← All work Case study · Agency
Seat 1 · Agency · Merkle Sokrati (dentsu), for VIU

Ten markets, one learning loop

Analyst, Merkle Sokrati · account shared with another analyst, split by market20183 min read
~70Kapp installs a month
10.5%install rate
10markets across Southeast Asia, the Middle East and North Africa
02 · The situation

Install costs were too high, and every market was learning alone

ClientVIU, a leading OTT streaming platform owned by PCCW
My roleAnalyst; split the markets with another analyst
MarketsIndonesia, Malaysia, Myanmar, the UAE, Saudi Arabia, Qatar, Kuwait, Oman, Bahrain, Egypt
ChannelMeta app installs

The markets differed in language, device mix and content taste, and some had no account history at all. Cost per install was too high to hit subscriber targets. The same generic approach ran in every market, so a lesson learned in one country never reached the others, and new markets started blind.

The problem wasn’t any single campaign. Learning was slow and local. With ten markets and limited history, the account needed a way to find what worked quickly and spread it.

03 · The call

Ten markets, some with no history. How often should the account learn?

Make your call to see mine.

04 · What I built

A weekly loop that spread what worked

Step 1 · Before

Ten markets, one generic approach

The same setup ran everywhere. A lesson learned in one country never reached the others, and new markets started blind.

Step 2 · The loop

Analyze, change, brief, every week

Each week: performance by format, placement and operating system, a change in the account, and a recommendation for the client.

Step 3 · Findings

What the loop found

Video took 78% of spend and drove 2.88x more installs than image. On iOS, Instagram ran at a 38% lower cost per install than Facebook; on Android, 32% lower.

Step 4 · Rollout

Every finding reached every market

Each finding was rolled out across markets, so new launches started from evidence instead of from zero.

Exhibit · step 1 of 4
Ten markets learning separately, a weekly analyze-change-brief loop, findings on video and Instagram cost per install, and each finding rolled out to every market ID MY MM AE SA QA KW OM BH EG 10 markets, each learning alone each finding rolled out to all 10 markets Analyze Change Brief every week installs: image vs video image video 2.88x cost per install on iOS Facebook Instagram −38% cost per install on Android Instagram −32%
Now showingTen markets, one generic approach
05 · What happened

One loop ran ten markets at scale

~70Kapp installs a month across 10 markets
10.5%install rate
280Kvideo views a month, a secondary signal
  • Video creative drove 2.88x more installs than image creative.
  • Instagram ran at a 38% lower cost per install than Facebook on iOS, and 32% lower on Android.
  • Single-image ads outperformed carousel on engagement.
  • New market launches started from tested findings instead of from zero.
06 · What I'd change

I’d optimize for subscribers, not installs

Installs were the agreed metric, but a cheap install that never subscribes is not cheap. Optimizing further down the funnel would have aligned the campaigns with the business model.

Across many markets, the speed of learning matters more than any single result.

Building new capability · Streaming / OTT · SE Asia, Middle East, North Africa · Meta app installs

Hiring for a role like this?astitvasawhney@gmail.comMore ways to reach me