Can Uber Break 1-million Riders In Charlotte?
Based on some of the job postings around, it comes as no surprise to anyone that the everyone’s favorite taxi/technology company, Uber, has been expanding at breakneck speed. As they plan to enter the Charlotte, North Carolina market, some may wonder if this service is viable in a city that is not New York, San Francisco, or Washington, DC. I decided to run some numbers and do some fancy-pants statistics, and the preliminary figures are: its possible to get towards 1 million even in this market. I used some math to estimate.
Now, I’ve been a fan of the company for a few months and think they have done a great service to the world by starting to break up the taxi oligarchies, and allowing new drivers to enter the market. There’s even some social good that comes along in doing so.
But anyone knows that scale is the enabler of a strong business, and having enough of a market is necessary for Uber to actually turn a profit in the Queen City. The way to think about this is to look at the data out there.
Population Characteristics
Charlotte is the 32nd largest metropolitan statistical area in the country (MSA). According to the most recent Census 5-year estimates (American Community Survey, for those data junkies), this MSA has a population of 1.7 million, with 704,000 (40.4 percent) of the residents living in the urban area of the City of Charlotte.
When we’re talking about taxi riders, though, we are likely talking more about the adults, so we need to look at the population 18 and over. So looking only at adults there are 1.3 million in the MSA, with 519,000 (39.7 percent) in the City of Charlotte. Here’s a chart, because its easier to visualize:
Population is just one way of looking at a market, but when you’re thinking about taxis, its a crude way of looking at it. What matters, really, is the number of trips being had by taxi, and fortunately there’s some data out there for that as well.
Any transportation planner worth their salt knows that the National Household Travel Survey gives us oodles of data on American travel behavior. I know, there are limitations, and sometimes some flukes, but isn’t that the case with any data set? So I drilled down into the numbers and this is what I came up with. (Note: You can see my original query here, and you have to scroll down to get to the the Charlotte, MSA)
The data is kind of hard to read, but basically in its raw form its the annual number of trips, in millions, and you have to divide by 365 to get daily trips. Its a cross tab, so you have the trip purpose on the left, and the mode across the top. I’ve provided both the daily, and the annual trips for your convenience:
So putting it together with population data, we can estimate that the average person in the Charlotte MSA engages in between 3.6 and 4.8 trips a day (depending if we count the under 18 population or not)
Looking at this cross tab, you notice some pretty interesting facts (remember this chart above I made is in daily trips):
There are only about 6,500 taxi trips per day in the entire Charlotte MSA, or 2 million per year.
Taxi use in the Charlotte area is pretty much used for work trips (meaning business, at work trips, not commuting); and social/recreational trips.
The taxi market is pretty small: about 0.09 percent of all trips.
However, the taxi share is 0.35 percent for social/recreational trips; and 0.38 percent for work trips.
Opportunities in Other Modes
While the taxi mode share is small, there can be some opportunity for Uber to pick up market share from other modes. Look at the bike/walk trips: annually, there are 230 million of such trips, and 67 million of them are in the “social/recreation” purpose. While some of the walk/bike trips are going to remain, from personal experience, when Uber enters a market some of those trips become taxi trips when I need to get out of the rain or snow.
Some other speculation can be made for the transit, car, and other modes, but we’d still be just speculating as what can be picked up.
So how much can they pick off? Well, this is where statistics comes in.
An interesting method for developing potential statistics is to use the PERT distribution and Monte Carlo simulation. The PERT process is something that has been used in the military and NASA, and Monte Carlo simulation has been used, well, all over the place.
So without boring you with deep stats methods, the bottom line is this: PERT distribution lets you give a professional guess. You only have to estimate a low, most likely, and high “educated” guess at some number to create a full on distribution. Monte Carlo simulation is used to randomly pick a number from the distrubtion.
Overall, doing a bunch of simulations (I like 10,000), Monte Carlo methods let us develop a distribution for the results.
Assuming Some Market Penetration
Like I said, PERT requires you to have an educated guess, pretty much, for a low, most likely, and high value for any variable. Now, the variable I’m looking at is the potential market penetration for each mode in Charlotte.
For the taxi share, I assumed that a “high” market penetration is around 40 percent. This is because 40 percent is around what Uber targeted for a year in New York City.
That is, there are 13,000 yellow cabs, and Uber planned to launch with 105 cabs, add 100 drivers a week, and end up with about 40 percent when you do the math. Just trust me…Ok, fine. 40.8 percent. And yes, I know that 40 percent of cabs using Uber is different than the percent of customers using it. Lets assume for now.
Everything else here is an assumption. I figure on the low end, 5 percent of the taxi share in a year, and somewhere around 20 percent as a reasonable target (half the New York market penetration).
Everything else are some range of guesses. I figure one quarter of a percent for picking off modes from other shares. That can be pretty big, if you assume that one quarter of a percent of all car trips (all 1.2 billion) will ride Uber.
The chart above lets you see the range for this simulation. Again, they are assumptions, and if you have better ones let me know. Lets run the numbers anyways.
Based on the PERT inputs from above, and over 10,000 simulations, here are the results:
Uber can expect an annual ridership of 6.2 million per year. On a 95 percent confidence interval, this is 3.9 million trips on the low end, and 8.5 million trips. Here’s the distribution for your own enjoyment:
Taxi Market Penetration Only Better Results?
Now, some of you might be saying, “Hey, Mike, this seems pretty big for Charlotte.” And my gut instinct is that assuming even a 0.25 market penetration of all car trips might be a bit generous.
Well, the market penetration PERT assumptions for the taxi market at least is based on goals from New York. So, lets focus on just on that.
Here I’m going to focus in on just the existing taxi market, taking out any market penetration in bike/walk, transit, cars, and the “other” category:
Just focusing on taxis, the most likely annual ridership is 414,000 Uber riders. The old 95 percent confidence interval would be 208,000 on the low end and 767,000 on the high end.
Well, one analysis had the trips in the millions, and the other not even getting half a million.
I’d say that the taxi-only results would be a more conservative estimate of this, and probably a baseline to start from. Maybe some market penetration can happen with bike/walk and transit trips, but we don’t know what those are.
Hey Uber, you wan’t a better answer? I’d love to know some actual data to make better estimates for this simulation.
For you aspiring transportation planners out there, this is where market surveys come in handy, where you’re asking people their mode share. This would inform better PERT assumptions and give us a nicer analysis.
For now, this is a cool way of looking at it. Let me check in in a year, but looking at my taxi estimate, and adding some fudge factor, I’d say 1 million annual riders in Charlotte would be a good reach goal and something one might strive for. I’ll bet, lets say, one drink on that unless I have better ridership data. Who knows? Maybe they can peel off some single-occupancy vehicle riders and transit users, and get that number up quite a bit.