snapcraft.io

a/b testing | analytics

Snapcraft.io is the home of the Snap Store - the App Store for Linux. The website hosts both the store, allowing users to browse and install applications, and the developer experience, helping the creators of snaps monitor and release new versions of their software, and publicise it in the store.

One of the main goals for Snapcraft, the product I am the lead designer for, is to attract more snap users, more people who are downloading software from the store. When I began working on the product in April this year, I decided to tackle this issue by a series of little, iterative improvements, tracked by A/B testing on the Snapcraft website. I decided that this is the best approach as it’s highly cost and effort-effective and would allow me to learn more about the audience, while not introducing breaking changes.

As the website is serving those two, quite separate, user stories the homepage is aiming to capture the attention of both audiences which ends up confusing most users and leads to a high bounce rate and low number of pages per session. This is easily uncovered by looking at the web analytics - users landing at the homepage click mostly on the Snapcraft logo or the logos of the featured companies, skipping over most of the content on the page. By watching recordings of user sessions, this was confirmed as the users were mostly split into two categories, the first one looking for the snap store, and the second one, looking for documentation - snap users and developers.

Considering the main aim of the product is to attract more people to the store, the first change I introduced was removing the hero, that currently doesn’t get much attention from users, and replacing it with the store search bar. This simple change has produced 30% higher install intent, reduced the bounce rate by 11%, and increased the average session duration - significantly improving all the key metrics in just 10 days. This test will run for a few more days, after which I will analyse the results and propose further improvements based on the metrics and recorded user sessions.