English Product Growth
Free-reading usage in the English-language service trailed the Korean service, and adding content did not produce a stable lift in weekly retention. I treated this as a growth problem across acquisition, activation, discovery, and return visits—not a translation problem. I worked with five functions to run hypothesis-led changes, and shifted the strategy toward personalization when a result did not hold across cohorts.
- My role
- English Product & Iteration Lead
- Decisions I owned
- Growth-portfolio model · pivot to personalization
- Team involved
- Content · social · growth · design · engineering
- Observed outcome
- Android install conversion rate 16.3→33.4%
- Leadership period
- 2025.03—10
- ASO comparison
- H1 2025
- Organization
- 5 functional teams
- Capabilities
- Growth Strategy · Activation · Personalization
The situation: Free-reading usage was lower in the English-language service
When I placed the two services side by side, free-reading usage was 44.75% in the English-language service versus 67.82% in the Korean service. Retention was 8.89% among the one-use group and 30.86% among the ten-use group.
I treated this as a signal that repeated value discovery and retention appeared together. It was not evidence that more uses caused retention.
How I ran the growth work
I managed acquisition, activation, content discovery, and return visits as one journey, aligning shared metrics and milestones across content, social, growth, service design, and engineering.
Before each release, the team and I agreed on the hypothesis and target metric. After launch, weekly dashboards and cohorts told us whether to continue, revise, or stop.
Running store pages matched to each ad context
We launched 39 ASO updates matched to each ad and content context: 29 on Android and 10 on iOS.
- ASO updates launched · Android 29 / iOS 10
- 39
- Android install conversion rate · from store-detail views
- 16.3% → 33.4%
- iOS install conversion rate · from store-detail views
- 2.6% → 4.8%
- Custom-page average conversion rate
- 28.9%
Redesigning activation and the recommendation structure
I chose to make birth-data entry the primary action, then worked with design and engineering to reduce the large opening visual. The intent was to let people start the core experience before reading an explanation of the service.
We reorganized the home experience from seven content sections to 16 and consolidated 79 inconsistent tags into 22. We then used history and tags to drive recommendation rules.
- New-user drop-off
- 44% → 38.4%
- Question-page recommendation interaction rate
- 30.4% → 41.5%
- Result-page recommendation interaction rate
- 45.8% → 50.3%
- Result-page recommendation events per user (EPU)
- 4.5 → 6.3
The hypothesis that failed, and the next pivot
Increasing the supply of free content did not create a stable lift in weekly-cohort retention. Results changed with the content released, so volume alone could not explain them.
I shifted the next strategy from adding more supply to improving data-informed personalization.
A good average in one period did not mean the hypothesis had held up.
What the data can and cannot show
Both figures show the share of store-detail-page viewers who went on to install the app. Across the periods compared within H1 2025, Android install conversion was observed at 16.3% in the earlier period and 33.4% in the later period; iOS was observed at 2.6% and 4.8%, respectively. The 39 figure is the number of ASO updates implemented, not the number of successful experiments.
I reviewed retention through weekly cohorts and their variation, not one overall average. A result that did not repeat became a reason to pivot, not a win to report.