The vehicle utilization or pooling ratio determines the profitability of ride pooling services. This blog post introduces some hacks as food for thought. It may improve the efficiency for ride pooling and carpooling systems with a focus on suburban and rural areas.
Traditional mass transit services run on fixed schedules and routes or lines with large vehicles. Passengers adjust to these schedules and fixed routes. This increases the likelihood that many passengers share these vehicles at certain times, leading to a high occupancy, vehicle utilization and an efficient use of the vehicles.
In rural and suburban areas this poses a problem: due to the low density and longer distances these areas are often unprofitable and only financeable either via subsidies or with high pooling ratios. This is why many rural areas are underserved with publicly available mobility services as a result. In these areas the only alternative to owning a car is public transport or taxi services. The first one is often insufficient, the latter one too expensive for many. As a result, in the countryside you often live in a mobility desert. You own a vehicle – or you are not mobile at all.
Mobility as a Service (MaaS) with on-demand ride pooling services can offer point to point connections or act as feeder systems to mass transit in rural and suburban areas. On-demand services without fixed routes just lack the temporal and spatial focal points of scheduled times and stops at fixed routes. Ride requests are usually scattered across the time scale and a wide area. This issue doesn’t matter in high-density rural areas of city centers at busy hours: there will always be ride requests along almost any route, which can be pooled. But this picture changes as soon as you look outside of the city centers or at off-peak hours. This is why many ride hailing and ride pooling services only operate within city centers during busy hours – at locations and times when public transport usually serves these passengers sufficiently as well. So the societal benefit of such “raisin picking” ride pooling services in city centers is rather questionable. They would be much more beneficial in areas or at times underserved by public transport.
Just to avoid confusion in the terminology of the two forms of pooling: ridepooling refers to commercial transport providers, who serve passenger requests with pooled rides on demand, usually with mini vans. Carpooling in contrast refers to private, noncommercial rides offering empty seats to passengers traveling the same route. Also see the New Mobility Glossary.
Here are some approaches, which can help improve profitability of pooling systems by increasing the vehicle utilization and pooling ratio in suburban and rural areas:
- Straighten the routes
- Use time windows and book in advance
- Decompose trips into segments
- Reduce approach journeys
- Add carpooling
Hack #1: Straighten the routes
If passengers are able to walk a certain distance, they could go to pick-up points at a main route and so avoid the vehicle to take long detours. This won’t directly increase the car utilization but it increases the overall speed and reduces the cost of the service, which will attract more passengers and indirectly increase utilization and pooling rates.
This approach is not specific for rural areas. It also works well in dense city centers. Ride pooling services such as Uber Match or MOIA use this approach already and take advantage of the “spatial flexibility” of their passengers and place their pick-up points along main roads. Science has picked up on this topic as well, see e.g. the paper from Andres Fielbaum on optimizing a vehicle’s route when users might walk.
A bit of active mobility, i.e. walking or cycling to the next pickup point usually keeps people fit and doesn’t hurt – unless they have a lot of luggage or are walking impaired. For better accessibility for walking impaired people, mobility providers could set up personal pick-up points at the doorsteps of these passengers.
Hack #2: Use time frames and book in advance
We have mentioned the timely scattered ride requests before, which pose a challenge for pooling. Most on-demand services try to serve their passenger requests instantly. The rationale stated for this approach is most often that the customers expect this without questioning this assumption. The statement is probably true if expectations are set as such. Instant service is on one end of the service scale, fixed schedules on the other end. If the first one is not profitable and the latter one not meeting customer’s needs. So why shouldn’t we look for a potential sweet spot in between these two extremes?
Most often the need to take a ride doesn’t come as a surprise. People usually know well in advance when they have to be at a certain location and can arrange rides accordingly. Also people often have some temporal flexibility. Both facts could be used to aggregate passenger’s requests on a time scale by asking for booking in advance and for a time frame or time window instead of just one point in time. Both measures will increase the likelihood to find matching rides to be pooled. Pooled rides have lower costs per passenger kilometer traveled, which can either meet public transport companies’ budget restrictions or lead to financial rewards for the passenger’s temporal flexibility and pre-booking. For the few instant ride requests with no temporal flexibility left, a regular taxi will do the job – with its respective price tag.
For carpooling services it is a best practice to ask the passenger for a time frame instead of just a fixed departure point in time. The larger the time frame, the higher the likelihood to find matching rides. So passengers have an intrinsic motivation to give a broad time window. A similar approach could be used for ride pooling.
In a fictitious carpooling simulation with random car rides and low density suburbs, the ride matching probability with different time frame lengths was calculated. The results in this example show that an one hour time window increases the pooling probability from close to zero to almost 50%. So the usage of a time frame can be considered a prerequisite to make pooling work.
One auxiliary condition for the successful use of time frames is that these time frames are known to the mobility service provider well in advance so that ride requests can be matched and scheduled accordingly. Cancellations of such pre-booked rides have an opposite effect. The challenge using this approach will be to give passengers good reasons to
- reliably book well in advance and
- offer a broad time frame.
Hack #3: Decompose trips into segments
A scientific study about ride sharing or carpooling covering the Los Angeles area1 indicated that in low density areas or areas with long distances between origin and destination the carpool matching probability gets so low that pooling is not a viable option in most cases. A two-sided critical mass problem – both for commercial ride pooling systems and private carpools. Not a surprise.
On the other hand transport capacity in rural areas is often not an issue – considering all the empty seats traveling in private cars.This is why private peer-to-peer carpooling can be a solution for rural areas. The transport capacity is there. The car’s rides just don’t cover the entire route at the requested time but they cover parts of it. Composing these parts, hops, legs or segments into a joint ride significantly increases the matching probability. Airlines practice this approach successfully by using connecting flights at big airport hubs instead of offering point to point connections everywhere.
In the example above the likelihood to find a matching ride for trip #1 is very low. But you will likely find a matching ride for trip #2 into the city and another one for a subsequent trip #3 departing from the city. Even if only one segment is a pooling ride, the other segment could be covered with another mode of transport – making it an intermodal travel chain.
When we decompose ride requests into sections and fulfill these requests with a combination of ride segments, we need to consider two aspects:
- Passengers are resistant to change modes during a trip. Every change adds travel time, inconvenience, and bears the risk of missing a connection. Ideally there is not more than one change required per trip. The best place to change is the location with the highest density of trips on the route. It would be the ideal location for a mobility hub.
- The second segment should be flexible enough to cover delays of the first leg in order to assure a reliable connection. The second segment could also be served with another mode of transport.
When we used this approach in the aforementioned simulation, allowing for just one change, we could further increase the matching probability from 50% (orange line) up to close to 95% (green line) given an one hour time frame. So combining temporal flexibility with decomposing a ride up to two segments increases the matching probability for carpools from a dissatisfying “nope, won’t work” to a whopping “almost sure”.
Hack #4: Reduce approach journeys
Other than private carpools, commercial ride pooling services usually don’t pick up passengers on their way, which they would drive anyhow. In case they have only one passenger request to serve, the ticket fees do not only need to cover the “effective” ride, which the passenger wants to take, but also the approach and return ride from and to the base. In dense city centers these approach journeys might be short and there will likely be pooling requests on the way and at least one passenger request for the return ride. Serving rural areas this can be very different as the example below clearly shows: This example is a real journey I took with a ride pooling feeder service offered by a rural community for their citizens and tourists without cars in Austria: it is an incredible and exemplary service – but it has its downsides.
I just had to get to the next bus stop 2 km from my home in order to take the bus to the train station in the next city. Due to heavy luggage walking was not an option. My request was served by a taxi company based in a city 12 km away from my home. The taxi drove 24 km in total in order to serve a 2 km ride request. Alone, no pooling. If we calculate the car occupancy as an efficiency metric, we need to divide the value-adding effective passenger ride distance by the total vehicle distance traveled. Comparable to an autonomous car service, the drivers do not count for car occupancy as they just operate the car but would not drive this route without my request otherwise. So the effective car occupancy on this ride was just 2km/24km = 0.08. To put this figure into perspective: if I would have taken my own car my car occupancy would have been 1.0 – and even if my neighbor would have given me a lift and then returned “empty”, the effective car occupancy would still have been 0.5.
From the entire ride I just had to pay for the effective 2 km ride and the community paid for the 12 km approach ride and the 10 km return ride. What a waste for the community! And no comment on the carbon footprint.
This example nicely illustrates the importance of short approach distances, which can be a challenge in rural areas. To avoid such inefficiencies mobility service providers in widespread areas should not rely on vehicles at centralized locations but have some vehicles located in highly frequented locations across the area – depending on the provisioning costs.
An alternative to drivers who are constantly on stand-by and incur costs would be to make use of part-time drivers across the whole area, who only get active and incur costs in case they serve a ride request. The gig economy business model.
The beauty of rural areas is that people are usually more willing to help than in cities. Many communities take advantage of this fact and organize local mobility services on a volunteering base, which nicely leads to hack #5.
Hack #5: Add carpooling
As seen in the example above the most climate-friendly and cost-efficient chauffeured ride service would be the one with ideally no approach journey, i.e. my neighbor giving me a lift – if possible if the neighbor drives to the destination anyhow. The private peer-to-peer carpooling model. I would just need to know which neighbor could give me a lift at what time – the classic carpooling system task.
The problem with carpooling is its lack of predictability. You never know if there will be a suitable ride available or not. So passengers would need a backup option, a plan B. This backup could be an expensive taxi or the aforementioned commercial ride pooling system. Together with scheduled and fixed route based public transport such an offer can improve mobility in suburban and rural areas and reduce car dependency.
The beauty of this combination would be that communities can provide a reliable mobility service in rural areas banking on the commercial transport service providers combined with the cost-saving private carpooling option, if available. As carpooling will usually be much cheaper and have a lower carbon footprint, it should be the preferred option. Finding the right incentives for drivers and passengers to offer and accept private rides will be key for its success. All could be offered as one integrated mobility service in the regular public transport routing and booking application.
Spatial-temporal flexibility increases matching probability and in consequence the pooling ratios. Each of the discussed approaches has its benefits but they are most effective in combination.
A solution combining these approaches could be much more effective and efficient in suburban and rural areas than a simple “pick up where they are when they call” software. It requires a software platform that
- works with fixed pick-up points and can guide passengers to these locations,
- can handle pre-booked rides,
- supports time frames for passenger requests,
- allows dispatching from multiple vehicle locations and
- combines commercial ride pooling with private carpooling.
Rate this post.