INTEREST PREVIEW
& DISCOVER

Duration

July - Oct 2018

team

1 PM, 2 Front-end, 2 Back-end, 1 DS, and 1 Product Designer

Role

A/B Testing, Prototyping, Data Analyzing

Output

Mobile App (iOS, Andr), Web, Mobile-Web

Service Overview

Vingle is an international online community that empowers more than one million monthly active users around the world to share their interests with each other. The platform connects and recommends users with universal interests including over 3,000 topics in music, film, sport, and social issues. The service is available in 26 languages where users come to create their own universe.

Project Statement

One of the core issues of the platform was the limited engagement within an interest community and the restrictions around transitioning from one community to another. As a result, Vingle's users could not fully utilize the core values of the service, which reduced their enjoyment of the platform. Hence, the team began strategizing to connect the discontinued UX flows.

DISCOVER

Users lack opportunities to discover new interest communities due to the popularity bias of the platform.

Vingle already had features such as "Related Interest" and "Search" to allow users to have a more seamless experience. However, these tools were underutilized by the user base. This complicated the process for users to transition between various interest communities.
‍
In order to consume content on the platform, users predominantly scrolled through the newsfeed of their respective interests. Vingle's algorithm rewarded interest communities with higher levels of engagement by placing them higher on the newsfeed. Therefore, interests with less participation were relegated to the bottom of the page.

Consequently, the platform's users were limited regarding the amount of interest communities they could participate in. This reduced the exposure to communities which were less popular.

Figure 1) Tag count histogram & Tag-based co-occurrences between interests.

SOLUTION 01

The home page should not be the source of content consumption, but rather the feed which empowers users to explore and join new interest communities.

We found that 85% of the users stayed on the home page. They recognized this feed as the main source for content consumption, which hindered the transition into various interest communities. Therefore, we listed the interests on the top of the home page, which ultimately encouraged users to explore new interests and transition into them.

With two client developers, a back-end developer, and a data scientist, I reached a consensus on how to tackle the various issues, which included the necessary base work and methods to maintain the advertisement section on the page. Then we distributed the updates to 20% of the user base for a/b testing and continued to optimize our design solutions.

REsult 01

Group3 positioned in the top of the homefeed with a random sampling of users' followed interests showed the highest Click-Through Ratios. On the other hand, Group1, Group2, and Group4 were positioned in 5-6th place with various algorithms that aimed to increase curiosity of users.
‍
Interestingly, users in 20s and 30s recorded significantly low average CTRs of Interest Preview serving QnA and Talk. The results led us to un-weighted the contents for the particular age groups.

Figure 2) Daily Interest Preview CTR (i.e. #. of users clicked interest previews * 100 / #. of users viewed interest previews); group1: mixed content types, group2: recommended interests, group3: uppermost positioning, and group4: popular interests.

Figure 3) Proportions of clickers' age group.

Figure 4) Preview Content CTR (i.e. #. of users clicked preview contents * 100 / #. of users viewed preview contents).

SOLUTION 02

Expose users to interests they are not currently following but may potentially have a regular visit

Recommend interest for users that they may like. Create more exposure to users by implementing elements like ranking as it should be a page that users should be able to search or browse new contents or interests.

REsult 02

Apparently, each gender groups showed different behaviors given similar sets of cards. While males recorded  a higher average read count, females were more active in the sense that they read more diverse sets.
‍
Two histograms represent users who decided to vertically scroll down in the screen and the corresponding Click-Through Ratio on contents at each position. Counts position 1 was naturally high and the CTR was spiked up in position 3.

Figure 5) Discover usages per demographics; Female Reads (top left), Male Reads (top right), Female Users (bottom left), and Male Users (bottom right).

Figure 6) Content CTR per position.

Figure 7) Users’ maximum scrolled positions.

Next step

An Interest chart and Interest Map were also proposed as a prototype afterward. It began based on the idea that the service can turn usersmore active by providing a map that allows users to have complete visibility on all data of over 20,000 interests.

By showing the real-time data, it allows some users toput more effort to develop the popularity and the performance of their main interest. It also allows usersto find less popular interest or contents to develop into something more popular. As a whole, it will generate the overall development of the community for interests with both high and low traffic.