Electrifying Ford GoBike: One month of electric pedal-assist data.

In January of 2018, the Ford GoBike system announced an eBike pilot program. The 1-year pilot was designed to integrate 250 pedal-assist electric bikes into the existing Ford GoBike station based network in San Francisco.

Ford GoBike genZe Pedal-assist bike.

The pilot included a partnership with electric bike and scooter manufacturer genZe to provide the electric-pedal assist bikes. These eBikes work by giving riders a boost while pedaling. Riders can achieve speeds of up to 18 mph (29 km/h). The bikes include an LCD with a speedometer and a field swappable 345Wh battery pack.

The eBike program officially launched to GoBike members in the San Francisco zone on April 24, 2018, at no extra cost to riders.

Today, the eBikes have been in service for over two months. Anecdotally, between social media and a few trips to San Francisco, it appears that the eBikes are very popular with riders. But of course, we were curious to see real data on theperformance. Fortunately for us, Ford GoBike publishes a trip history data set. My friend Ayser Armiti and I decided to peek into the data to visualize how the eBikes were used over the month of May.

Identifying the eBikes

The first task was to distinguish between regular bikes and eBikes in the dataset. Unfortunately, there is no attribute that identifies the eBikes, as all bikes in the system receive just a unique ‘bike_id’. After a few days of trying to track down a list of electric Bike ID’s without any luck, we had to figure out a different way of isolating the eBikes. We ended up developing a heuristic method to identify the electric pedal-assist bikes in the data.

To start, we looked at the total number of unique bikes in the fleet by tracking every ‘bike_id’ in the entire dataset. When we captured a new ‘bike_id’, we assumed that it was a new bike introduced to the fleet. Here’s a look at all the unique bikes we captured throughout the entire history of Ford GoBike:

New Bikes vs Total Fleet Size.

In total we identified 2,958 unique bikes in the system. Given that some of the bikes have been destroyed, damaged, or lost, we knew we were somewhere in the same ballpark as Ford GoBike’s public fleet stats of 2,600 bikes + 250 ebikes = ~2850 total bikes.

Next, we needed to isolate the eBikes from the regular bikes in the data. We knew that the eBikes deployed on April 24, 2018. As you can see in the figure below, there was a signifigant spike in new bikes introduced to the fleet at the end of April and beginning of May. In total, we identified 232 eBikes.

New bikes deployed over time. Note the spike of new eBikes in April/March.

Trip Speed

We wanted to check if our heuristic was reliable by comparing the median speed of identified eBikes with the rest of the fleet. To calculate this, we took trip durations from the dataset and estimates of the trip distances using the Graphhopper bike routing engine between each trips ‘start station’ and ‘end station’.

Average speed of bikes vs. eBikes

The figure above shows the speed distribution for bikes and eBikes. We can see that our heuristic recognized eBikes with a high degree of accuracy. The mean speed of an eBike (Orange) was about 9.3 mph (15 km/h) and for a pedal bike (Blue) was about 7.7 mph (12.5 km/h). If you look closely at the pedal bike distribution, you can see that some bikes have a median speed of 9.3 mph. These are likely eBikes introduced after May 2nd. Now with the eBikes identified, we can take a look at how they performed!

Trips per day

The figure below shows the total number of eBike trips per day and the number of unique eBikes used per day.

eBike total trips per day vs. number of unique eBikes used.

It’s important to note that we are unable to identify how many total eBikes are available during the day, we can only see which ones have been used at least once.

Zach Lipton on Twitter posted: “it quickly got to the point where there were around a dozen available for the whole city.”

Looking at the trend, it appears that eBikes decrease in availability as the month goes on, signaling the operational challenges eBikes face in service with charging and rebalancing.

Average rentals per day: eBike vs. Normal Bike.

During May, we can see that pedal bikes on average got up to 3 trips per day per bike, while eBikes get around 8 trips per day per bike.

Trips per day per vehicle: ebike vs. pedal bike.

Below is the total number of unique bikes rented per day with the total number of trips per day throughout the history of Ford GoBike. As you can see, the number of utilized bikes per day (orange) is much lower than the approximate total number of available bikes in the system (2600+).

Total rentals per day vs. number of unique Bikes tented per day.

Trip Durations

Do riders take eBikes on longer trips? The trip duration for eBikes appears to be a bit less. Remember, the average speed of an eBike is faster (9.3 mph) than a regular pedal bike (7.7 mph).

Distribution of Trip Durations: eBikes vs. Normal Bikes.

Trip Distances

The next figure shows the breakdown of trip distances between eBikes (Orange) and regular bikes (Blue). There is just a slight trend in longer trips with eBikes. As Kyle Grochmal points out, this could indicate that there are not enough eBikes in the fleet and they just end up replacing regular bike trips.

One point of comparison: JUMP claims to be achieving an average distance of 2.6 miles with their dockless ebikes in San Francisco. The nature of station based bikeshare where you have start and stop at a specific location, could also be a contributing factor to the shorter average trip distance.

Riders age

Here is a rider age distribution between eBike vs. regular bike trips. As you can see, the rider age distibution looks quite similar:

Distribution of trips by Riders age: eBike vs. Normal Bike.

Trip types

There’s the classic bikeshare problem with bikes that don’t seem to return to the stations at the top of hills, causing balancing headaches for everyone. Are riders more likely to take eBikes to stations at higher elevations? Here is a distribution of elevation ascension for eBikes and regular bikes. It looks like eBikes are a bit more favorable for those hill climbs :)

Trip Ascension: eBikes vs. bikes

In the GIF below, we can visualize where bike trips are ending. Green represents eBikes and blue represents regular bikes. The thickness of the roads are the percentage of trips that passed through that road segment. The red circles represent stations where trips ended and the circle size represents the percentage of trips ending at that station. Take a look at the stations on the center left of the map, near ‘Ignatius Heights’.

Green = eBikes. Blue = Bikes. Red = Stations where trips ended.

Fleet size vs. Total Trips

Lastly, here is the total fleet size in relation to total rentals. We can see a positive trend with the increase in the fleet size over time. One thing to keep in mind is that fleet is suppose to be up to 7,000 bikes by the end of 2018.

End

After one month of data, we can see that eBikes are popular in the Ford GoBike fleet. As multiple people have pointed out, riders bias towards eBikes if they are available at the station. There doesn’t seem to be enough supply of eBikes in the fleet to enable riders to take new types of trips on the system. With the eBike popularity, there are plenty of operational challenges with keeping them in service and charged. Hopefully we can see if there are improvements in availablility over the coming months in the data. But one thing is for sure, eBikes are here to stay. Now it’s up to the SFMTA to allow more to operate in the city. :)


UrbanBrain is a data science consulting and machine learning development firm specializing in mobility and transportation.