7,777 votes: How we generate the Minnebar schedule

Update: Here is the interactive scheduling visualization from this session. Enjoy! (Source code at the bottom of that page.)

Did you know we Minnebar organizers create the session schedule based on your votes? It’s true! An algorithm turns thousands of session votes into one giant schedule. When you click that “Yes! I might attend” button, you affect which sessions happen when.

This poses an interesting problem: How do we decide that one schedule is better than another? How do we quantify that?

If you are only interested in 2 sessions, then you want them scheduled at two different times. Easy! But what if you are interested in 30? There are only 7 time slots, so you can only see 7 of those sessions — and you almost certainly will get to. Are some of the schedules where you get to see 7 sessions better for you than others?

In earlier versions of the session scheduler, the answer to that question was “no.” In recent years, we upgraded the answer to “yes!” Come find out how we did it.

In addition to being a fun little brain teaser, this talk has lessons for organization trying to make data-driven decisions that sit at the intersection of data, UX design, and human behavior.

This is a follow-up to my earlier session on the topic. The session before focused on simulated annealing, and the advantages of using a zero-training probabilistic optimizer (which is a fancy way of saying “just let the software guess a lot”). This session delves into the details of what we are optimizing.

This talk does not require advanced math or programming background. There will be some fun tidbits for the math lovers in the audience, but the concepts will be accessible to all.

• Development
• Other