Ryde Is 30% Less Annoying

This week was a big week for data analysis on Ryde. I was getting reports from several users that simply walking around their office was triggering trips. I put in some quick bandaids initially, but it was time to peek into the problem a bit more deeply.

This analysis is based on 3111 trips recorded by Ryde, about over 1000 of which were marked by users as bike trips. 

The first thing I found was that 70% of the notifications users were receiving weren't even bike trips! I also found that 30% of trips that were marked as "Not Bike" fell into two edge cases that I could easily patch...

Too Fast

Ryde already had a speed limit of 100mph for filtering out some really crazy data. But a quick look at the data above showed that trips that were too fast were generating a ton of bad notifications. Only 0.4% of trips marked as "Bike" had average speeds of over 30 kph (as the crow flies) however, 15% of all trips marked as "Not Bike" had 30kph average speed or higher.

This was a clear win for all the Ryders. However, I don't want to ever lose a ride because it was too fast. That's something to be proud of, not lament! So I set the speed limit to 40kph and not 30kph. I will re-evaluate this cap when there is more available data.

Too Short

Ryde already has a limit for trips that are too short. I implemented it early on to filter out trips that started and ended in the same place. Now, however, I was getting complaints from users about trips that were just a walk around the office.

From my analysis 350 meters (as the crow flies) looked like the tipping point. This rule change will disqualify 1.5% of all trips marked as "Bike", but will eliminate 19% of all notifications for trips that were not done by bike. I didn't feel there was a strong downside to this decision since Ryders who have gotten credit for exceptionally short bike trips feel as though they are cheating.

Next Steps

Through my analysis it became clear that focusing on short, low speed trips (presumably walking) represents the greatest opportunity for further reductions in erroneous notifications. The next stage in solving this problem will be a more challenging undertaking, but one that I'm very excited about. Look for further reductions to "Not Bike" trips over the coming months.