Why testing a more generic flow:

Testing a generic flow allows us to iterate quickly. We can spark interest from communities that haven't been involved yet by offering a non-invasive feature that simply places existing guidelines in a more accessible format. This is an organic way to encourage communities to eventually provide their own specific guidelines without requiring upfront, individual outreach from our team.

By comparing generic guidance to community-specific guidance, we will gain the data needed to decide whether next year's strategy should focus on one, the other, or a hybrid of both.

The main hypothesis here is that "generic" guidance might perform well enough to avoid the bottleneck of individual community outreach.

What does it look like?

In terms of what the more generic flow could look like and how much it would differ from the current community-based guidance:

Current guided flow for reference T414409: [Design] Article Creation Guidance - Initial intervention

• Title matching / type of article: this does not need to be modified from the currently designed experience
• Reference check: instead of community-specific lists, we use a global list of discouraged sources (e.g., social media, common news sites) derived from general consensus or existing global guidelines.
• Notability check: generic notability guidelines that are not community specific and same logic as the current experience
• Community guidance (initial advice and content) here we can check for existing community guidance that can add value if presented summarized, or include generic guidance for most common types of articles.

We can prepare a first version of these generic outlines and compare it to what we learn from the pilot communities in T417249: Collaborate with pilot wikis to define the initial guidance

How we can go about testing it if communties are not involved?

Activate by topics as we will do for the regular flow but looking at article deletion or survival rates by topic x wikis autonomously.

Questions:

I think an A/B/C test (control vs. community vs. generic) would be ideal for direct comparison, but I am concerned about traffic volume. Splitting the traffic three ways may delay statistical significance. We should evaluate the traffic levels of our pilot wikis to decide if we can run a three group test or if we should sequence these experiments instead. Also consider if we test the generic flow on a separate wiki that is not invovled in the guidance exercise.