top of page
shift predictions.png

ML shift predictions

Smart shifts for optimized scheduling

problem statement

We found that users were spending 25% to 50% of their time on our app just posting shifts and another 25% to 50% canceling shifts. There were many clicks involved to post these shifts to the schedule and it was very frustrating for users. 

Users were also frequently posting more shifts than they thought they'd need to make sure they didn't end up understaffed. Often they would over-post shifts and then cancel unneeded shifts close to the shift time causing frustration for the healthcare workers who had picked up those shifts. 


We assumed that the reason most facility operators were canceling is because one of their own staff picked up the shift causing them to be double booked. We hypothesized that we could fix this problem by incentivizing them to not double book shifts and fine them when they cancelled last minute. We were shocked to discover that the #1 reason was actually change in census according to the survey we sent out. Our root problem was actually significantly different than we had originally thought, forcing us to re-think how to address it.

opportunity solution tree

Instead of focusing on negative reinforcement for late cancelations, what if we focused on helping them predict census and post the right shifts to begin with?

I put together an opportunity solution tree to explore options for helping users post the right shifts. 

personas & jobs to be done

I also created personas and JTBD statements for each persona to better help me empathize with the user. I focused on how I might address the jobs to be done of our internal users at CareRev as well.

competitve analysis

I spent time gathering examples of how other apps surface AI & ML suggestions and took notes on what I observed. 

wireframes & iteration

From there I began creating wireframes and socializing those ideas to gather feedback. I met with our client facing teams, engineers and users to dig into this idea and iterate. 


After we aligned on an approach and I worked out usability issues with testing, I was ready to hand my designs off to the engineering team and step into a supporting role. 

bottom of page