Applying Data to Solve the Mystery of Multi-Space Parking Meters
The Art of Parking Analytics Redux
By Matt Darst and Eduardo Cardenas Sanchez
Like the plot of a film noir whodunit, the world of transportation data is murky. Separating what we suspect about parking and mobility from what we can prove is difficult. There are precious few clues, and those that do exist are buried in a cloud of data thicker than cigarette smoke in a 1940s speakeasy. The facts are convoluted, and the bigger the data, the bigger the mystery.
The stakes are high. Figuring out parking is essential to move people and reduce congestion by shifting demand and getting drivers parked quickly. Inefficiency siphons off money like an extortionist, leaving less to invest in vital infrastructure. But it’s not all about the “cabbage” as they say. Where there’s congestion there’s also driver confusion and distraction. That makes modal alternatives like bicycling or walking less attractive. Unless cities fix parking, innocent people will continue to get hurt. Some might even die.[1]
Sure, reports and dashboards might augur trouble, but they don’t do much to help you get to the bottom of it. They don’t communicate patterns or their importance. That means the detective work rests on you, the parking and mobility manager. Unfortunately, parking professionals are already tasked to the hilt managing employees, money, supply, and customers. There’s very little time to dig into the data.
Sometimes you need help. That’s when you call an expert, someone hard-boiled and skilled at navigating the dark labyrinth of data. In the movies, that’s usually a private investigator. In the real world, it’s a data scientist.
Instead of carrying a revolver and a badge, data scientists are armed with statistical and economic theory. Using data classification, data clustering, machine learning, and predictive modeling, data scientists can uncover hidden patterns. These clues are critical. They help identify trends, shine a light on the impact of technology, and track the effectiveness of business processes. And, unlike a report or dashboard view, data scientists will clearly communicate what they’ve found to their clients, whether good news or bad.
The difference between reports and advanced analytics is as black and white as an old Bogie film.
The Shadow of the Parking Meter
Advanced analytics can improve parking and mobility, including the customer experience, operations, and sustainability. When the trail’s gone cold, data can be used to gain insights and guide decisions about setting the right hourly rates, correlating time limits, and optimizing enforcement. Getting these policies right is imperative; curbside parking is a finite resource. Lots of people want it for a variety of reasons.
But before changes to rates, time limits, and hours are contemplated, parking professionals must consider the parking meter mix. We have a duty to get it right. Analytics can help determine the appropriate technologies on each block, making parking safer, easier, and more accessible across an entire city and region.
Oddly, parking managers and administrators rarely question assumptions about meter technology like we do parking policies. We’ve taken the manufacturers’ narratives about the benefits of single-space and multi-space meters as gospel, hardly giving these anecdotes a second thought. This post endeavors to interrogate these assumptions under the harsh glare of a swinging sodium light bulb of data. But before we begin our questioning, let’s recap how we got here.
Flashback: A Brief History of Metered Parking
The first parking meters were introduced in Oklahoma in 1935 to manage parking demand and generate turnover for businesses. It worked, improving sales at local businesses. By the following year, their success led to the New Republic labeling parking meters as “the next great American gadget.”
Throughout the 1940s and 50s—the golden era of film noir, coincidentally—single space parking meters (“SSMs”) spread across the U.S. and sprouted like stalks of corn in neat little rows on city streets. By the 90s, these meters had become fairly ubiquitous. Metered parking stalls were either marked with paint or in snowier climates designated by parking meter poles (referred to as “marked” or “demarcated” parking). For more than 50 years, the technology changed very little. Rates failed to keep pace with demand for three primary reasons:
- Public sentiment. With rates failing to increase in many large cities for decades, hourly fees lost some of their utility. Customers and merchants came to think of hourly meter rates as nothing more than a tax.
- Political will. Elected officials were reluctant to advocate for unpopular rate increases despite their importance to turnover.
- Technology. Traditional parking meters could not accept credit cards, so rate increases would have required motorists to pillage their couch cushions and shakedown shop owners for coins.
The American parking revolution actually started in Europe. Overseas manufacturers introduced computerized devices, or multi-space meters (“MSMs”), that took credit cards and supplemented batteries with solar power. The first MSMs (also known as kiosks or pay boxes) in the United States accepted payment for a particular space (“pay by space”) or provided parking receipts for display on the dashboard (“pay and display”). These meters resolved the technology divide. But with the proliferation of kiosks came public backlash rooted in a sense of loss. Some call it sehnsucht. For others it’s toska or saudade, a sense of nostalgia that’s hard to define with the English language. After all, parking is a commonality like eating or breathing, impacting all of us regardless of our individual wealth and fame. Feeding a meter is just something we do, and in that sense, parking meters have helped define who we are as a society.
Some American companies rebelled against this trend. They built their own smart single-space meter equipment, leveraging shrinking microchips and existing meter housings to network single-space meters. At the same time, MSM manufacturers doubled down on their technology by offering pay by plate kiosks, technology requiring a motorist to key in his or her license plate number.[2] Soon, app providers, domestic and foreign, were in on the action, making it easy for drivers to pay for parking with a cell phone or app by entering the space or zone number or their vehicle registration tag.
A minority of cities implementing MSMs took advantage of existing stall markings or meter poles to designate a space. A 2015 survey by the District of Columbia Department of Transportation (“DDOT”) found just 10 of 24 respondents (42%) demarcated at least some portion of MSM parking. Whenever MSMs were used in a pay by space environment, the spaces were always marked using paint, space number poles, or some other form of demarcation. However, the remainder (58%) chose to not mark their spaces. Many cited the expense and effort to maintain painted space markers, while half said that more cars are able to park on streets when spaces are not demarcated.
The Search for Parking Supply
Curb space is a valuable asset. Cities should seek to optimize the number of vehicles that can park on the public way. But the question remains: Do MSMs really create more parking supply?
A lot of esteemed parking and mobility professionals seem to think so.[3] Some offer anecdotal evidence based upon manufacturer representations. Some provide generalizations that may not necessarily prove out in the real world. Others have conflated the idea of "parking space" with "parking supply," failing to note how the manner in which the available parking footage on a block is used will impact supply.
City officials have a duty to their constituents, merchants, and commuters to question assumptions to determine if, like a bad alibi, they risk falling apart under scrutiny.
Following the Clues
At first blush, the claim of added space makes some sense. Moving from a pay-by-space environment to an undemarcated one (by removing painted lines or meter poles) should, theoretically, allow for more curbside parkers. On a block with nine or ten contiguous stalls, you could effectively gain two feet from each one assuming an average vehicle length of 16.5 feet. So, theoretically, city planners could realize 10% or more parking supply.
Of course, this calculation doesn’t necessarily work on blocks with eight or fewer contiguous parking spaces where the amount of space gained back is less than the size of an average vehicle. This is critical as blocks with fewer than nine parking spaces represent a significant portion of urban curbside parking systems—often a majority—as represented by a sampling of a few cities below.
In addition, the ability to gain a space or two assumes that there are no fire hydrants, loading zones, alleys, or other breaks in parking on the block. Break up the contiguous parking space, and it often becomes physically impossible to make room for another vehicle. The spaces must be uninterrupted for the math to properly work 100% of the time.
I know of no cities in the U.S. where there are at least 9 uninterrupted parking spaces on every single block face. Consequently, there should never technically be a uniform 10% system-wide bump from the addition of pay by plate or pay and display MSMs. The potential increase in curbside space will always be limited to a subset of the total metered blocks.
Another common mistake is in applying the 10% bump as a revenue/paid use multiplier. Municipalities can only benefit from gained parking spaces when occupancy is at its highest. Adding additional parking footage does little to improve revenue or the customer experience when there’s already ample parking available.
To the degree a MSM can increase supply, though, we must also address another factor: the potential impact of human behavior.
The Case for Marking Spaces
In Indianapolis, St. Louis, Cincinnati, and, most recently, as a pilot in Washington, DC, the recommendations of the Manual on Uniform Traffic Devices are being followed by marking spaces with meter poles, numbering, or curb markings.
While the Manual on Uniform Traffic Devices supports demarcation of parking spaces (see above), the Manual provides little more in the way of additional guidance. That, regrettably, leaves parking policymakers in something of a quandary. What priorities should they set to steer their decision-making? Should they focus on revenue? Costs? Efficiency? Or should they more heavily weigh the customer experience and congestion reduction in their policies? Parking professionals have largely had to solve this dilemma on their own.
Parking is more than just pulling or backing a vehicle into a space. Parking is a process, and a difficult one at that. The process begins before the driver ever puts the key in the ignition. It starts when he or she contemplates the trip and includes the drive, understanding signs, and searching for available spaces. It’s comprised of paying for parking, adding time as needed, and even dealing with parking tickets.
Fair parking systems gain legitimacy[4]. Legitimacy helps government to effectively realize its goals of improving revenues and reducing costs. In other words, customers want the parking process to be easy. When cities are responsive and give customers what they want, drivers are more likely to park legally. Public acceptance will drive revenue improvements and efficiency.
From a customer service perspective, single space parking meters seem like an easy option for customers. Unlike MSMs, there’s no need to walk down the street to wait in line to pay. Pay and display and pay by plate MSMs not only require a motorist to amble over to the kiosk, they may even require the customer to go back to his or her car to display a receipt or take a look at their license plate number.
Pay by space MSMs are slightly more complicated than SSMs but remain pretty easy to use. After walking to the MSM, the driver simply pays after entering a three- or four-digit space number then goes about his or her business. No return to the car is required.
This, of course, begs a question: Is the space number or the license plate number more difficult to remember? That’s difficult to say. There’s no empirical evidence to suggest a plate is easier to remember than a 3- or 4- (or even a 5-) digit number. In fact, just 40% of Americans know their license plate number. When memorizing larger numbers, we tend to “chunk” information, essentially breaking down a number into three or four digit segments. So, a plate generally requires the memorization of more chunks and variations. There are technically 78,364,164,096 possible permutations of letters and numbers in a 7-digit license plate compared to just 99,999 variations in a numeric 5-digit number. Still, it’s certainly easier to remember vanity plates and registration numbers that lend to mnemonics (e.g., using the work “peekaboo” to remember the letters “PKB”).[5]
Parking operators, especially those that leverage the customer experience to achieve other program goals, cite a number of additional reasons for continuing to demarcate spaces.
Pay by plate systems require near real-time communication. Enforcement personnel may not be able to verify whether or not a license plate is paid if:
- Communication speed (“latency”) is slow. There are a number of applications that must “speak” to each other: payment apps, parking meters, enforcement handhelds, and back-office systems.
- Offline meters create communication failures.
- Motorists or officers key license plates incorrectly. There is no safeguard against license plates that fail to conform to a standard format or length.
Whether due to latency, offline meters, or data entry issues, the potential for erroneous citations is very real. It’s an issue parking organizations face every day. In one eastern U.S. city that has largely transitioned to a pay by plate system, the error rate for meter transactions was reported at 1.3% (about 1.16% of that figure being directly related to human error). That may not sound like much, but it equated to 300 errors daily in 2015, up from 120 errors per day in 2013. While “fuzzy logic” algorithms can correct many of the errors based upon officer entries (by assuming officer entries are 100% accurate and matching against them)[6], voiding tickets behind the scenes presents specific problems. Customers that receive an erroneous citation, for instance, never get notice of these voids in the mail. Consequently, the burden remains on the vehicle owner to contact the City’s customer service line or contest the citation in order to learn if the citation has been non-suited. Further, payments towards bad tickets must be processed and refunded.
Undemarcated Parking: When Two Plus Two Equals Three
To address the impact of human behavior on supply, data scientists donned their fedoras and conducted an in-depth analysis using modeling and field data collection in two cities (we'll call them "Urban Neighborhood 1" and "Urban Neighborhood 2"). Our objective: to determine whether a specific parking configuration—demarcated or not—is more beneficial from a supply perspective.
Both cities employed free flow curbside parking, and our team utilized cameras to capture data on a number of blocks. Computer vision algorithms detected vehicles and parking events, specifically the vehicle arrival and departure times.
We captured the position of vehicles on each block and the lengths of cars, trucks, and motorcycles throughout the day for several weeks. As demonstrated below, we determined the distances between vehicles and the starting and end points on a given block. We were conservative in our approach and assumed that whenever the distance between two vehicles (or a vehicle and the beginning/end of a curbside parking zone) was 18 feet or greater, there existed an opportunity for another vehicle to park.
When seeking to forecast the impact of moving to undemarcated parking on supply, use, and revenue, parking managers should avoid simply applying a multiplier to current system performance. A factor alluded to above is variable demand. In order for cities to theoretically realize a benefit from pay by plate or pay and display MSMs across all hours of the day, demand must be high. When parking demand is low, the addition of potential supply provides no value. As demonstrated below on a representative block with nine potential parking spaces in an metropolitan neighborhood ("Urban Neighborhood 1"), occupancy was less than 50% almost one-third (30%) of the time.
The potential benefit, then, is at best limited to 70% of the parking events. However, as discussed below, the implementation of pay by plate or pay and display MSMs may actually have negative value.
Just the Facts
In Urban Neighborhood 1 we examined when five, six, seven, or eight vehicles were parked. Assuming undemarcated parking allows for up to nine cars on a block that’s 176 feet long, we’d expect to see enough remaining available parking for the addition of four, three, two, and one vehicle, respectively.
That is not what we found. Instead, we discovered that the number of available parking spaces was often less than the theoretical maximum. For instance, when five vehicles were parked, on average just 2.84 more cars could be added, a loss of 1.16 vehicles. The potential for parking nine vehicles was achievable just 3.9% of the time. Worse, more than 14% of the time there was actually a loss of two or three spaces.
As demand increases beyond five spaces, the ability to realize maximum occupancy (9 vehicles) declines. As described below, undemarcated parking not only failed to achieve 100% occupancy in all cases when six, seven, or eight vehicles were parked, it actually reduced available parking by:
- Up to 44% when five vehicles were parked,
- Between 11% and 33% when six vehicles were parked,
- Between 11% and 22% when seven vehicles were parked, and
- 11% when eight vehicles were parked.
Ultimately, less parking availability equates to a poorer customer experience. Against that backdrop, parking professionals should determine whether less availability might also lead to less revenue and more congestion. And to the degree there is a revenue benefit, is it possible that the improvement in municipal revenues resulted from more citations due to customer confusion or user error? Is it likely that credit card acceptance or other policy changes like rates or hours, and not supply, improved revenue? Would similar revenue increases have been realized if the SSMs were adjusted to shorten the parking spaces?
While pay by plate and pay by cell may reduce “piggybacking” on another motorist's purchased time, an Indianapolis study showed that piggybacking in pay by space systems can actually be quite low (especially with shorter time limits). And when pay by space MSMs are implemented, it’s very hard for drivers to take advantage of existing time. To what degree are the benefits of pay by plate and pay and display parking negated by trip chaining? These are all areas requiring further study.
The Ripple Effect
Why are parking spaces potentially lost in undemarcated systems? A motorist parking in an undemarcated space is provided no information about how he or she may optimize supply for other drivers. There are no markings on the public way, no meter or number poles, and no signs providing guidance. Because there’s little if any direction, the driver will presumably park as close to his or her destination as possible without any thought to maximizing efficiency. While that may not necessarily cause concern when supply is low, it may become a problem as demand increases throughout the day.
Like a pebble tossed into a pond, sloppy parking has a ripple effect. The initial event can incrementally impact behavior in the hours to follow. We experience this effect most often in traffic when an accident earlier in the day continues to create a jam long after any signs of it are gone. We also witness this emergent phenomenon with traffic waves when we park in undemarcated spaces.
As demonstrated here on a previously undemarcated block in Urban Neighborhood 2, parking behaviors compound and cement as use increases. Parking crystallizes with demand, forming distinct peaks in use and creating an uneven distribution of space in between parked cars.[7]
The result may be less efficient use of curbside parking space. Since there are seven distinct peaks noted above, let’s examine what happens on this block with seven vehicles are parked.
As noted below, when seven motorists are parked on this particular Urban Neighborhood 2 block, it’s very difficult to fit the theoretical maximum of eight vehicles. In fact, occupancy can only be maximized 16.5% despite the fact that the seven vehicles occupy an average of just 71% of the 156 feet of curbside space. In this case free flow parking effectively reduces available space by 12.5% compared to demarcated parking.
The Impact of Perception
Have you ever wondered why a group of witnesses to a crime may all describe the events and even the suspects differently? Why might a someone describe a suspect as a tall, middle-aged man in a suit while another person might swear the perpetrator was short in stature, young, and wearing a track jacket? Our perception is inherently biased and can often be erroneous. We focus on different details, and our recollection can be contaminated by later events. It makes solving a crime that much more difficult.
Our unique perceptions not only influences our ability to be an effective eyewitness, they also impact our ability to park. Xerox Research Fellow Chris R. Dance studied this phenomenon. When we approach a potential parking space on an undemarcated block, we must contemplate whether or not our car will fit. That decision-making process is influenced by our perception of the available space as well the length of our car. And our perception is influenced by our familiarity with the car and our own individual biases. Dance concluded that “a vehicle length of 14.75 feet requiring 3.2 feet of extra space...is consistent with a standard deviation of perceived lengths ranging from 0.2 feet to 2.5 feet, depending on the individual.”
If a driver doesn’t allow for some margin of error, he or she risks the embarrassment of making a series of “reversing moves” to park in an undemarcated space. If unsuccessful, the driver will need to find another space altogether. Some drivers might even pass up prime parking because of this uncertainty and fear of public scorn.
Drivers parking in a pay by space environment, however, are more likely to reason that since they previously fit in a demarcated space they should be able to fit their cars in a metered space again. In this way, the lines or poles that delineate the parking space are a lot like the height strips at the entrance of a convenience store. The latter provides guidance to eyewitnesses, helping them to provide accurate descriptions to the police. Similarly, Dance reasoned that demarcated parking provides cues to motorists and helps them improve their perception.
But it’s more than just perception. In a demarcated environment, the gaps are generally uniform, anywhere between 18 and 24 feet. On undemarcated blocks, there could be more gaps that are less than 18 feet (10 feet, 12 feet, etc.). That means that larger vehicles may need to driver further to find a parking space.
To test this, Dance simulated random vehicle departures using data collected on a number of blocks in Urban Neighborhood 2 and determined the distance between the remaining cars on undemarcated streets. He modelled a couple of scenarios for the necessary extra feet required to park (e.g., 2 feet to 7 feet) and found that:
- Larger vehicles may need to travel 4 times further in an undemarcated environment than in a demarcated one to find parking.
- Demarcation reduced the average distance to the closest parking space by between 5% and 50%.
- Motorists are more likely to find a space in demarcated environments (90%) than undemarcated ones (60%).
The Case of the Stolen Parking Spaces
Granted, we’ve presented just a few studies here. Every city is different and within each city, every block is different. Further, vehicle sizes differ from region to region and even within a municipality. Our studies are based on national averages, and, as such, fail to recognize potential differences in average vehicle length from city to city. Still, the claims that undemarcated environments increase supply have largely gone untested. Often anecdotal information is cited as fact. Each city should endeavor to understand the unique implications of free flow parking systems in its neighborhoods.
Efficiency is a critical factor in the decision to demarcate parking. The enforcement of pay by plate systems, for instance, can be supported by license plate recognition technology. In addition to efficiency, though, cities will need to other factors, like customer convenience, availability predictions and traveller information systems, and demand pricing strategies. City officials should be careful when highlighting the reasons for moving to undemarcated parking so as not to give undue weight to claims that may require additional study or could attract unwanted press inquiries.
In truth, these various systems can likely coexist, especially when pay and display or pay by plate MSMs are positioned on longer streets (1) where demand may not be as high (thereby avoiding challenges with queuing), (2) that offer significant turnover, allowing, as Dance notes, quick self-organizing to achieve efficiency (thereby addressing the wave effect), (3) where there is significant variability in vehicle types and lengths, and/or (4) that offer pedestrian and modal alternatives to address the longer distance needed to find a space.
[1] Parking is a pain. That’s chiefly because drivers can’t find parking spaces when they need them. As a result, roads get congested. Travel times increase, exacerbating pollution. Insufficient parking thwarts the goals of good mobility policy: complete streets, solving the first and last mile problem, and making streets safer (“Vision Zero”). Further, parking congestion creates confusion and frustration. The resulting aggravation leads to people getting hurt. Or worse. One of the leading causes of pedestrian deaths? Distracted drivers, often hunting in futility for a parking space. Sometimes, motorists even turn to violence, intentionally harming pedestrians, bicyclists, and other motorists.
[2] Andreas Jansson, Managing Director of Cale America, set forth a number of benefits of pay by space MSMs (+) and some potentially negative considerations (-) in a recent interview with John Van Horne for Parking Today (see the summary below).
[3] The esteemed professor Donald Shoup praised MSMs in 2002, noting that they lead to “better urban design” and “more parking spaces.” He wrote, “One multi-space meter replaces 20 to 30 individual post-mounted meters, and the pay-and-display procedure does not require striping the street to mark individual curb spaces. Multi-space meters thus reduce street clutter, hardware, and signs.” Further, “because individual spaces are not marked on the street, more cars can typically park at the curb than when the permanent placement of individual meters requires every space to be long enough to accommodate full-sized cars.” A variety of cities have said the same thing, most citing manufacturer’s claims that undemarcated parking leads to a 10% improvement system-wide or the addition of a space per block. In some cities, officials relied upon observations to make determinations like: “1 to 1.5 spaces are gained” or a "19% improvement in space" is realized per block.
[4] Legitimacy is the judgment by stakeholders, including drivers, merchants, pedestrians, and bicyclists alike, about the rightfulness of official conduct.
[5] The number of potential plate iterations could grow as states contemplate revising license plates to include symbols. Studies show that although 70% of crimes involve a vehicle, partial license plates are only recalled 15% of the time. Researchers have argued that witnesses are more likely to remember a plate number if it includes a symbol, like a star or circle. That hypothesis, though, is still being debated. In any case, MSM keyboards would need to be reconfigured to accept symbols in addition to alphanumeric values.
[6] Officer citations, by the way, are never 100% accurate. In every city, citations are regularly dismissed, nonsuited, or suspended because the vehicle registration number was keyed or written incorrectly.
[7] Twenty feet at the very end of this block is rarely used because it was allocated for persons with disabilities.