Increasing Page Views
Quantitative A/B Testing
Minor coding of HTML/CSS
Most visitors to the site would land on a Healthline article from Google, read it, and then leave. Being on the engagement team, one of my goals was to increase time spent on the site by getting the user to read more recommended articles. The content team wanted users to discover and read many other related articles that would be of benefit to their health knowledge.
Many, many variations were explored through the process of A/B testing and observing metrics like time spent on the site. My role would be to create designs, and implement them in Optimizely myself when possible with minimal coding. For drastic changes that required a lot of front-end coding, I worked with internal developers, or external developers I would hire from sites like Upwork.
Below is an example of one variation, though many more variations were done. I can present more details in-person.
Hypothesis #1: At the end of an article, rather than presenting multiple choices for other articles to read, present a single option with high prominence.
Original variation: 5 recommended articles at the end of an article
Variation 1: Recommend a single article, with high visual prominence
After 14k sessions run in Optimizely, the results were flat with respect to increasing visit duration. The conclusion was that having a single, prominent option in this style did not increase visit duration on Healthline.com.
A DISRUPTIVE ITERATION
While I was testing small variations of the above design, my manager and I simultaneously worked with a freelance designer to come up with a disruptive variation. A disruptive variation in this case was a full re-design, where I re-imagined what the reading experience could be. In addition to design input, I worked with the development team to manage implementation and integrate Google Analytics. I then managed the testing in Optimizely.
The disruptive variation
We saw a huge spike in time spent! I thought we had a huge win. However, through more iterations of testing, I deduced that since this was a lightweight prototype, it was performance (reduced page load time) that resulted in the time spent on the site, not the design itself. =/
After running several tests throughout the year, I concluded that in order to increase visit duration, the two areas of focus must be:
1. Improving the algorithm for the recommendation engine itself. Recommending articles that are more relevant will lead to more clicks (yes, I A/B tested this).
2. Performance matters. A LOT. Users will click around more when the site loads faster.