Wikimedia Discovery/Meetings/Portal roadmap 2016-03-08

Portal team roadmap planning

Deb, Jan, Julien, Moiz, Mikhail, Dan, Tomasz, Kevin

Continuation of this meeting: Portal roadmap 2016-03-04

Document under discussion: https://meta.wikimedia.org/wiki/User:DTankersley_(WMF)/Sandbox/Roadmap

Note: Deb will no longer be "presenting" the roadmap at the upcoming Product offsite, because it has shifted to unconference format. 

Goal:

Figure out sequencing of Q4 goals:

• Test a more streamlined display

• Test collapsing all language links into a selector tool 

• Create a set of portal test pages (draft name: Portal Labs pages)

• Complete a production release

Discussion

Moiz: Portal Labs does not represent user functionality, so building it is not suitable for a Q goal

Jan: Like the idea of going down 2 roads simultaneously (small+incremental vs. revolutionary)

Moiz: Incremental changes are available almost immediately. That's the work to earn the revolutionary work, which meets our needs more than users. Focusing an entire quarter on non-user facing seems problematic. 

Deb: One main thing we want to do is have the page localized. That would be a huge user-facing goal. 

T: What would be the outcome of that output?

D: User satisfaction for non-English users.

T: Do we have a way to measure satisfaction

Jan: We have clickthrough rates. How it relates to satisfaction is a bit blurry.

Moiz: We are measuring languages of users coming into the page.

Deb: The upcoming test will help us understand this better. We would have to make assumptions now. 

Jan: First test (making search easier to find) was successful, so people want to search. 

Moiz: Showing images in the dropdown probably helped. Can we do something more with that finding, without running another test? I would like to do more with that test. 

T: Would be nice to be grounded in data, but we'll have to take a guess. 

Deb: Images were important, bigger button. Looked more modern, so people used it more. 

Jan: About 40% of the people who saw suggestions clicked on one instead of some other link. Proves that images can help. 

Deb: If we can localize the page, so I come in using a different language, and we can make sure the content is localized, that would be good. Stretch goal would be to fall back to English (or something) if that language is not available. 

T: "As a user, I would like localized content" (Scribe missed some here)

Deb: Can we actually show search results from other languages?

Jan: That's a multi-lingual (multi-project) search, which is a ways away. Some names have redirects, so in some cases something might show up. 

M: Before we localize this page, which is a big multi-step task, we should run tests like the universal language selector. Right now we know that 35% of people who land here run a search, 50% do nothing, 1% click on languages below. With that, I don't want to build a pipeline that would only address 1% of the need. Right now might not be the time. 

D: Before we do anything like that, we need to do our homework. We need to know what we can do. The test will help. Looking back to the doc: Are these the things we want to do? Some are pie-in-the-sky. Do we want to take those on first, or later. 

M: We mostly agree that we want to do everything there; just a question of when. Another thing we should explore is trending articles. That builds on one successful test (showing images). Also the iOS app going live tomorrow has "most read" (similar to trending articles), so they will have real data in 2 weeks. They have also worked out API kinks. We can learn from that. Doing similar on desktop would learn from them, and would build on our earlier test, and is a more visual thing (which would be fun for the team). 

T: Higher risk, higher reward

D: Localization: People might not even notice the first time or two. But it will be "cool. It's there." Trending and visual changes will be more obvious.

T: For trending, be sure to check with ops re: cacheing. 

M: Yup. 

D: Would recommended articles be localized? If we don't have anything in their language, what would we do? Maybe mobile data would help clear that up. 

M: iOS team has figured a lot of that out. They work in all languages (some better than others). Pick a language, and top articles will appear. 

D: If a smaller language isn't updated as much, and I visit every day, would I see the same articles all the time? 

M: First iteration wouldn't be perfect. 

T: Fewer unknowns with this one. We have precedence to follow. With localization, we don't have that. 

(scribe interrupted so a couple minutes of notes were lost here)

Julien: I agree about trending articles. 

Jan: Trending is not based just on pageviews, right?

M: Highly recommend joining forces with iOS beta. They have done some nice work to pick a good set of articles. 

T: There is already precedent that signpost has been using to select top articles. We don't want to filter, so we would not be doing anything that other hasn't already done. 

M: If they see that a page is getting traffic from just one platform, that raises a flag, so they can correct for it. It's community-driven. 

Julien: xhamster is in the top 3 articles on my beta iOS app. It seems to be the same normal API call. 

(Some discussion about who is seeing what; updating the app seemed to fix it)

T: What open questions to we have? Trending seems more grounded. Localization has more open questions. 

Mikhail: Trending articles are for enwiki. If we pursue this, people with no interest in English will get them. 

M: You can switch the preferred language on the app to get that language. Still some bugs, but it works. 

D: Comes back to: What should we do first? Should we get localization done first so we can do it effectively? Or just move forward with trending articles?

Jan: The articles are already localized. Just need to serve the right articles to the right people.

D: Will we have enough knowledge to do trending articles well? 

M: We won't be localized as far as the static text (e.g. "click on a trending article" won't be translated)

D: Do we want to localize first? Or trending first?

K: Radical overhaul to page would require community discussion. Could take time. Small wins could help team morale and build credibility of the team. 

Jan: Yes, trending would take a radical overhaul

M: Can we vote Yes or no on "Trending articles first, then localize"

Unanimous (not counting T/K abstaining)

Deb: I'll rearrange the Q4 goals, and switch it to one main goal. 

T: Eng can enumerate risks, including cache effects. 

K: And Chris needs to be in the loop on the community side

D: Include these issues in the roadmap?

M: I would vote no. 

T: Let engineers decide where to have the discussion. Goal is to maximize alignment and answer questions before the next quarter starts

M: Check with Ori etc. (regarding performance)

T: Eng can pick who to contact

D: Track in phab or elsewhere?

Julien: Phab is fine (Jan +1, keeps it public)

M: Really happy because this should take us to the point where the 60% of people who do nothing can find content without searching. Avoids the need to do search. 

T: Be sure to check with Josh about how the filtering works. 

M: This is great. I have been trying to get trending articles in for a long time.