Generate next-analytics Reports on Riff Text Interaction

[jaedoucette]

User Stories

As a scientist, John wants to know whether speaking to other people more often in RiffEDU is related to doing well in a Riff course, so that he can tell the NSF how great we are and convince the NSF to give us a swimming pool of cash to fund future research projects.

Edith is a senior executive for a company that makes razor sharp metal drinking straws. Edith has commissioned a RiffEDU course for her employees to learn how to work better together so that their frequent workplace injuries stop affecting her bottom line. Edith wants to know whether the current crop of students is engaging more with the material. She asks Riff for a report that includes the number of text-based interactions her employees had during the course. She gets a new report back the same day, and decides to purchase 5 more RiffEDU courses.

As Riff's CEO, Beth wants to tell a panel of GoldFargo Investments Inc. bankers about how Riff is revolutionizing team effectiveness, so that they'll give her a giant cheque in exchange for a tiny fraction of the company. She shows them evidence that students who spend more time using Riff's text chat feature stay engaged with the course at twice the rate of competing EDU solutions, and closes the deal.

Difficulty, impact, and usage score

This involves adding new processing logic for the Mattermost SQL databases to the next-analytics' preprocessing module. The logic will extract, for each student:

• The number of posts the student made.
• The number of posts made in the student's PLG channel.
• The number of posts made in the student's capstone channel.
• The number of reactions to posts the student made.
• The number of replies to posts the student made.
• The number of reactions the student made to others' posts.
• The number of replies the student made to others' posts.
• The 'h-index' of the students' posts, with citation counts replaced by the number of words in the post, or some other measurement of the distribution of post lengths.
• The h-index of the students' posts, where a citation is any reaction or reply to the post.
• The median Flesch–Kincaid readability index of posts the student made.

How does this tie into our current product?

Extracting these features will allow the generation of reports much like those that are already produced for video usage.

Who asked for this?

John, as part of the datapipeline design.

[adonahue]

@jaedoucette - I'm having trouble understanding how to prioritize this work. A couple of Questions:

• How do these changes to the Mattermost database fit in with our plans to make Riff pluin-able (i.e. to have as little custom code as possible )? Should this (and other pipeline work) be in it's own plugin?

• What would the reports say that isn't provided through other metrics?

• Who generates the reports and how are they delivered?

• Would this work block any portion of the test coverage effort?

• Does the data collected from this work give us the ability to do additional metrics or analysis, beyond the scope of this story?

[adonahue]

Perhaps worth doing in conjunction with story #242.