Lyft priority pickup waiting times
When I moved to California at the beginning of the year I didn’t have a drivers’ license. To get to the office in the morning and back home in the afternoon I used Lyft rides.
Lyft rides have an add-on option called ‘Priority Pickup’, where you can spend a couple of dollars to decrease the amount of time you wait before your ride is available.
The ride ordering screen shows estimated waiting times for the different services ‘Priority Pickup’ and ‘Standard’. Once you choose a service you are matched with a driver, and you get a different estimate of when that particular driver will arrive to pick you up.
After having used Lyft every day for a couple of weeks, it seemed that the estimate for the ‘Priority Pickup’ waiting time was never accurate compared to how long I had to actually wait. It seemed that I could have just picked the ‘Standard’ pickup option.
To test this, I started gathering data. Every time I ordered a Lyft ride I saved a screenshot of the screen showing the estimated waiting times for the two different services, picked ‘Priority Pickup’, and then saved a second screenshot showing the time to arrival for the driver I had matched with.
Over the course of about a month, I collected screenshot pairs from 43 lyft rides, and then copied the different estimated times into a spreadsheet.
For every post-match estimated waiting time, there are two pre-booking estimated waiting times: one for ‘Priority Pickup’ and one for ‘Standard’. Naturally, pre-booking ‘Priority Pickup’ waiting times are always shorter than ‘Standard’.
Once you are matched with a driver after selecting ‘Priority Pickup’, is that estimated waiting time typically closer to the promised ‘Priority Pickup’ estimate or the ‘Standard’ estimate?
We can look at the distribution of the differences between the ‘Priority Pickup’ pre-booking estimate and the post-match estimate, and at the distribution of the differences between the ‘Standard’ pre-booking estimate and the post-match estimate. Then we can compare these distributions.
Visually we can see that the ‘Standard’ difference from the post-match estimate is shifted to the right of 0. So the post-match estimated waiting times are typically better than the pre-booking ‘Standard’ estimated waiting times.
Through linear regression we can learn that, on average, the post-match estimated waiting times after selecting ‘Priority Pickup’ are two and a half minutes shorter than the pre-booking estimated waiting time for ‘Standard’.
The post-match waiting times are, on average, half a minute longer than the pre-booking estimates, but there is uncertainty in the estimation of this average, where the possibility of pre-booking estimates being spot-on is contained in the 95% confidence interval.
It turns out my impression of the ‘Priority Pickup’ option not improving waiting times was wrong. There are a few cases where the post-match estimates are much longer than the pre-booking waiting times estimated for ‘Priority Pickup’, visible as the red dots in the upper left corner of the scatter plot above. My original impression is a very common cognitive bias, where the cases when I had to wait for the ride for longer than I had paid for are more memorable due to the associated disappointment. Without collecting this data I would have continued to believe that selecting the ‘Priority Pickup’ option wouldn’t make a difference.
A notebook with the code for the analysis is available at GitHub
