Elizabeth Windram

When I started working on search, I knew there was a lot to clean up, but also wanted to be cognizant of understanding why things were the way they were. I started probing into details with the engineering team and PM team as to why things were and started to make a set of recommendations of things we absolutely needed to fix. Additionally, in tandem with the styleguide I was also doing, I cleaned up result blocks and the search UI benefitted from that work as well.

In addition to the entire left column lacking any sense of structure, it was also just really difficult to differentiate what the different types of results were. Even if a person knew to expect videos, playlists, channels, and recently uploaded videos, it was really difficult to differentiate these types of results and understand how they were ordered. A few of the key issues were:

• There were section titles for result types, but those result types weren't grouped, so the titles weren't helpful in sectioning or labeling.
• The different result types looked different enough to cause lot of visual variation, but they didn't look different enough to be easily recognizable.
• New videos had a different format because we weren't as confident that they were a great result, but rather than minimizing attention to them, we actually drew attention to them by having but only for the first of that result type.
• The newest video from a channel was displayed with the channel result, but it wasn't aligned to anything and was floating by itself.

As a team, we discussed the issues, my proposed solutions, and other people's approaches as well. We divided up the proposed changes into the things that we all agreed were much better and changes we wanted to experiment with and explore further variations. Typography changes, thumbnail borders, and the video block clean-up fell into the former category and we didn't think we would see much impact on metrics. (For these we did a 1% holdback.) For result type titling, I created several variations and we experiment with several of them. Our biggest concern was watching the pivots into these results (e.g., filtering to only "new" results, or to only "channel" results). Eventually, we settled on a much more discreet label and expected most filtering to happen from within the "filter and explore" section.

The work above dovetailed nicely with the work I was concurrently doing for the YouTube styleguide. We used 1% experiments on search to get feedback on a lot of the changes we were proposing for the styleguide and started rolling out many of the site-wide changes on search.

I was the only designer on search at this point. I worked closely with a PM and a small team of engineers, most of whom were based in Tokyo.