Based on testing, remove background video on Student Life.
Second level pages, such as Student Life and Academics, have heroes with background video and a CTA to watch a video inside a modal. This design is based on reasonable, yet not thoroughly tested assumptions. Can we do further research on our site?
Test variations and combinations on the Student Life page.
We tested five variations against each other.
We ran a split test where users were randomly served one of the five versions. This test was run for three months and had 4,359 experiment sessions.
Version | Video Placement | Hero | Video Plays | Video Play to End | Page Value |
---|---|---|---|---|---|
0 | Modal | Video | 19% | 4% | .02 |
1 | Text | Video | 22% | 8% | <.01 |
2 | Text | Image | 15% | 5% | .04 |
3 | Modal | Image | 16% | 5% | .04 |
4 | Text | Image | 22% | 3% | .05 |
There is not a clear pattern between video placement and video engagement or video
placement and page value. However, there is evidence that heroes with an image perform
better than heroes with a background video. Each of the three heroes with static images
performed better than the heroes with auto-playing videos. The version with the CTA,
Version 4, has the best page value. Oddly, there is a reverse correlation between
video complete and page value.
Since the previous test had 3 strong contenders with similar results, we retested those versions along with a new version. We ran a split test where users were randomly served one of four versions - all with background images. This test was run for three months and had 6,295 experiment sessions.
Version | Video Placement | CTA | Video Plays | Video Play to End | Page Value |
---|---|---|---|---|---|
0 | Text | Visit | 20% | 3% | .05 |
2 | Text | None | 22% | 5% | .05 |
3 | Modal | Watch | 12% | 4% | .05 |
4 | Text | Ask | 27% | 5% | .06 |
The results were fairly similar. Overall, the version with the Ask a Student CTA performed
the best. There is not a clear pattern between video placement and page value. Contrary
to the previous test, there is no reverse correlation between video complete and page
value. Placing the video in a modal seems to get less video plays but the equivalent amount
of video completes.
Testing the lessons from the previous test, we swapped in an image for the video on the Science page. We ran an A/B split test where users were randomly served the two versions. This test was run for 2 months and had 2,578 experiment sessions.
Version | Hero | Video Plays | Video Play to End | Page Value |
---|---|---|---|---|
Control | Video | 3.6% | 2.7% | .04 |
Alt | Image | 3.8% | 2.6% | .03 |
The results were fairly similar. Overall, the version with a video background performed
better than the image, this is contrary to the finding above with student life.
Testing the lessons from the previous tests, we swapped in an image for the video on the Arts page. We ran an A/B split test where users were randomly served the two versions. This test was run for 3 months and had 1,120 experiment sessions.
Version | Hero | Page Value |
---|---|---|
Control | Video | .04 |
Alt | Image | .03 |
Like Science and unlike Student Life, the version with a video background performed
better than the image.