Wellington Scoop

Hubs and spokes: the design flaws of the new bus network

by Lim Leong
Here is the plain English version of why a hub and spoke transport network will never work in Wellington without an underlying support infrastructure. I invite you to compare what US consultant Jarrett Walker said with what I am saying, and see which version makes more sense.

There are two fundamental physical transport network design topologies – “hub and spoke” vs “point to point”.

A point to point network has a direct route from any point to any point in the network.

A hub and spoke network organises routes as a series of “spokes” that connect outlying points to a central “hub”. To get to your destination, at least one transfer is required unless the hub is your destination. The main benefit of a hub and spoke network is that it simplifies the network configuration by reducing the number of routes required to service all points within the network. Delta Air pioneered the hub and spoke transport network paradigm in the 1950s.

The theoretical hub and spoke network design is a well studied subject and its benefits, drawbacks and constraints are well understood. In the real world, it is often a hybrid network of some sort. The old bus network was principally a point to point network with some hub and spoke design. The new network is the reverse – principally a hub and spoke network with some point to point routes.

Hub and spoke design is a valid network architecture and it can be and has been successfully implemented overseas (Singapore, New York, London are examples). However, for it to be successful there must be underlying support infrastructure like multi lane roads, dedicated bus lanes, bus priority lanes, purpose-built hubs.

The reason the underlying infrastructure is so important is that a hub and spoke network relies heavily on transfers and for transfers to work, everything has to be on time. A delay on one route will generate further consequential delays at the hubs and connecting routes causing a domino delay effect throughout the entire network.

Does anyone think they can make a bus be on time even on a good day with Wellington’s single lane, steep and winding roads? The answer is of course it is impossible to be on time without the support infrastructure. In contrast, the old network was largely point to point and not rely on transfers. Delays on a point to point network only affect that route with no consequential delay.

The reason the hub and spoke design was chosen is that GWRC wanted to reduce bus traffic on Lambton Quay, citing congestion and capacity reasons. Hub and spoke offers “demand aggregation” meaning high frequency buses can take you into a hub (for example, the Railway Station) and then the less frequent feeder buses will take you across town. In this way, bus traffic will be reduced on Lambton Quay as compared to the old point to point network.

The trouble is that a lot of people are now driving cars in to CBD due to network failure. CBD congestion could be worse than before, negating the purported benefits of a hub and spoke network. Personally. Personally, I am taking the car – ferrying kids to and from schools for exams, after school activities etc because one can’t rely on buses for important appointments.

This hub and spoke network is never going to work in Wellington because none of the important routes have dedicated bus lanes or multilane roads. One minor hold up and you are history. A possible workaround (albeit a nasty one) is to schedule lots of slack/buffer into the time schedule but that automatically makes the journey 2 – 3 times as long when compared to the old network. I bet this is what GWRC will be looking at as a cheap fix.

Another really bad side effect is even a short journey less than 3kms (eg. Northland to CBD) now incurs a transfer which is unnecessary in my view. Designing a 8-hub model for a compact city like Wellington knowing there isn’t any supporting infrastructure is blind faith at best and professional negligence at worst.

It is hard to fix a hub and spoke network. A simple change on one route or timing changes can have unintended bad consequences throughout the network because everything is inter-dependent. The whole network has to be re-balanced with any change.

In a data transport network, there are network traffic analysis/simulation programs which let you assess the systemic impact of route changes on the entire network. In the case of the Wellington bus network, I bet my bottom dollar that there is no such simulation model because it is impossible to model the traffic behaviours of Wellington’s single lane, narrow and winding roads. IMHO, this hub and spoke network is unfixable because none of the design pre-requisites are present. The only solution is a rollback (partial or full). A roll back will immediately reduce the adverse effects for commuters and it will also buy time to consider longer term public transport options.

In summary, there won’t be any meaningful improvements until GWRC is prepared to acknowledge the basic design flaws.

Lim Leong was a management consultant with KPMG and is now a senior Programme Manager responsible for complex delivery programmes at Spark NZ


  1. David Mackenzie, 9. October 2018, 11:32

    A very good analysis. However, the GWRC will fight vigorously against any return to the previous routes. It would be necessary to replace the GWRC council first. That requires government intervention: either a local act to set up a board, or to replace the GWRC with a commissioner.

  2. Mike Mellor, 9. October 2018, 14:53

    Overall a good analysis, the emphasis on the necessity of punctual running being important. But there are some things that need comment:

    “The new network is…principally a hub and spoke network with some point to point routes” – not so. While it is much more dependent on hubs than previously, most journeys are still covered by point-to-point routes, with just half-a-dozen feeder routes. In the peaks, every suburb except Vogeltown has through buses to/from the CBD (but not necessarily along the whole of the Golden Mile).

    “Does anyone think they can make a bus be on time even on a good day with Wellington’s single lane, steep and winding roads?”, “it is impossible to model the traffic behaviours of Wellington’s single lane, narrow and winding roads” – it’s congested roads that are the problem, not these roads; buses can and do keep time here, and traffic behaviour is predictable.

    “The reason the hub and spoke design was chosen is that GWRC wanted to reduce bus traffic on Lambton Quay, citing congestion and capacity reasons” – this is not actually the case, since at peak times nearly all the hubs are bypassed by through services to the CBD. Bus numbers along the Golden Mile have been reduced by having larger buses and by terminating peak-only routes from the north and west at Brandon St, not by hubbing.

    “The trouble is that a lot of people are now driving cars in to CBD due to network failure” – it would be good to see any data and evidence.

    “none of the important routes have dedicated bus lanes or multilane roads” – not so: there are examples of both on the important spine routes and elsewhere. But they are not enough. At least along the spine routes, GWRC and WCC need to be implementing a programme of allocating road space to the most space-efficient modes. It’s very disappointing and short sighted that GWRC failed to recognise this in the planning of the network.

    “schedule lots of slack/buffer into the time schedule but that automatically makes the journey 2 – 3 times as long when compared to the old network” – all public transport schedules have an element of slack time in them, but to suggest that this would double or triple journey times, let alone automatically, is absurd.

    “this hub and spoke network is unfixable because none of the design pre-requisites are present” – so the hub-and-spoke elements of the current network are fixable with proper infrastructure (e.g. bus priority and convenient hubs), decent timetabling (e.g. connections consistently of just a couple of minutes) and a proper appreciation of passenger demand, all of which could and should have happened before July.

  3. Lim Leong, 9. October 2018, 16:44

    @Mike Mellor. A few further clarifications regarding your points.
    “While it is much more dependent on hubs than previously, most journeys are still covered by point-to-point routes, with just half-a-dozen feeder routes.” I really would like to see hard Snapper evidence from GWRC, which claims only 5% of journeys requir transfers but anecdotal evidence suggests otherwise. To go to a lot of places now requires at least one transfer. GWRC had said that around 8,000 – 10,000 trips are transfers and this looks like far more than 5% of journeys to me. We also should not forget many people have gone into cars/Uber/Taxis since going live because they can’t rely on connections to work for them. This could well skew the Snapper data. Lastly, if only 5% of journeys require transfers then why design a 8-hub model to start with? 3- 4 hub model is more than enough for the size of Wellington.

    “All public transport schedules have an element of slack time in them, but to suggest that this would double or triple journey times, let alone automatically, is absurd.” This is not absurd at all especially for short journeys with the current network. Take a real life example. Before July this year, the direct journey from/to Wellington College to/from Northland, direct route used to take around 30 – 35 mins on a good day. Now the total journey time with the new network with transfers is easily 1 hour. If you add a further 10 mins of slack time to the schedule, the journey easily takes twice as long compared to the old network and that is assuming you don’t get bumped off at peak time. I am sure there are lots of other examples throughout the whole city.

    It is now 3 months since going live. The bigger question is can anyone explain to the people of Wellington what benefits has this new network actually brought for commuters and ratepayers. I would like to see the hard evidence and data and not PR spin (eg. counting transfers as complete trips). All I am seeing is there are well over 1000 people attending the 7 – 8 public meetings expressing their disappointment and anger.

  4. Michael Gibson, 9. October 2018, 17:31

    This excellent analysis by Lim Leong rightly gives Northland as a typical example of what is wrong by saying: “even a short journey less than 3kms (eg. Northland to CBD) now incurs a transfer”.
    What on earth can be good about increasing the length of such a journey? The buses from the Western suburbs which are forced to use the Railway Station as a hub have to turn left at the bottom of Bowen Street before returning the x hundred metres and again passing the bottom of Bowen Street before travelling along Lambton Quay – that is, of course, if one can get the right bus at the Railway Station. And the buses are guzzlers – can the extra trip be good for the environment?
    This is typical of the same sort of ridiculous problem elsewhere. Councillors: just fix it!!

  5. Tania, 9. October 2018, 18:21

    I think these ideas are valid and well researched.
    GWRC please take note. Bus users have suffered enough.

  6. C Benson, 9. October 2018, 18:22

    Mike Mellor, thanks for calling out lack of service from and to Vogeltown. Many of the commuters have voted with their feet and are driving, scootering, cycling, walking or bussing to the Kingston terminus to start there. Those who have persisted with the transfer have not been able to board the packed no. 3 or have watched as the transfer bus drives off leaving them to wait another 15-20 minutes. This happens both ways. Vogeltown now has a pretty much empty bus sailing between Kingston and Houghton Bay with extended evening and weekend hours. The waste and cost to the ratepayer is shocking

  7. Maria Gobbi, 9. October 2018, 21:25

    This is an excellent article explaining why the hub and spokes model is not working well for Wellington, and why tinkering with some of our bus routes will not fix things for most of us.

  8. h, 10. October 2018, 11:16

    Thanks for keeping the focus on the buses. I’m worried that, once the very worst of the problems are kind-of-sorted, everyone will be so exhausted with the topic that we’ll be left with all the sub-optimalities of the new system.

    I’ve got two main points here (mainly repeats of others points)

    1) the Snapper data isn’t going to show the full extent of the transfers needed under the new system, because the connections are so unreliable that I at least am doing everything I can to avoid needing to make a connection. For instance, the Number 21 is supposed to go from Courtenay Place in the evening and intersect with the Mairangi at Victoria University for a transfer. This almost never works (for instance with the buses supposed to leave at 7pm last night). The best way I have found to get from Courtenay place to Northland is to climb the 177 steps of Allenby Terrace & get on the Mairangi at the intersection between Salamanca rd and The Terrace – this is not a transfer. Requiring this level of personal effort to make the bus system work is awful (& inappropriate for people with heavy bags or mobility problems) – but won’t show up in the Snapper data as a problem.

    2) If they were hell-bent on a hub system why didn’t they include the cable-car as a hub? It has a purpose built shelter at the top & an utterly reliable 10 minute service & gives access to Kelburn/Northland from the centre of the CBD. It would have been easy to route the Mairangi buses past the top of the cable-car for the transfers. It would also have the advantage of bypassing the worst of the congestion/ dodgy roads. It was originally built as a commuter service & is owned by the people of Wellington. It’s ludicrous that it’s currently not part of the commuter network & is priced for tourists not locals.

  9. Dave B, 10. October 2018, 13:29

    Why don’t we consider hubbing for car-users? Build a great big carpark somewhere on the outskirts and force them all to change onto a bus. Think of the benefits to the city in getting rid of the traffic!
    Far-fetched? Yes. Tongue-in-cheek? Yes. But uncannily like what is being forced on bus-users.

  10. Jonny Utzone, 10. October 2018, 15:08

    Dave – we already have an out of town mega car park hub. It’s called Petone rail station! We need to start charging for car park space at rail stations like Vancouver, Toronto and the UK have been doing for 40 years. Come on NZ, get with the CO2 reduction plan! Its time for ratepayers to stop subsidising car parkers.

  11. greenwelly, 10. October 2018, 15:50

    @Jonny U, Dave B – Plus Porirua has 800. In total there are around 6000 Park n Ride spaces across the network.

  12. John Rankin, 10. October 2018, 16:10

    I hope we will see a continued focus on making bus services more reliable, rather than giving up on providing connections that widen the kinds of trips people can make by bus. Because:

    1. Buses don’t run on time not through some set of unique Wellington features but because we have not prioritised and invested in the infrastructure needed for on-time performance. Why would we find this acceptable? Doing the hard work necessary to deliver a reliable service shows you respect your customers. Or not.

    2. As several others have pointed out, there are many things we know we can do to make bus services more reliable, some more expensive and more difficult than others. Some changes are easy and cheap. There is no better time to apply pressure for these changes than now.

    3. If the rumours are true, LGWM will shortly be coming out with a recommended programme of investment that includes rapid transit, probably based on light rail. One of the critical success factors underpinning the economics of rapid transit is reliable connections to feeder bus services. If we don’t have a reliable bus service, we are setting a $1bn investment up to fail. Getting the reliability we will need (between 1 minute early and 2 minutes late, all day every day) is likely to be a 10 year journey. We need to start yesterday.

  13. Morris Oxford, 10. October 2018, 16:15

    Dave B, car-hubbing would mean getting into another car. Not a bad idea for Wellington where bus-hubbing needs to be avoided at all costs.

  14. Michael Gibson, 10. October 2018, 16:24

    I have just read h’s suggestion about the top of the cable car. I have yet to have an adequate explanation about my being kicked off a bus at Vic when I wanted to stay aboard and it was going to the top of the cable car to turn around. I suppose the American designer promoted this in 2011 when he finished his work and got paid.

  15. kiwi_overseas, 10. October 2018, 16:40

    Putting aside the operational issues of not enough drivers and the wrong size buses on some routes, GWRC & NTZA need to answer the following regarding risk management of such a large network change:
    1) As a PT funder, did NZTA technically peer review and sign off the very large bus network structure changes? If not why not (does the process need updating?), and if so what were the findings?
    2) Did the analysis GWRC undertook include vehicle simulation to ensure there were no vehicle congestion issues? Was this peer reviewed?
    3) Did the analysis GWRC undertook include demand simulation so that effects of the new network & transfers etc could be considered? (may not have been done as there are few single agent demand/vehicle simulators available). If it was done was it peer reviewed?
    4) If 3) wasn’t done, was the demand and new bus network analysed using the strategic travel demand model for existing demand (not just mainly the forecast future demand as for a business case)? If so what were the results and were the results peer reviewed?

  16. Lim Leong, 10. October 2018, 17:36

    @John Rankin. I do not disagree with your analysis that a transfers based system can be made to work with the underlying support infrastructure. A transfer based system is probably a pre-requisite to support any future rapid transit option. However, we all know that establishing infrastructure takes investment and time, neither come easily nor quickly with the central government’s and local government’s bureaucracy. There are no quick fixes for a hub and spoke network.

    We don’t have to look far, this new network is 7 – 8 years in the planning, and we still do not have any decent hubs built in time to support the launch. It will take years to establish the necessary underlying infrastructure to support a proper hub and spoke transfer based system. Meanwhile there is absolutely no justification for commuters to continue to suffer the failures of this new network. I know of many people who have reverted to using cars while those without cars are incurring financial burden of using Uber/Taxi. There is also a significant group of people who neither have cars nor the financial means and they continue to spend hours of the life everyday fighting this new network. There are also older people who feel they are trapped in their homes due to their mobility issues and fear of using the transfer system which is known to be unreliable.

    My recommendation is GWRC should acknowledge the fundamental issues. Do a rollback (partial of full) but keep parts of the new network which are working. This will bring stability and reduce the sufferings for commuters and it will also buy time to consider and establish the underlying infrastructure.

  17. Graham Atkinson, 11. October 2018, 10:29

    Michael Gibson: not allowing passengers to travel beyond the terminal point on any route is standard practice worldwide. I am assuming you were travelling from either the Lambton Interchange, Courtenay Place or Newtown on a service with the destination “Victoria University”. This is the terminus and all passengers have to be off the bus when it commences any “Not In Service” travel.

  18. John Rankin, 11. October 2018, 12:43

    @LimLeong: I would add that the Public Transport Spine Study identified a suite of comparatively low-cost improvements which would have delivered “bus priority” on the main PT spine. These would have gone a long way towards preventing the kinds of problems the new network design has created. Once the recommended “Bus Rapid Transit” option was put on the back burner, why did GW then opt for the “do nothing” option?

    The risk of a rollback is that it will not necessarily deliver reliability. In my view, being on time is a reasonable customer expectation for any bus network, not just one with a hub and spoke design. Even if we do a network rollback, we still need make it a priority to address reliability problems.

  19. Lim Leong, 11. October 2018, 14:34

    @John Rankin. I fully agree with your statement that “Even if we do a network rollback, we still need make it a priority to address reliability problems”. GWRC also needs to pay attention to some Eastern and Western Suburbs who are now poorly served by this new network.

    @kiwi_overseas. Valid comments from you regarding risk management and due diligence. I hope GWRC can be more transparent on these matters as I think one of the biggest issues is a lack of realistic traffic simulation and demand modelling, aggravated by lack of or insufficient testing. One critical lesson from this debacle is that the customers’ views must be taken seriously as part of the demand modelling.

    @h. I share your concern that Snapper data is not going to show the full extent of transfers required in the new network even for very short journeys. Many people are staying away from any transfers (myself included) due to their unreliability.

  20. Matthew Gibbons, 12. October 2018, 15:52

    The low population density in Kelburn and Northland (I live in the latter suburb) and policies that favour the construction of parking spaces rather than housing, mean that the Northland bus service is not very frequent or heavily used. Even before the bus changes, when I was in the central city I usually ended up taking the Karori bus to the tunnel. I then either walked the remaining distance or sometimes transferred to a Northland bus. With limited passenger numbers, running Northland buses all the way to Courtenay Place doesn’t make a huge amount of sense to me – the focus should be on maintaining coverage of streets not covered by other bus routes.
    More suburban intensification on the Terrace, Kelburn and Northland, including eliminating car parking requirements, would be desirable and boost passenger numbers. Furthermore, having a state secondary in Karori or Northland would avoid the need for students to travel across town to high school each day.

  21. Nora, 12. October 2018, 16:37

    I live on the Wilton/Wadestown/Mairangi route of Bus 22 and in Courtenay Place I often wish I didnt have to travel to the Railway Station or The Terrace to catch this bus to Norwich Street where I am normally the last off – so why continue to Johnsonville which has a great train service.

  22. Lim Leong, 12. October 2018, 17:59

    @Matthew. I am afraid you got your facts wrong. You obviously have not seen how crowded the new Bus 22 at peak time is and how much the old Bus 22 and 23 got used even at off peak with the old network.

  23. Casey, 12. October 2018, 18:28

    Matthew: Buses are a public service for all citizens, including those who are not as fit and able as you are. A secondary school in the Karori/Northland/Wilton area has been suggested for at least the past 50 years, and isn’t going to happen any time soon.

  24. Patricia Thompson, 12. October 2018, 19:18

    Matthew, if you were to try catching the 22 you would find it is heavily overused. Not only is it the bus carrying Johnsonville students to and from the University, it is pretty much the only bus running through Kelburn and Northland to the station/Terrace/Molesworth with its very high concentration of workplaces. There is the paltry 37 which only runs a couple of services a day. The morning 22 and 37 buses are so overused that they either don’t stop by the time they get to Kelburn or passengers are squashed on in pretty appalling and unsafe conditions. My daily experience, when I can get on a 22, is that seats are still full, even when students pile off at Vic. The 22s leaving the city after 5pm are usually full to standing from one bus stop out from the station. As long time loyal bus users for environmental reasons as well as convenience, my family are now relying on Ubers quite often to get home. Given that I am pushing 60, and usually wearing a backpack with a laptop, I don’t find it easy to take the 18 or 21 options and then face the long walk to or from Salamanca. On days like this, with heavy rain and high winds, I get to work drenched if I do that. The same goes for walking to and from the cable car.

  25. Steve Matthews, 12. October 2018, 19:38

    @Matthew Gibbons. Have you actually tried to catch a 22 in Kelburn between 8 and 9am? After a few weeks of being crammed cattle-like Into extremely over-used buses, I’ve given up and now walk to work. And coming home in the evening, it’s often standing room only after one stop. I’m lucky I can walk to work – lots of people can’t and have to continue putting up with this horrible overcrowded service.

  26. Keith Flinders, 12. October 2018, 23:23

    Here is a list of the peak hour services from/to Northland/Mairangi now, and as they were pre 15 July this year. Total number of services are the same, but the destinations in the morning don’t take account of where the passengers want to go to. Evening peak services don’t start or pass through the stops they did before mid July. Dropping of the 17 sees some Karori commuters working near the railway station use the 22 in the evenings.

    Starting the 22 in Newtown would suit a lot of its users and take pressure off the 2.

    Buses running to and from Brandon Street ignore where many Karori and Northland/Mairangi users want to get to or leave from. Not all work near Brandon Street, a major point that Metlink has chosen to ignore.

  27. Michael Gibson, 13. October 2018, 9:46

    “Buses running to and from Brandon Street” also ignore the need to cut down on wasteful travel. They go straight ahead at the bottom of Bowen Street and around the block instead of turning right and taking the shortest route.

  28. Keith Flinders, 13. October 2018, 12:46

    At http://www.revoltwellington.co.nz are also the comparisons between old and new Karori peak hour bus schedules. See the “Information Vault” link at the bottom of the opening page.

    Overcrowding issues still hinge on the reduction of services to/from Courtenay Place.

    In September more peak services were added to route 2 and will remain until December. Even so the afternoon/evening peak still sees fewer services than pre mid July. What happens to these added extra services from January 2019 hasn’t been advised, yet.

  29. Andrew Bartlett, 14. October 2018, 8:14

    Keith Flinders, It certainly is strange that, after all the work to have services organised as through running etc, that the 3 and the 22 don’t link up. (Perhaps they got boxed in to their ‘bus units’?). As you suggest, linking the two and increasing services would also address long-standing concerns in Newtown regarding service capacity without dramatically increasing traffic on the golden mile.

    It would be a long route however, and we saw how well that worked on the 1…

  30. Mike Mellor, 14. October 2018, 11:35

    Andrew Bartlett: The reason behind linking services across the CBD was to reduce the number of buses along the Golden Mile, which wouldn’t apply to the linking the 3 and the 22 – they are in different units, but both are operated by NZ Bus. A linked 3/22 would be a long, circuitous route, but so is 24, which seems to run reasonably well – though GWRC is not telling us anything about performance by route, despite having all the information at its fingertips.

    Keith Flinders: re the Brandon St terminus, the much (and unfairly) maligned Jarret Walker said in his 2011 report, “Stops between Hunter and Panama could also serve as a “short turn” option for northern operators who do not, for various reasons, need to operate all the way to Courtenay Place, but they would need to be south of Panama…for this to be acceptable” (note 10, p59). It’s a shame that his proposals were replaced by a very different network that was never properly consulted on.

  31. Matthew Gibbons, 14. October 2018, 21:33

    Admittedly I invariably walk to VUW in the morning, and my experience of the Northland bus is largely confined to offpeak times at the weekend and the evening. However, passenger numbers at these times are not very high, the buses are infrequent, and I think my point about concentrating Northland buses during these hours on streets not served by other buses makes some sense. I see the peak time changes have disadvantaged some people, but running more buses to different parts of the CBD might slow down the buses.