Wikimedia Discovery/Meetings/Portal planning 2016-03-17
Portal Planning 2016-03-17
Deb, Moiz, Jan, Mikhail, Dan, Kevin
Topic: Q4 (and beyond)
Goal is to make the page easier to use. Entice people to use existing features more. Bring in new feature ("trending articles")
Want Mikhail's confirmation that the planned A/B tests will provide useful/actionable data/results.
Check with Jan on technical facilitation
Each test is expected to run for a week. The full cycle per test (coding, refining, analyzing) might take a month.
[Moiz shows mocks of A/B tests]
- Collapse all bottom links (other languages and other projects) into fancy dropdowns. Dropdown will group languages by article count, similar to what is already on the page.
- Starting from the existing point, move globe links to a line below the search box.
- Combine the first and second test. (Hopefully push to production here, depending on test results, and community feedback outside the A/B tests)
- Final target endpoint, which would be like #3 but with trending articles added. (Hopefully push to production here, depending on test results, and community feedback outside the A/B tests)
Moiz will post "final" mocks to wiki and solicit feedback. Post to [Discovery] mailing list, along with portal improvement wiki pages.
- We would remove the article counts from the top-10 links, and hiding other article counts inside a dropdown. So we need to make a strong case to justify this.
- Sister project links don't really convey what they do just from their name, so maybe we could improve by including descriptions in that dropdown.
Questions for the team:
How can we improve those experiences?
Mikhail: How about an A/B test keeping sister project links on the page, with descriptions?
Moiz: I wrote a draft 3-5 word description of each project.
Deb: I like the suggestion; We want to be able to have the language picker, but people are interested in it. I like the dropdown treatment, but how would it work on mobile? Maybe there could be a button to show all the links the way they are used to.
Moiz: I need to mock up mobile versions of all these pages. That's on my TODO list.
Jan: I wasn't expecting to see the sister projects in a dropdown. I kind of like them out on the page. There are lots of considerations--is the language selector the same as the ranking list, or separate? The mobile and non-JS cases need to be considered. Features like trending articles wouldn't be available without JS but we want to make them available in some way. I like the clean look of the proposed final page. I assume we'll still be doing the localizing thing where we put the user's preferred language(s) and not generic top-10. I like de-emphasizing the ranking because it doesn't help readers, but obviously editors/admins like the rankings.
Moiz: I also prefer the sister projects in the footer, but Deb wanted to see them above the fold. We can discuss further.
Jan: We should talk more about it. Would that conversation happen in phab? There is one ticket right now; do you want to split that into more tasks?
Deb: We can run an A/B test to see if the sister projects are better in a dropdown or footer. The logos or names in the footer didn't seem as helpful, but with the descriptions maybe it would work.
Moiz: If every test takes us a month, each one pushes everything back, which is painful. Earlier mocks had it in the footer. It is a common pattern; like a site map.
Deb: The question I have about that is: How do people know there is something below the fold. Users might not scroll down.
Moiz: That's a design and product decision. We need to prioritize the elements on this page. We can't put everything above the fold.
Deb: But if there is no indication that there is something below the fold, they might not even try.
Moiz: Research shows that scrolling is a common behavior; many people scroll first as soon as a page is loaded. Especially on mobile. But we can put some indication that there is something below.
Jan: I agree about prioritizing things. We could have a visual indicator that there is more below.
Deb: Relevant article: http://conversionxl.com/above-the-fold/
Moiz: Our "call to action" is to search or to click on a (relevant or) top-10 link. Maybe trending would become the first, with search second.
Kevin: We don't really know why people are coming to this page, so it seems presumptuous of us to say the call to action has to be search. But this is wikipedia.org so the sisters seem like they would be secondary.
Deb: I'm not happy with the term "de-emphasize". We want people to find it. Below the fold is fine, but I object to the term "de-emphasize".
Moiz: We need to have a hierarchy on the page, which means you emphasize some things and de-emphasize others.
Deb: We don't want trending to overpower the search.
Jan: With the language stuff, so far the A/B tests here are pretty easy. The significant work would be the universal language selector. And obviously the trending articles.
Deb: So if we don't use the universal language selector, but instead use the new mock from Moiz, I would think that would be easier to develop?
Deb: As a PM I love the universal language selector, but I understand it's difficult.
Jan: I would love to rework the selector, but we should see how our time goes.
Deb: If we use the universal language selector, we might need some alternative mechanism to show the article counts and ranking.
Moiz: Was talking to Pau, asking why the language selector wasn't more widely used. (Scribe: Not sure what the reasons were.) They have a JS-only version which is outside of MediaWiki. But I agree with Deb that doing the easy way for now makes sense. In the future, we could change that dropdown to the language selector.
Kevin: Leaving sisters below the fold could save effort vs. putting them in a dropdown.
Deb: Right. For that test we could leave them below the fold, but have the 3-word descriptions. Would be interesting to see if they got more clicks due to the description.
Jan: We could put a just-above-the-fold hint that there is more down below.
Deb: The fold is at a different place on each device.
Dan: My advice: Get Chris looped in. It's fine to have these discussions with subsets. Get the plan on a wiki page (which is the team's responsibility). Chris's job will be to figure out how to disseminate that info, and get feedback to where it belongs. This looks good. The presentation looks good.
Deb: I talked with Chris yesterday, and he's OK with it.
Dan: Just include tons of disclaimers (plan not commitment; mocks are an indication not detailed pixel-perfect; etc).