Latin America (Algeria, Brazil, Turkey, Saudi Arabia)

AI UGC These markets respond to people, not interfaces. AI generated influencer style content created an authentic, scroll-stopping feel, and gave us the flexibility to test message angles at scale without traditional production costs.

Rather than producing one size fits all assets, I developed a market segmented creative strategy based on platform behavior and cultural context.

Europe (UK, IT, DE, FR)

MENA & Latin America (Algeria, Brazil, Turkey, Saudi Arabia)

Rather than producing one size fits all assets, I developed a market segmented creative strategy based on platform behavior and cultural context.

Europe (UK, IT, DE, FR)

MENA & Latin America (Algeria, Brazil, Turkey, Saudi Arabia)

How We Tested: Every variable had a reason

Each round of testing was structured around a single question, isolating one variable at a time so results could actually mean something.

Step 1 Message Testing

Six core messaging angles were defined and tested as separate ad groups: Making Friends, Killing Time, Talk Freely, People Watching, In Control, and Finding Date. Spend was controlled per group to ensure comparable data.

Step 2 Structure Lock

Intro was the only variable. Body and outro were identical across all variants, so any performance difference pointed directly to the hook, not production style or visual treatment.

Step 3 Visual Testing

Top performing messages got a second round: same script, different visual formats, influencer to camera, gym mirror selfie, situation-based visuals, to separate creative format from message strength.

Step 4 Scale

Every concept that passed the brief became a production matrix. One creative idea translated into visual variations, three aspect ratios (vertical 9:16, horizontal 16:9, square 1:1), and two durations, a 6-second bumper and a 10-second+ performance cut. Multiply that across platforms (Google, YouTube, TikTok, Meta) and markets, and a single approved concept could generate 15–20 final files before it ever ran. Total output for this project: 190+ deliverables across 8 active markets.

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 variants. The hook was the aspirational vibe, not the gender.
Product UI mockup: UK. Designed to show the card-swiping experience without overwhelming the viewer in the first 3 seconds.
Late night loneliness angle: Algeria. Localized script, AI-generated talent, Arabic subtitles formatted by custom tool.
"Every app is too serious": Vividi. Positioned the app as the lightest option in the category.

What We Found

The Wanna-Be Effect

The gym mirror selfie concept outperformed everything Not because of the setting, but because of the energy. Aspirational, authentic, "I want to be that person" visuals drove higher CTR and better downstream metrics than scripted testimonials. Real-feeling > polished.

Light intent, better funnel

Users who came in through low-pressure, entertainment-first messaging ("kill some time," "every app is too serious") converted better than those driven by strong intent hooks ("make real friends," "find someone to date"). The more serious the promise, the harder the app had to work to deliver, and it didn't always land.

Misaligned intent is a funnel killer

Dating-adjacent messaging drove strong click-through but fell apart at registration. Users arrived expecting Tinder, found a video chat app, and left. The message was doing its job, the wrong job.

Female ratio as a creative signal

People Watching had competitive CPR and strong CTR, but female ratio dropped to under 8%. A high-performing ad that brings in the wrong audience isn't a win. Gender ratio became one of our primary signals for message-audience fit, not just an afterthought metric.

Format ≠ Message

Locking body and outro while testing intros proved that hook format and hook message are different variables. The same message delivered as a talking-head vs. a text overlay vs. a situation based visual produced meaningfully different results, which became the basis for the next test round.

The Wanna-Be Effect

Light intent, better funnel

Misaligned intent is a funnel killer

Dating-adjacent messaging drove strong click-through but fell apart at registration. Users arrived expecting Tinder, found a video chat app, and left. The message was doing its job, the wrong job.

Female ratio as a creative signal

Format ≠ Message

The Wanna-Be Effect

Light intent, better funnel

Misaligned intent is a funnel killer

Dating-adjacent messaging drove strong click-through but fell apart at registration. Users arrived expecting Tinder, found a video chat app, and left. The message was doing its job, the wrong job.

Female ratio as a creative signal

Format ≠ Message

Results

Across markets tested, top-performing creatives drove measurable improvement in registration rate and female user acquisition, the two metrics that matter most for Azar's ecosystem health.

The testing framework, production workflow, and custom tooling built during this project now form the foundation for ongoing creative iteration across both Azar and Vividi's active markets.

190+ assets produced across 8 markets

6 message angles tested, ranked, and actioned
2 custom AI tools built to reduce production time
Female ratio adopted as a primary creative KPI
CTR and registration rate improved on winning variants

Azar & Vividi Performance Creative

What if the ad that works isn't the one you'd expect, and the only way to know is to test at scale?

Other Works

Other Works