For a surprisingly large percentage of commutes in San Francisco, Skip is faster than driving. Especially for commutes where bike lane infrastructure exists. In this study, we take a closer look at the science behind that claim — how San Francisco Skips traffic!
Every year, the average San Francisco commuter spends over 100 hours sitting in traffic, at an estimated cost of $1,624 per driver. Rush hour driving speed has fallen every year and now averages 12 miles per hour.
After 10 months of operating in the SFMTA Powered Scooter Share Pilot, we have data that shows what we’ve believed all along, intuitively: new transit technology like Skip can be the best option for everyday commuting.
Focusing on San Francisco Skip rides that pass through or finish at the Caltrain or Montgomery Street BART station, we examine two of our most common destinations that also rank among the Bay Area’s most popular transit hubs.
TL;DR: For more than 70% of trips examined, Skip is far faster than a car!
For example, if you are heading to Caltrain from The Mission, Google estimates an evening drive time of about 13.5 minutes, while Skip typically takes just 8. From the Ferry Building or Embarcadero yields a similar savings: 12.5 minutes driving, but just 8 minutes by Skip.
The Skip advantage can be even greater when commuting to Montgomery St BART from The Mission or Mid Market. You’ll typically save 5+ minutes on Skip versus driving.
Let’s look closer at the science behind our study.
To highlight the impact at rush hour, we examined all historical anonymized Skip trips taken to Caltrain or Montgomery St BART. We drilled down further to weekday evening trips in the second quarter of 2019, leaving roughly 10,000 trips (4,300 Caltrain and 5,700 BART).
From that rush hour data, we converted each Skip ride into a time-stamped sequence of geo-coordinates along the route (sample below).
By spatially organizing trip data, regional travel times via Skip vs. car can be compared from different parts of the city. At Skip we use the open-source H3 library to model spatial data. H3 overlays a map with a grid of uniformly sized hexagons.
Hexagonal tiling has many advantages, most important of which is a fixed grid that guarantees a given hexagon ID always corresponds to the same region of space. This allows confidence in comparing results across experiments and across time.
Unlike neighborhood maps and zip codes, hex grids require no prior knowledge of semantics or physical geography, are less prone to biases, and removes concern about normalizing results, because each tile covers the same surface area.
Finally, hexagons reduce discrepancy among distances from the center point to any given edge point (vs. a rectangle, for example).
Applying hex tiling to the sample trip above:
- Step 1: Discard points after the destination of interest; assume that a trip that passes through Caltrain or BART also ends there.
- Step 2: Overlay H3 hexagonal tiling.
- Step 3: Identify H3 tile that contains each geo-coordinate point
- Step 4: Retain only the latest time stamped point in each tile, measured in minutes against the first point at the destination.
This fourth step is important to call out because of diverse rider patterns. If a casual trip criss-crosses the same hexagon multiple times, we only want the last transit to count in calculations (we also filter out trips longer than two hours to reduce casual rider influence).
From trip to dataset
Transforming each rush hour trip from a sequence of points to a set of tiles completes the full dataset and allows travel times to be computed.
Each tile crossed through on a trip (barring destination) is a separate observation, with this “unbundling” resembling the Uber Movement project methodology, though solely focused on two destinations vs. the entire city.
The resulting dataset has three columns: Trip ID, Hexagon ID, and Travel Time:
The median travel time (in minutes) across all trips across a given tile looks like this:
Finally, visualizing median travel times for all San Francisco tiles with at least 10 observations results in a clear picture of typical travel weekday evening travel times to Caltrain:
Taking Skip from Union Square or the Ferry Building is about 6–8 minutes, while it takes about 20 minutes to get to Caltrain from Fisherman’s Wharf.
Key finding: bike infrastructure can offset longer distances; going the two-thirds of a mile from Townsend and 8th to Caltrain only takes ~ 2.5 minutes on Skip, thanks to a designated bike lane.
To calculate time saved on Skip vs. driving, we use the commercial Google Maps Distance Matrix API, comparing trips starting at 5:45 p.m. on a Thursday evening from the center of each tile and ending at Caltrain or Montgomery St. BART. The estimated driving time is subtracted from the Skip travel time to show time saved (tiles with fewer than 10 trips aren’t shown).
Important note: we don’t include any time for ride-hailing or for parking (or exiting a parking structure); just the straight drive time from point A to point B.
Blue means Skip is faster, red means driving is faster, and white is a tie; darker shades indicate more extreme differences.
For more than 70% of the map, Skip is in fact the clear winner!
Heading to Caltrain from The Mission, for example? Google estimates an evening commute time of about 13.5 minutes, while Skip typically takes just 8 minutes. From the Ferry Building or Embarcadero yields a similar savings: 12.5 minutes driving, but just 8 minutes by Skip.
The Skip advantage can be even greater when commuting to Montgomery St BART from The Mission or Mid Market, you’ll typically save 5+ minutes on Skip versus driving.
There are commutes that are faster by car, of course. Results from the northern part of the city, especially Fisherman’s Wharf, show longer Skip travel times. We have some early hypotheses for why.
- First, these areas are more prone to random noise in the data
- Second, we believe tourists tend not to be as concerned with time to destination
- Finally, the lack of bike lane infrastructure contributes to less time savings.
This type of real world study is fundamental to Skip’s mission to energize cities by making mobility accessible to everyone.
The lessons from this study will also drive innovation in our customer experience, guide our fleet positioning, and inform our community and transit agency partnerships. We believe the future of cities will be shaped by responsible data sharing.
At Skip, we’re working on a variety of hard problems at the intersections of data science, machine learning and connected hardware. If solving real world problems at scale interests you, our team is hiring!