Best-in-Class App Search in China
In China, the Google Play Store is not mainstream. All large phone manufacturers have their own App Stores that contribute significant revenue in addition to core mobile phone sales.
In 2015, Quixey started working with Meizu, one of the leading mobile phone manufacturers in China, to power search across the Meizu App Store on all their phones.
Meizu previously used another leading app search service; the goal was to significantly improve app downloads to generate additional revenue.
Integration took about a month, including A/B testing to compare two main performance metrics:
- Search Results Page CTR - Higher CTR means more relevant search results
- App Details Page CTR - Higher CTR means the app meets the user’s expectation
Quixey showed significant improvement over the leading vendor (red line), as shown below for both key metrics (all percentages are relative). The data below highlights the months following the launch.
During the first week, Quixey’s metrics were only marginally better, but over time, Quixey’s results improved further. In 4 months, Quixey CTR was over 20 percent better for search results, and over 35 percent higher on app details.
These improvements would mean an app store with a daily search volume of 10M would receive around 2M additional downloads every day.The Best App Search in China
There were several differentiators that contributed to building the market-leading app search product in China.Data from Multiple App Stores
App stores usually use only their own data, but Quixey gathers data on a daily basis from all major app stores in China. The impact of this:
- Search results are more comprehensive. Previously, the Meizu App Store did not include some of the apps available to the market.
- Search results are more relevant. We used and combined download counts, ratings and other signals from multiple sources, improving our ranking.
Quixey creates machine learned ranking models. As more signals were integrated into the models, we optimized our search results ranking. This contributed to the bulk of the early improvements.Data Feedback Loop
Our Data Feedback Loop uses user engagement data, including searches and downloads, to improve the overall relevance of search results. This is extremely effective because of the recency effect; user engagement data are collected in near-real-time and fed back into the ranking model. The outcome is that search results are more relevant to user intent, and trending apps can be promoted quickly.
The figure below depicts Quixey’s Ranking Model Data Flow, including our Data Feedback Loop.
Ranking Model Data Flow
Combining a variety of techniques led to Quixey building out the market-leading App Search offering in China. We are now working closely with the Meizu team on:
- Further monetization of the app search experience via promotion of relevant apps
- Enhancing the user experience with Quixey’s Deep Search technology.
Stay tuned for more exciting updates from Quixey in China.