Written evidence submitted by Mr John Geddes (BUS0023)
I am a lone individual, living in a rural area and approaching the age-group where the availability of public transport could make a big difference to my life.
As a former McKinsey consultant and mathematical modeller, I have devoted time to scrutiny of DRT proposals to try and work out whether they really can be expected to deliver what they promise.
“Bus Back Better”, the National Bus Strategy for England, makes extensive reference to the use of Demand Responsive Transport (DRT) as a likely way forward for provision of public transport in rural areas.
My research suggest that DRT is not delivering value for money in such contexts – nor is ever going to be able to do so. In most deployments it will deliver little more effective capacity than a taxi service (and perhaps less), at much higher cost.
This matters because many BSIPs propose DRT to replace subsidy for timetabled services in rural areas. The operating cost will be fixed – the unknown will be what proportion of transport needs such a service can meet, and the evidence is that the number of passengers served will be low.
If such changes go ahead, the number of rural people unable to access public transport will soar, and it will be only the very richest who can responsibly plan to spend later life (and the risk of eventually being unable to use a car) in a rural area where a taxi will be the only option for most of their travel needs.
I have looked in vain to find any comprehensive published study of the performance of DRT in rural areas.
Among the reviews that have been undertaken into rural DRT, all appear to limit their curiosity to the opinions of those lucky enough to get a taxi-like ride to fit their needs (spoiler alert: they think it is great, just as lottery winners would consider their ticket purchase good value). Invariably, reports avoid any analysis of the cost per passenger – the Transport Focus report on Fflecsi DRT in Wales (Feb 2022), for example, gives absolutely no consideration to operating costs.
I have undertaken my own data analysis on one DRT scheme local to me in Derbyshire. I looked at the (very cursory) data that had been recorded for the “Derbyshire Connect” scheme around Ashbourne for the year before Covid started. Making what I think are reasonable estimates of the extent of home-school journeys that are mixed into the data, I estimate that in “DRT” mode the vehicle was averaging one passenger for every six miles driven. Given the topography of the operating zone, that is close to what one might expect from a taxi (one or two people being driven an average of four miles, with the vehicle returning empty). Such service provision would be far cheaper delivered using a taxi than with a 16-seater minibus, or possibly by a timetabled service using a full-sized bus, in cases where the authority is already covering the fixed costs of the vehicle in order to provide home-school journeys.
I have also used social media to seek out others’ experience of DRT. There is no natural channel to reach DRT users (and unsuccessful would-be users) but I did manage to engage with several people who have an interest in how DRT is developing. Nobody could point to AADRT schemes that were achieving high loadings – again and again, reports talk of vehicles achieving loadings little better than a taxi.
To try and validate my views, I have also spoken with Stuart Eccles, the key official behind Call Connect in Lincolnshire, one of the UK’s most celebrated and long-running DRT schemes. Stuart gave me a generous 45 minutes (at a weekend) to answer my questions. His answers suggest that while DRT will “never be cheap”, it can perhaps offer a good customer experience at a sensible cost – but only in very specific conditions that will not apply in most areas and, crucially, only if other services are carrying the majority of the fixed costs. (Stuart Eccles said he was not planning to make a submission to your committee – if you are looking for expert input to any of your sessions, he would make an excellent invitee).
“Demand Responsive Transport” is a very broad term, used to cover very different services. The BSIP proposals for rural areas, as far as I have surveyed them, seem to be intending the “anywhere, anytime” model (“AADRT” in my terminology) where one or more vehicles are assigned to a geographic zone, and passengers are invited to request journeys between any two locations in the zone, at any time within the scheme’s operating hours. Journeys are allocated on a first-come, first-served basis when bookings open, typically 7 days ahead. A few areas may be contemplating the system supported by one of the software companies in the field, where booking requests are only considered as passengers are ready to travel.
The ostensible justification for moving from subsidy of timetabled bus services to the provision of DRT is to provide transport to those who need it, at a lower cost per passenger than can be achieved using timetabled services.
The contrast between the picture painted and reality is large – to the extent that I have come to question whether lower-cost journeys are really the driving force in all cases.
The more cynical explanation could be that over-optimistic councillors and overwhelmed officials are complicit in engaging in wishful thinking, egged on by software and vehicle suppliers who have spotted a lucrative market opportunity.
AADRT allows authorities to set their subsidy at whatever level they choose, with the ability to say to any potential complainant that they are indeed providing a transport service that fulfils their statutory obligation.
The adequacy of a lottery for places in fulfilling the statutory obligation is legally unchallenged, and the public relations damage of a user complaining “I couldn’t book the trip I wanted” is much less than former passengers complaining that “our 10.30 bus has been cut”. As an additional attraction, in future budget rounds, the service can be thinned further without major PR damage. Passengers may have had the odds of a securing a trip come down from 1 in 3 to 1 in 4, but they are not going to be able to prove it, and it is too complicated to make a compelling complaint media story.
Crucially, AADRT needs to be able to use appropriately-sized vehicles, for which other services are carrying a large proportion of the fixed costs.
In Lincolnshire, my understanding is that they are well on the way towards “Total Transport” – the same vehicles serve Adult Social Care, Home-to-school transport and AADRT at different times of the day. Adult Social Care usage apparently makes a major contribution towards the fixed costs of the vehicles, and school transport another sizeable share. That leaves AADRT paying a fraction of the fixed costs, plus the marginal, per-mile, costs.
I have no way of knowing whether the allocation of fixed costs for Lincolnshire is reasonable: if Adult Social Care were carrying a disproportionate share of the costs (because external transport providers there are extremely expensive, perhaps), then the cost of AADRT will have been be artificially reduced compared to a realistic contribution that could be expected in other areas.
Using AADRT vehicles also for home-school transport has the potential to spread costs, but only to the extent that the sensible size of vehicle for AADRT is roughly the same as for home-school transport. For AADRT, the right size will be a small vehicle because the typical passenger count is low: Lincolnshire runs several 8-seaters; the Fflecsi review suggests adding small vehicles to their fleet. But only in the sparsest rural areas is that a sensible size of vehicle for home-school provision. Buses of 20 seats upwards are the typical vehicle of choice for home-school transport in many rural areas.
If AADRT uses a separate fleet of vehicles (or results in either AADRT or school transport carrying extra costs because they are obliged to run wrong-sized vehicles) then the true cost per passenger journey will rise from very-expensive to ruinous.
Where currently a County Council is currently funding a combined operation (school journeys from the Schools budget, middle-of-the-day public services from the Transport budget), a separate AADRT fleet will mean that the Schools budget will face a higher cost as bus operators are no longer receiving anything from the Transport budget. A school-service-only contract will cost considerably more to deliver than a contract which dovetails with a Public Transport contract for the middle of the day.
AADRT loses the aggregation that is achieved by a timetabled service.
A timetabled service enforces compromise on where you will travel and when – so the 1030 bus from B to C may be carrying people who would prefer to travel B-C at 0930 at others who would prefer B-C at 1130. It may also carry people from B who would rather shop in D, but will make do with C as the bus goes there. Move to AADRT and each passenger manifests their ideal itinerary as a separate demand without compromise.
The flexibility is great for those who hit lucky – they have the convenience of a taxi for the price of a bus. But this newly-disaggregated demand will not fit with efficient use of vehicle or driver.
A bus timetable is constructed to have the vehicle earning revenue for as much of the time as possible. But current AADRT systems cannot plan efficient schedules.
If you are going to add trips to the schedule as requests come in (the general AADRT model) then if you are in a busy urban area with many vehicles covering each zone, there may be only a modest reduction in utilisation because of the randomly-planned schedule.
But as zones get bigger and the number of vehicles gets fewer, so the efficiency of vehicle use will plummet. A multiple-vehicle area with a single destination hub may achieve an acceptable utilisation, from the experience reported by Lincolnshire. But if you can’t afford that level of provision, you end up widening your catchment (to the point that a zone contains more than one popular destination) and reducing the number of vehicles. Vehicle utilisation will drop dramatically, as the randomly-constructed schedule sees more and more (and longer) “dead mileage” runs as a trip from A to B is followed by one from C to D and then from B to A again
I understand that Lincolnshire use human intervention in the construction of schedules for a single-vehicle zone even with a single dominant destination, because unconstrained first-come-first-served scheduling has proved to result in extremely poor utilisation. And they have avoided setting up any AADRT zones with multiple destinations because the likely utilisation is too low. There is evidence for this view: I understand that an experiment in Northamptonshire, with a single zone covering Towcester and Brackley (two towns of comparable size, 12 miles apart) suffered very poor vehicle utilisation because of the many empty repositioning moves between the towns.
The challenge for the AADRT proposals in BSIPs is that many plans appear to be for a very thin service: single vehicles in large zones. This is just where AADRT achieves the worst vehicle utilisation, and therefore the highest cost per passenger. If an AADRT vehicle is darting around between points where only one or two people want to make similar journeys, and then running empty to pick up one or two people for a completely different journey, the limited capacity is going to result in relatively few of the former timetabled bus users being able to make a suitable reservation under AADRT.
(AADRT could achieve higher efficiency, by moving away from making instant decisions on accepting or declining bookings – but at the expense of a much more complex system. See Appendix for details.)
It would be worth checking a few of the BSIP bids in detail to see whether AADRT proposals are designed to achieve competitive contract pricing.
AADRT vehicles are expensive, and an operator buying them needs to be sure either that they have a secure contract for long enough to recover the bulk of the purchase cost, or that there will be a strong resale market if they have a shorter contract that is not renewed.
If you fear that you may end up with unwanted vehicles on your hands in three years’ time (at the same time as other operators may find themselves in exactly the same position), you won’t be confident of good resale value. If you bid at all, you will want to pad your quote with a generous margin to cover this risk.
I understand that Lincolnshire finance the vehicles themselves, and then let contracts for their operation. This seems a good recipe for achieving competitive costs as it removes the “vehicle risk” from the contract and makes it more attractive to bidders. In relatively sparse rural areas, you need to make your contract attractive to every possible operator if you are to have a chance of competitively-priced bids.
AADRT has the major problem of removing predictability from public transport.
From my (admittedly limited) research, it appears that even the strongest proponents of AADRT don’t pretend that DRT is suitable for journeys when people need certainty more than a week ahead.
A timetabled bus service is (Covid era excepted) locked in for at least six weeks ahead, and in practice most services go many months without major adjustments. The local health centre, dentist, chiropodist and hospital (and perhaps even Job Centre!) can have a copy of the timetable, and can offer appointments to fit the bus times.
Those planning a long-distance journey on public transport can book a less-expensive fixed-time train or coach ticket when they go on sale several months ahead - with the reasonable expectation that there will be a bus at the time shown in the timetable.
But AADRT wrecks these usage cases (and many others). The flipside of the flexibility that is touted as a key selling point is that the routing and timing of the service becomes completely unpredictable. If you are the first person to get through when bookings open, then you do better than with a timetabled bus – you get to choose exactly when and where you will travel. But if you aren’t so lucky, you may find that by the time that you get through to book, the schedule can’t add an extra trip to meet your request. AADRT runs out of schedule slots well before it runs out of seats.
If the schedule includes an earlier trip from your community to your destination, then you could join that and wait around. But that is far from certain. There is absolutely no promise that the AADRT system will be able to carry you where you need to go on a particular date, let alone at a particular time.
So the AADRT user cannot commit to being at a particular place at a particular time any further ahead than the time at which the “booking window” opens. That is most commonly 7 days ahead – far too late for many purposes. When a hospital asks what time you want to see the specialist, six weeks ahead, the AADRT user can hardly say “can I let you know 7 days ahead, when I’m going to find out when I can get a ride?”
Some systems work on “live” scheduling, which is even worse for predictability. Users only make their request when they are ready to travel. And if you are unlucky when you make your request (because the schedule is already filled with journeys that are incompatible with your needs), you won’t get to find out until it is too late to book a taxi instead, even if you could afford it.
Aside from the very special set of circumstances where AADRT can be claimed to meet the needs of a large proportion of those requiring transport at acceptable cost, I would suggest that it would be better to:
March 2022
AADRT could achieve higher efficiency, by moving away from making instant decisions on accepting or declining bookings. There’s a lot more hassle for the user, but potentially much better utilisation of the vehicle and driver as requirements can be shuffled around to offer a more sensible itinerary.
For a set period (say 10 days ahead to 8 days ahead of the day of travel), the system invites requests: the passenger sets limits on when they can leave home, by what time they need to arrive at their destination, and the minimum time that they need to stay at their destination before returning home.
These requests are like tenders for a contract – a promise to go ahead if an offer is made.
At the close of the first phase (perhaps 8 days ahead of the day being scheduled), someone (or some software) looks at the requests and builds the best possible schedule, measured on how many requests can be accepted within their travel time limits, without any passenger suffering an inordinately protracted journey.
Finally (7 days ahead in this example), requestors are contacted to tell them whether the system has been able to accommodate them, and to tell them where and when they can join the vehicle.
After this point, additional passengers may be accepted if they can fit within the set schedule, but these passengers’ journeys won’t shape the schedule.
This approach is expensive, as it involves systems and/or people to make the extra contact.
This approach is inconvenient, as the user has a period when they have committed to making a journey that fits their parameters, but don’t know whether they are going to be accommodated within the schedule.
But it does offer a less wasteful use of very expensive vehicles and drivers. It would require careful modelling to establish whether the improvement is enough to make AADRT a cost-effective method of providing transport in rural areas. It is quite possible that better scheduling would improve the cost per passenger from “unbelievably high” only down to “eye-wateringly expensive”.
I have not been able to find any AADRT systems that currently work in this way, nor any published plans for systems of this type.