Don’t stress: bus timetables aren’t being “scrapped”

Timetables are set to become more dynamic with on-demand complementary shuttle services

Yesterday, the Minister for Transport announced that Transport for NSW will trial “on-demand” public transport next year as part of their Future Transport Roadmap. A number of media outlets reported “New South Wales Government to scrap bus timetables” based on initiatives promised to “transform the mass transit network”.

As one would expect, a flurry of fury followed after the announcement. Punters complained of the inability for current buses to stay within their current timetables as it is. How are they to deliver services quality “without a timetable”?

I don’t think the media did a great job at explaining what the plans were. So, let’s break it down. There are two parts to this puzzle:

  1. Timetables are becoming more dynamic
  2. On-demand services are being introduced

The truth is, timetables aren’t going way but are becoming more flexible. There are also additional on-demand services to help make the trip to timetabled services more efficient.

Hub and spoke model

You may have heard of the hub-and-spoke model. That’s where commuters take a short service close to their homes to a major transport hub to reach their final destination.

This reduces the number of low demand, point-to-point services required to get commuters to and from their destinations whilst still maintaining flexible route options.

The challenge with the current system is that spoke services (the short hops between homes and hubs) have long routes within the suburbs to get to as many pick up points as possible. This means that it could take a long time for commuters to get from their home to the hub regardless of whether all the pick up points have passengers.

On-demand spoke services

The on-demand trial that was being mentioned is about improving commuter connections to and from transport hubs. As the Future Transport Roadmap says:

The future of personalised transport will involve customers being able to book flexible, on-demand local services to make first- and last-mile connections to and from mass transit hubs.

Page 41, Future Transport Technology Roadmap 2016

On-demand services would complement existing “spoke” bus routes with routes being optimised for booked demand.

Diagram showing how a hub and spoke model with on-demand services could work
Diagram showing how a hub and spoke model with on-demand services could work

On-demand example:

Imagine the resident living in the middle of the suburbs, around a 15 minute drive from the train station. Currently, the options may be for the resident to drive their car to the station and commute to work. However, parking spots are limited.

Catching a bus is also an option. However, the closest bus stop may be a 10 minute walk away and only runs during peak hours. Worse still, it’s a bus service which is route is long and stops at many locations within the suburb before reaching the train station.

The on-demand public transport model tries to solve this. A commuter can “book” what is effectively a shuttle service between their home and the closest train station in advance. The route and times for this on-demand service will be generated continuously based on who’s booked a service.

Dynamic timetabling in trunk routes

It’s something that Sydney Trains have been doing for years. Despite having seemingly static timetables, Sydney Trains timetables are generated at least once a day to account for things like track work, special events and “operational issues”.

Some bus routes are also brought in especially in time for special events. For example, an example I know well is the Central to Moore Park shuttle during major sporting events.

The promise made in the Future Transport roadmap is that these dynamic timetables will reach more modes of transport (including buses). These timetables will also extend in reach, modelling patterns based on weather, demands based on day of week.

Using the supply/demand insights, develop an algorithm that optimises the timetable for day-of-week, weather and planned/unplanned events

Page 88, Future Transport Technology Roadmap 2016

It’s also about being able to generate new routes and increased frequency when new demands arise. With the Opal data that Transport for NSW has on their hands, they could potentially generate new high demand, point-to-point routes to cater for new businesses opening up or when new developments are built.

Conclusion

Don’t stress: timetables aren’t going away. They remain very important for the operation of transport services.

The good news is that the timetables will likely be adjusted more frequently based on demand on each route at particular times.

On-demand services will likely complement existing spoke services to make them more efficient and convenient for commuters.

Hopefully, this will less crowded services and quicker journey times into the future.

2016 Senate Election tibits

Every election, the AEC releases what are effectively digitised versions of every single “formal” Senate ballot paper in the election.  This year, NSW’s dataset is a 1.06GB CSV file.

Since the Senate voting rules have changed, I thought I’d have a bit of fun and take a look at how well Australians (specifically NSW voters) followed these new rules.

‘Below the line’ vs. ‘Above the line’ voting

First off, let’s look at below the line vs above the line voting.  The new rules suggest voters place at least 12 boxes below the line or 6 boxes above the line.  I use the word “suggest” because incomplete votes are still “saved” under new savings provisions which came with the changed rules.

Let’s take a look at the NSW breakdown of formal Senate votes:

  • Below the line*: 185,387
  • Above the line*: 4,234,502
  • Unspecified**: 72,310

* For the purpose of this summary, ballots considered ‘below the line’ if they do not contain numbers marked ‘above the line’.  Ballots considered ‘above the line’ are the inverse. ** Unspecified ballots are ballots saved by various ‘savings provisions’ and are still considered formal votes.  For example, these could include ballots where the voter placed ‘1’ above the line and 2-6 below the line.

Voting below the line

The new voting rules made it easier for voters to vote “below the line”.  Not only did voters not have to fill in every single box on the paper, the savings provisions are more generous than ever making more incorrectly filled ballots considered “formal”.

The missing ‘1’: Right off the rank, there is a data anomaly from these generous savings provisions.  There were more below the line voters which had placed a ‘2’ in their ballot paper then voters who had placed a ‘1’ below the line.  Around 2,054 “below the line” ballots neglected the ‘1’ compared to 608 voters who neglected the ‘2’.

Voting only ‘6’ below the line: Many voters also got confused with the 6 “above the line” or 12 “below the line”.  Around 2.5% of NSW below the line ballots were only filled to the number ‘6’, rather than the ’12’ which was prescribed by the AEC^.

Overall compliance: Having said all of that, 96.7% of NSW “below the line” voters followed the instructions correctly and placed a number ’12’ below the line.

Beyond 12 below the line: However, the subsequent drop-off was huge.  Only 48,505 ballots had the number ’13’ below the line, compared with 179,196 for the number ’12’.

It was reported that many voters were advised to 12 numbers below the line, and 12 only, which could have played a role in the substantial drop-off.  However, it could also be the case that the majority of voters only want to do the “bare minimum” in terms of filling out a ballot.

Only about 26% (48,505) of below the line voters placed a ’13’ below the line. 4,332 ballots (2.34% of below the line ballots) had all 151 boxes filled in.

^ N.B. I am using the number of voters who placed a ‘6’ minus the number of voters who placed a ‘7’ anywhere below the line as a proxy for the total number of votes below the line. Further analysis is required to determine the exact number of voters who placed exactly 6 numbers below the line.
 

Voting above the line

Only 1 above the line: Despite a vocal campaign from the AEC to make voters aware of the changed Senate voting rules, around 6% of above the line voters continued to simply place a single ‘1’ above the line^. This was the previous above the line voting method.

Beyond 6 above the line: As with the below the line vote, there was a substantial drop off beyond the minimum number of boxes prescribed to be filled. 3,766,149 (3.8 million) voters placed a ‘6’ above the line compared with a mere 224,416 ballots with a ‘7’.

Compliance (BTL vs ATL): Interestingly, above the line voters were not quite as acute in following the AEC instructions with only 89% of above the line voters managing to place a ‘6’ above the line (compared with 96.7% BTL voters complying with the 12 number ‘rule’).

^ N.B. Derived from the number of ballots with a ‘1’ above the line minus the number of ballots with a ‘2’ below the line. This is merely a proxy. Further analysis is required to determine the exact number of voters who placed exactly 1 number above the line. Excludes ballots with both below and above the line numbering (i.e. unspecified)

Thoughts…

Here are a few take outs:

  • Most voters continued to vote above the line
  • A substantial number of voters continued to only place ‘1’ above the line when voting, meaning their votes were immediately exhausted after their first nominated group or candidate
  • Most voters are doing the bare minimum when voting for the Senate (i.e. only filling in 6 ATL or 12 BTL).

If I had more time, I’d analyse the exact number of numbers per ballot (rather than the pseudo counting I have going on at the moment). I’d also convert ATL votes into BTL votes, which would then allow preference flow simulations to see exactly which votes ended up making the quota for the 12 NSW Senators we elected. But alas, too much data and too little time on my hands 🙂

Widespread delays plague Multi-Technology rollout

Analysis: Some areas delayed by up to 8 months, with 290k premises delayed by at least a month

The company responsible for building the National Broadband Network has updated its rollout schedule, revealing wide ranging delays of over a month in 105 multi-technology mix (MTM) rollout areas around Australia, affecting around 290,000 premises.

These rollout areas predominantly uses the Coalition’s preferred Fibre to the Node (FTTN) technology, where the company rolls out fibre to the neighbourhood and reconnects with the existing copper to the home.  Despite promising rollouts using the FTTN technology to be faster to complete, the company had reportedly been facing issues including slow rollout design approvals from power companies who will have to power the nodes in the streets.  The reasoning behind the latest set of delays is unknown.

The areas worst affected by the delay are Cygnet in Tasmania and Mornington in Victoria, with a delay of 8 months shifting completion dates from late 2016 to mid 2017.  The is followed closely by another rollout area in Mornington, Victoria as well as South Hobart and Margate in Tasmania and Garfield in Victoria with delays of between 6 and 7 months.

The rollout in Fletcher, NSW and suburbs near Claremont, Hobart, Tasmania have been set back by around 5 months.  Another 8 rollout areas, covering around 22,400 premises in parts of Victoria, Queensland, Western Australia, Tasmania and New South Wales have been delayed by 4 months.

For a full list of affected areas, refer to the table below.

Another 22 areas, not listed below, were delayed by less than a month.

Service Area Module (SAM) Locality May completion date June completion date # of months delayed Approx number of premises affected
7CYE-01 Cygnet Oct-2016 Jun-2017 8 600
3MOR-02 Mornington 12-Aug-2016 Apr-2017 8 3400
3MOR-05 Mornington Sep-2016 Apr-2017 7 2300
7HOB-12 South Hobart, Wellington Park, Fern Tree Oct-2016 Apr-2017 6 1000
7MGT-01 Electrona, Lower Snug, Margate, Snug, Coningham Dec-2016 Jun-2017 6 2000
3GAR-01 Garfield, Longwarry, Bunyip 05-Aug-2016 Feb-2017 6 2300
2NLT-01 Fletcher Oct-2016 Mar-2017 5 3700
7NWT-02 Chigwell, Berriedale Oct-2016 Mar-2017 5 2800
3TAG-04 Traralgon Oct-2016 Feb-2017 4 2600
4IGH-01 Ingham Oct-2016 Feb-2017 4 3000
6RKH-01 Safety Bay, Rockingham Sep-2016 Jan-2017 4 3100
6RKH-04 Rockingham, Safety Bay, Cooloongup Sep-2016 Jan-2017 4 3500
6ARM-02 Cardup, Byford Nov-2016 Mar-2017 4 4700
7BUI-04 Hillcrest, Montello, Park Grove, Parklands, Burnie Jan-2017 May-2017 4 1800
7HUL-01 Huonville, Ranelagh, Franklin Jan-2017 May-2017 4 2600
2GUL-01 Gulgong 15-Jul-2016 Nov-2016 4 1100
2NRN-01 Spring Farm Nov-2016 Feb-2017 3 800
7DER-04 Don, Stony Rise, Devonport Nov-2016 Feb-2017 3 3000
4SHF-02 Trinity Beach Sep-2016 Dec-2016 3 3200
2SHH-04 Flinders, Blackbutt Jan-2017 Apr-2017 3 3300
7ETD-01 East Devonport, Ambleside Dec-2016 Mar-2017 3 2600
7DSF-01 Dodges Ferry, Forcett, Lewisham, Primrose Sands, Carlton Dec-2016 Mar-2017 3 4300
7DER-03 Miandetta, Quoiba, Spreyton, Stony Rise, Tugrah, Aberdeen Jan-2017 Apr-2017 3 2900
3TAG-06 Traralgon 19-Aug-2016 Nov-2016 2 3400
2GLB-03 Run-o-Waters, Goulburn 22-Jul-2016 Oct-2016 2 2600
4WUR-09 Minyama, Buddina 08-Jul-2016 16-Sep-2016 2 3400
3MOE-03 Newborough, Moe 29-Jul-2016 Oct-2016 2 2800
2GLB-02 Goulburn 29-Jul-2016 Oct-2016 2 3200
3CRB-06 Craigieburn 29-Jul-2016 Oct-2016 2 4400
2GLB-04 Goulburn 22-Jul-2016 23-Sep-2016 2 2900
2ALB-01 South Albury, Albury 29-Jul-2016 30-Sep-2016 2 3600
7BUI-01 Emu Heights, Havenview, South Burnie, Wivenhoe, Burnie Dec-2016 Feb-2017 2 1200
3TAG-03 Traralgon East, Traralgon Sep-2016 Nov-2016 2 3100
4EDG-08 Kanimbla, Whitfield, Brinsmead Sep-2016 Nov-2016 2 2600
4FRV-01 Norman Gardens Sep-2016 Nov-2016 2 3300
7HOB-11 Moonah, New Town, Lenah Valley Sep-2016 Nov-2016 2 3000
3WBO-01 Warrnambool Sep-2016 Nov-2016 2 3700
7HOB-10 Sandy Bay Oct-2016 Dec-2016 2 3000
2BUP-03 Doyalson, Wyee, Blue Haven Nov-2016 Jan-2017 2 3100
2NRN-06 Harrington Park Nov-2016 Jan-2017 2 1800
3RYE-01 Rye Nov-2016 Jan-2017 2 4000
4BWE-01 Bowen Nov-2016 Jan-2017 2 2200
5MIC-01 Millicent Nov-2016 Jan-2017 2 2600
3CBR-64 Coburg North,Preston (Vic.),Reservoir (Vic.) Sep-2016 Nov-2016 2 1200
2BUP-01 San Remo, Buff Point 05-Aug-2016 Oct-2016 2 3200
3COL-02 Colac East, Elliminyt, Colac 05-Aug-2016 Oct-2016 2 3900
2MAI-09 Maitland, Telarah, Lorn 15-Jul-2016 09-Sep-2016 2 3300
3WAN-01 Wangaratta 15-Jul-2016 09-Sep-2016 2 3000
2MAI-06 Bolwarra Heights, Largs, Mindaribba, Bolwarra 22-Jul-2016 16-Sep-2016 2 2200
2ALB-04 North Albury, Albury 05-Aug-2016 30-Sep-2016 2 3400
2AVA-02 Avalon 05-Aug-2016 30-Sep-2016 2 2900
2MAI-01 Windella, Rutherford 12-Aug-2016 30-Sep-2016 2 1000
3MOE-04 Yallourn North, Newborough 19-Aug-2016 Oct-2016 1 3300
2MSV-03 East Bowral, Bowral 15-Jul-2016 26-Aug-2016 1 3500
2ALB-03 West Albury, Albury 29-Jul-2016 09-Sep-2016 1 3700
3SHP-09 Kialla West, Kialla 29-Jul-2016 09-Sep-2016 1 1500
2MTT-01 Mittagong, Willow Vale, Braemar 12-Aug-2016 23-Sep-2016 1 2000
4EDG-07 Earlville, Mooroobool, Bayview Heights 26-Aug-2016 Oct-2016 1 3400
3WAN-02 Wangaratta 15-Jul-2016 19-Aug-2016 1 2900
3WGU-04 Drouin East, Drouin 22-Jul-2016 26-Aug-2016 1 3400
7DER-01 Devonport 05-Aug-2016 09-Sep-2016 1 1800
2ERN-03 Green Point, Erina 12-Aug-2016 16-Sep-2016 1 3200
2KTB-02 Leura, Medlow Bath, Katoomba 12-Aug-2016 16-Sep-2016 1 3400
3WGU-05 Drouin 19-Aug-2016 23-Sep-2016 1 2800
4NEW-06 Cranley, Gowrie Junction, Torrington, Wilsonton, Wilsonton Heights, Cotswold Hills Dec-2016 Jan-2017 1 3300
2SHH-01 Shell Cove, Shellharbour, Dunmore Jan-2017 Feb-2017 1 3000
2NRN-05 Narellan, Narellan Vale, Elderslie Oct-2016 Nov-2016 1 2900
2PKE-05 Lake Heights, Berkeley Oct-2016 Nov-2016 1 3900
5GPC-05 Pooraka, Para Hills West Oct-2016 Nov-2016 1 3700
2PKE-02 Lake Heights, Warrawong, Cringila Oct-2016 Nov-2016 1 3700
3LOR-20 Lorne Oct-2016 Nov-2016 1 2500
3RMS-01 Romsey, Lancefield Oct-2016 Nov-2016 1 2700
3SUN-01 Sunbury Oct-2016 Nov-2016 1 3300
2NRN-03 Narellan Vale, Smeaton Grange, Harrington Park Oct-2016 Nov-2016 1 3800
4BDA-20 Babinda Oct-2016 Nov-2016 1 800
4FRV-03 Koongal, Lakes Creek, Nerimbera, Frenchville Oct-2016 Nov-2016 1 4300
4SAR-01 Sarina Oct-2016 Nov-2016 1 1900
6RKH-05 Waikiki, Cooloongup Oct-2016 Nov-2016 1 3700
6RKH-03 East Rockingham, Hillman, Rockingham, Cooloongup Dec-2016 Jan-2017 1 3300
6RKH-06 Waikiki, Safety Bay Dec-2016 Jan-2017 1 3500
6RKH-08 Rockingham, Peron Dec-2016 Jan-2017 1 3900
7CLT-01 Campbell Town Jan-2017 Feb-2017 1 700
5GPC-07 Mawson Lakes Jan-2017 Feb-2017 1 3600
3WDG-01 Wodonga, Leneva Sep-2016 Oct-2016 1 2700
3WAN-03 Wangaratta, Waldara Sep-2016 Oct-2016 1 1700
4NEW-09 Drayton, Harristown, Darling Heights Sep-2016 Oct-2016 1 3100
7EXE-01 Exeter, Gravelly Beach, Lanena, Rosevears, Swan Point, Blackwall Sep-2016 Oct-2016 1 900
4BUD-05 Palmwoods Sep-2016 Oct-2016 1 2100
4ROT-01 Rockhampton City, Depot Hill Sep-2016 Oct-2016 1 1400
4RED-01 Mount Cotton, Carbrook Sep-2016 Oct-2016 1 3000
4SHF-01 Trinity Beach, Trinity Park, Kewarra Beach Sep-2016 Oct-2016 1 1900
7NOL-01 Magra, New Norfolk, Lawitta Sep-2016 Oct-2016 1 2900
5BDT-20 Bordertown Nov-2016 Dec-2016 1 1600
7BIC-01 Bicheno Nov-2016 Dec-2016 1 900
5BRR-01 Berri Nov-2016 Dec-2016 1 2500
6ARM-01 Byford, Darling Downs, Haynes, Hilbert, Mount Richon, Wungong, Brookdale Nov-2016 Dec-2016 1 4000
6ARM-04 Forrestdale, Haynes, Seville Grove, Armadale Nov-2016 Dec-2016 1 3600
3MOE-01 Trafalgar, Moe 22-Jul-2016 19-Aug-2016 1 2700
6MDR-10 Erskine 27-May-2016 24-Jun-2016 1 2500
3PTO-05 Portarlington 15-Jul-2016 12-Aug-2016 1 2200
3PTO-06 St Leonards 15-Jul-2016 12-Aug-2016 1 2800
3OCG-02 Breamlea, Barwon Heads 22-Jul-2016 19-Aug-2016 1 2800
3SHP-10 Kialla 29-Jul-2016 26-Aug-2016 1 1300
7BUI-03 East Cam, Ocean Vista, Park Grove, Parklands, Shorewell Park, Cooee Feb-2017 Mar-2017 1 2900
7ETD-02 Latrobe Feb-2017 Mar-2017 1 2000

Analysis based on NBN Co’s rollout schedule, reproduced by Telstra on the Telstra Wholesale website.  A variety of sources were checked to confirm this information.

Opal CBD Increment: a quick follow up

A few people have written in about the CBD Increment since my blog post went live this morning. As it turns out, this CBD increment is “well documented”… in a 107 page handbook known as the Sydney Trains and NSW Trains Fares and Ticketing Customer Handbook.

I’ve tried looking all afternoon, and cannot find a link from either the Opal website nor on any portion of the Transport for NSW website discussing Opal or fares generally. It is, however, linked to from the Terms and Conditions page about paper tickets.

For those curious, the direct link can be found here. The part you’re looking for is page 74.

Quick summary

Basically, any train trip that traverses through or starts and ends at a CBD station (Central, Town Hall, Wynyard, Circular Quay, Martin Place, Kings Cross, St. James and Museum Stations) will incur an extra 3.21km distance in their trip.

There is one extra caveat though. Regardless of which of the CBD stations you get off at, TfNSW will calculate the end of your trip to a “Gateway Location” based on which line you came from… before adding the extra 3.21km. Let me elaborate:

CBD Gateway Station Table

Gateway Station Travelling via
Wynyard the Sydney Harbour Bridge
Central Redfern Station, or Airport Line
Kings Cross Eastern Suburbs Line

If you’re travelling from Macquarie University to any CBD Station, you will be charged the fare distance from Macquarie University to Wynyard (the Gateway Station for via Sydney Harbour Bridge) plus the extra 3.21km increment.

If you’re travelling from North Sydney to Newtown, you also have to add the increment.  You will be charged the distance fare from North Sydney to Wynyard (the Gateway Station for via Sydney Harbour Bridge) plus the extra 3.21km increment.  In addition, you will pay for Central (the Gateway Station for via Redfern Station) to Newtown.  Note, that the increment is only charged once.

CBDIncrement

The problem

The idea behind the CBD increment is so that periodical tickets (e.g. weekly tickets) can be sold as a “city ticket” meaning passengers can get off any any of the CBD stations with the same ticket.

However, this doesn’t make sense for the Opal system where fares are advertised on a distance basis. It’s misleading and disingenuous to advertise that Opal train fares are based on “track distance” when in fact, it’s based on a psudo-distance hidden away in a 170 page handbook.

A suitable analogy, in my opinion, is a grocer selling apples and oranges at $3.99/kg. However, hidden away in a 107 page handbook, the grocer says that oranges incur an extra 1 kg increment that can be found the aisle that sells milk. Surely, this is considered misleading advertising.

Like the grocer, Opal advertises different fares based on track distance bands with no reference to this psudo-distance calculation. Like the grocer, it hides the CBD increment in a lengthy handbook stored in a part of the website that doesn’t talk about Opal fares. Does this mean that CBD Opal fares constitute as misleading advertising?

While I personally don’t mind to pay extra for travelling through the busy CBD area, I think Transport for NSW needs to be transparent about it.  Fiddling with the distance travelled certainly doesn’t look great.

Just my two cents. Keen to hear people’s thoughts.

Opal secretly adds extra distance to CBD fares

This is why your train fares may be more than you think

Update: we now know how TfNSW calculates this increment… but there are still massive problems.

If you travel to and from a CBD station using an Opal card, Transport for New South Wales (TfNSW) may have been charging you a little extra every time you tap off.

It has been a relatively well kept secret until now, but the final IPART report into public transport fares revealed and recommended the removal of a hidden feature, known as the Opal ‘CBD Increment’. The report states that:

“the ‘CBD increment’ [adds an] extra notional distance to the distance travelled for rail trips that start or finish in the CBD”

Excerpt from IPART final report (Page 13)
Excerpt from IPART final report (Page 13)

I came across this issue after finding inconsistencies with distance calculations when building my Opal calculator, a easy-to-use tool to compare current Opal fares with those set to start in September. To my surprise, after exhaustive research, I’ve been unable to find any mention of this “CBD Increment” on the Opal or TfNSW website.

Even TfNSW doesn’t know this exists…

Reaching out to TfNSW to enquire about this, they seemed just as baffled as I was. After two phone calls, no one thus far has been able to explain to me what this CBD Increment is for, or how much extra distance is being added to each CBD trip. Although, they have promised to escalate my issue and come back to me with more information (this was two weeks ago).

What I know for sure is that this increment does exist. Having tested a few trips myself for research, it appears the distance increment is quite random.

Some affected trips

Here are a number of trips to the CBD which cost more than what you would expect if the fare was based solely on track distance:

CBD Destination From Station Track dist.
(km)
Expected Fare Actual Fare
Town Hall Ashfield 9.6 $3.38 $4.20
Summer Hill 8.3
Artarmon 9.2
Tempe 8.1
Wolli Creek 8.6
Arncliffe 9.8
Turella 9.9
Meadowbank 19.5 $4.20 $4.82
Auburn 19.9
Berala 19.6
Macquarie University 19.9
Macquarie Park 18.7
Turramurra 19.7
Riverwood 18.8
Museum Tempe 7.7 $3.38 $4.20
Arncliffe 9.4
Wynyard Tempe 8.9 $3.38 $4.20
Summer Hill 9.2

Note: these are a small selection of trips selected to test the CBD increment. It is not an exhaustive list of stations which are affected. Track distances are based on track information provided by TfNSW through its Open Data exchange.  Prices listed are Adult peak fares.
Some trips with track distance within the tolerances listed also appear to be unaffected by the CBD increment. I’m unable to to discern a pattern at this point in time.

A trip from Ashfield to Town Hall (a CBD station) has a total track distance of ~9.6 kilometres — just shy of the 10 kilometre fare band which will cost $3.38 for an Adult during peak time. However, when travelling on the train between these two stations, TfNSW charges for the higher 10–20km fare band, costing $4.20.

Will this stay?

Despite IPART’s recommendation, Transport for NSW has not indicated whether or not it will retain the CBD increment when the proposed fare changes come into force in September.

I’m still awaiting a response from TfNSW about my enquiry about this existence of this ‘CBD increment’. Let’s see what they say if and when they respond… I’ll update this post when that happens.

NBN technicians

Gigabit NBN: demand is there, but NBN pricing isn’t

It’s not that there’s no demand, it’s that NBN’s pricing model makes it impossible for providers to put any gigabit services to market.

There’s a common misconception that there is “no demand” for gigabit Internet services in Australia. After all, if there’s no actual customer taking up a gigabit service on the NBN — how can there possibly be any demand?

Apparently, despite providers around the world actively and successfully selling gigabit connections to speed-hungry residential customers, Australians are so “agile” and “innovative” that they see no need for higher speed connections that our counterparts in the rest of the developed world seek for.

At least that’s what the latest spiel from The Australian may lead you to think.

Of course, the paper is absolutely correct in saying that the only NBN customers on gigabit connections are “trial” connections by service providers. But what the article neglects on is the context: why Australians are always seemingly so special compared with the rest of the world. A inhibitive pricing model.

Where did the pricing go wrong?

I’ve written extensively about the issues with the NBN Connectivity Virtual Circuit (CVC) pricing model. Effectively, providers have to buy virtual bandwidth capacity (CVC) that’s shared across all^ the users from a single retail service provider (RSP).

If a provider buys only 100Mbps of CVC: at any point in time, the sum of all traffic across all users from that RSP will be capped at 100Mbps. It doesn’t matter if the provider has 50 x 100/40Mbps connection, the maximum the provider can transfer across all users is 100Mbps.

The problem is that it is set at an artificially high price to ensure revenue to the NBN company.

This means, to even consider offering gigabit plans, a provider must have enough customers at each point of interconnect to need at least 1Gbps of CVC… which comes in at $17,500 ($17.50 per Mbps before GST).

NBN will soon be introducing dimension-based discounts which will give more generous CVC discounts to providers who buy more CVC on a per-user basis, but it still puts 1Gbps residential plans in the realm of impossibility.

Let me be clear —it’s not that there’s no demand for gigabit plans in the residential market. It is that it is too cost prohibitive and far too risky for providers to offer gigabit services to consumers due to the NBN CVC pricing model.

Under the current pricing model, which to be fair to the current Government, was brought in when the NBN was first introduced by Labor… the NBN rollout must reach critical mass at each point of interconnect and have a significantly reduced cost per Mbps for CVC in order for NBN plans with speeds higher than 100Mbps to start showing up.

Providers overseas, for example AT&T, are offering gigabit services at USD$70 which converts to around AUD$100. With the current CVC model at an arguably high 1:100 contention ration (i.e. 10Mbps of CVC allocated for a 1Gbps connection), NBN costs alone would be over $300. That doesn’t even include interconnection, peering or backhaul costs.

I’ve argued in the past for a different style of dimension-based CVC pricing, where a proportional CVC contribution is paid per end user to make up for the revenue. In turn, NBN will manage the contention.

This would allow providers to offer 1Gbps plans to consumers before the rollout reaches critical mass, and increase the short term revenue for the project if there are more users who take up higher speeds. But such a radical change can’t happen without substantial industry consultation.

Regardless, it is a self-fulling prophecy to say that “there is no demand for gigabit services” when in fact the NBN pricing model is designed to prohibit gigabit services for residential connections.

^ effectively all customers serviced from the same NBN point of interconnect. There are limitations to how many users can be placed on a single CVC and which technologies can share a common CVC, but for all intents and purposes — this is irrelevant for this discussion.

Where’s the demand?

Even if the pricing issue is tacked… you might ask — why would there be demand for gigabit services? QHD content on Netflix only utilises 25Mbps of bandwidth; what applications possibly require more than that? While I can’t possibly accurately predict future applications, there is one obvious demand for high speed: productivity.

As “cloud” storage and computing becomes more and more widely used, there is increased need for burstable Internet connections. As a budding geospatial geek, say I need to transfer a large dataset from my computer to a cloud processing service (conceptually, backing up a large number of photos or videos to cloud storage would be an identical situation).

It would take me no less than 12 hours to transfer a 500GB raster dataset over a 100Mbps connection to my remote server. With a 1Gbps connection, this would reduce to just over an hour.

In this example — I don’t necessarily need the 1Gbps at all time. Perhaps on average, I need no more than a 25Mbps connection. But the productivity gains in having a burstable connection that can reach gigabit speeds when I need it to can be enormous. Rather than sitting and twiddling my thumbs for half a day to wait for my data to be uploaded, and another 12 hours to download it again, I can save some 20-22 hours all up simply by having the connection “on standby” (so to speak).

Infrastructure networks are rarely designed on an “average consumption” basis. If roads and highways, or water and power infrastructure were designed on what happens on “average”, traffic congestion, blackouts and water shortage would be a common problem during peak times.

Good power infrastructure, for example, is designed to cope with sudden demands for electricity. There is phenomena known as “TV Pickup” in the United Kingdom where there is a sudden increase in electricity demand during TV ad breaks — thanks to a large number of people simultaneously turning on the kettle across the UK.

Final words

So, this isn’t simple demand and supply here. Demand is not only driven by users… it is also a function of the NBN’s price structure.

The fact of the matter is, there is far more to the lack of gigabit services than meets the eye initially.

If we are indeed serious about the innovation nation, I call on both side of politics to put down their swords and look seriously at pricing reform for the CVC.  I seriously hope for some bipartisanship… if not at the technology front, then certainly at the pricing policy front.

Detailed analysis: How will Qantas’ on-board Wi-Fi impact NBN satellites?

Geospatial analysis of the daily Qantas flight paths and NBN satellite beam coverage shows how NBN’s satellite network could be affected.

Writer’s note: Qantas is a customer of ViaSat, not NBN Co. ViaSat intends to trial on-board internet using NBN Co’s satellites — however, they have indicated they intend to launch their own satellites (ViaSat-3) to deliver a long-term solution for on-board Wi-Fi globally. This Qantas-ViaSat-NBN Co deal is dependent on industry consultation being completed in June 2016.

Qantas has recently announced that it plans to offer a Wi-Fi service on board its A330 and Boeing 737 fleet from early 2017 by utilising capacity on NBN’s recently launched Sky Muster satellite.  However, many Australians living in rural and regional Australia have raised concerns that the Qantas service will cause further congestion on an already limited service.

The company responsible for building the National Broadband Network, nbn, had completed a review of satellite capacity in its Fixed Wireless and Satellite review where it identified 31 beams around Australia that would be oversubscribed or severely oversubscribed once a take-up rate of 65% is reached.  As a result, the company will introduce strict Fair Use policies to significantly limit the amount of data to 150 GB (priced at a premium) that can be used by each customer on a 4-week rolling average basis.

Consequently, the Qantas announcement has infuriated many who are within the satellite footprint.  Analysis of flight paths taken by Qantas’ domestic flights using their A330-200, A330-300 and Boeing 737-800 fleets show that all planes fly through at least one congested or severely congested satellite beam.

Key findings:

  • 334 Qantas domestic flights utilise the A330-200, A330-300 and Boeing 737-800 fleet on the day analysed.
  • Every flight flew under at least one oversubscribed or severely oversubscribed beam.
  • 58% of oversubscribed or severely oversubscribed beams will have minimal impact, with at most 2 planes flying under the beam at any given time.
  • The Sydney-Brisbane and Sydney-Gold Coast routes travel entirely within severely oversubscribed NBN Satellite beams (34, 29, 25, 20).
  • The Sydney-Melbourne, Melbourne-Canberra, Sydney-Canberra and Townsville-Brisbane routes travel entirely within oversubscribed OR severely oversubscribed NBN Satellite beams.

Analysis: by congested beams

Over half (58%) of congested beams are not affected or are minimally affected by Qantas planes.  Of the 31 beams considered oversubscribed or severely oversubscribed, 5 of them do not cover any current eligible Qantas flight paths.  7 beams will have at least one flight under the path at some point, and 6 beams will have up to 2 flights under it at the same time (within a 10 minute time frame).

Unsurprisingly however, the beams covering areas immediately surrounding capital cities will have the greatest number of flights under it at any given time:

  • Beam 42 (Sydney Beam) is the worst affected, with up to 7 Qantas flights within a 10 minute time frame travelling under it.
  • Beams 47 (Melbourne Beam) and 20 (Brisbane Beam) come in second, with up to 6 Qantas flights within a 10 minute time frame travelling directly under each of them.
  • Beam 37 (Adelaide Beam) comes in third, with up to 5 Qantas flights within a 10 minute time frame travelling directly under each of them.

It should be noted though, that for the day analysed — the Sydney and Brisbane beams only had the greatest number of flights under it for a single 10 minute time frame (between 8:50am – 9:00am and 7:30pm – 7:40pm respectively).

Finally, by considering the average number of Qantas planes under a beam over a 24 hour period — we see that Beam 47 (Melbourne) tops out at 1.94 planes with Beam 42 (Sydney), Beam 20 (Brisbane) and Beam 41 (Canberra) following closely behind at 1.52, 1.45 and 1.22 planes respectively.

Beam* Max concurrent
flights under beam
(over 10 min period)
%age of time with
at max plane
under beam
%age of time with
at least one plane
under beam
Avg. number of
planes under beam
over 24 hrs
47 (Melbourne) 6 2% 67% 1.94
42 (Sydney) 7 1% 63% 1.52
20 (Brisbane) 6 1% 63% 1.45
41 (Canberra) 4 5% 58% 1.22
37 (Adelaide) 5 1% 49% 0.86
24 (Armidale) 5 1% 53% 0.83
48 (Omeo) 4 1% 51% 0.76
66 (Perth) 3 1% 51% 0.63
34 (Newcastle) 3 3% 40% 0.57
33 (Bathurst) 4 1% 33% 0.48
25 (Grafton) 2 3% 33% 0.37
46 (Ararat) 3 1% 28% 0.36
3 (Townsville) 2 3% 28% 0.31
15 (Wondai) 3 1% 24% 0.30
29 (Port Macquarie) 2 1% 26% 0.27
36 (Yorketown) 3 1% 22% 0.26
35 (Port Lincoln) 2 2% 22% 0.24
7 (Mackay) 2 3% 19% 0.22
11 (Rockhampton) 2 2% 20% 0.22
31 (Whyalla) 1 15% 15% 0.15
44 (Kangaroo Island) 1 8% 8% 0.08
63 (Guilderton) 1 8% 8% 0.08
16 (Maryborough) 1 6% 6% 0.06
2 (Charters Towers) 1 3% 3% 0.03
1 (Cairns) 1 3% 3% 0.03
51 (Geelong) 1 2% 2% 0.02
49 (Bega) 0 0% 0% 0.00
52 (Inverloch) 0 0% 0% 0.00
54 (Burnie) 0 0% 0% 0.00
56 (Hobart) 0 0% 0% 0.00
68 (Bridgetown) 0 0% 0% 0.00

* Beam name is based on a suburb/town/city directly under the beam and may not be the official name used by nbn

Analysis: by flight

Each flight and its flight path were analysed to see which NBN Satellite beam it flies under. The results shows that every single Domestic Qantas A330 and B737 flight flies under at least one oversubscribed or severely oversubscribed satellite beam.

The most prominent are the Sydney-Brisbane and Sydney-Gold Coast routes, which flies entirely within severely oversubscribed beams (that’s beams 34, 29, 25, 20).  Sydney-Melbourne, Melbourne-Canberra, Sydney-Canberra and Townsville-Brisbane routes fly entirely through oversubscribed or severely oversubscribed beams.

For the full analysis output, click here!

Concluding thoughts…

I find the results of this analysis somewhat inconclusive. Firstly, unlike domestic US services — the number of Qantas flights expected to get the Wi-Fi service is quite small.

With under 350 flights spread out geographically and over a 24 hour period, I doubt the planes would have a material effect on congestion. Currently, the worst case scenario seems to be up 6-7 planes flying under a single beam at any given time. However, in the case of the 7-plane statistic, it happens only once in a 24 hour period. The speed at which planes travel also mean that they will typically fly in and out of narrow beams within 10-15 minutes, meaning any impact should be distributed across multiple beams as the plane flies through the airspace.

On the other hand, the bulk of the flights will fly under already oversubscribed areas. This is especially true for the beams serving the areas immediately surrounding the capital cities, which are all severely oversubscribed (bar-Darwin). These areas also have the greatest number of concurrent flights, represented by the “average number of planes over 24 hours” statistic.

So, no. I don’t think there’s an immediate threat to congestion. However, it does set a precedent. If more carriers get on board… and if international flights get added to the pool as well — things could well change in the future.

NBN Long Term Satellite beams
Diagram showing NBN Co satellite beams and risk of congestion as determined in the Fixed Wireless/Satellite Strategic Review (FWSat SR).

Assumptions made in this analysis:

  • The flight data analysed was from Wednesday, 23rd Feb 2016.
  • Qantas will only install the Satellite-powered Wi-Fi solution on their A330-200, A330-300 and Boeing 737-800 fleet.
  • In all cases where the plane transverses an area with both a wide and narrow beam, the congested, narrow beam is selected.
  • Plane locations are calculated in 10 minute intervals.
  • Congestion (oversubscribed beams) are based on results in the NBN Fixed Wireless and Satellite Review.
nbn™ logo (large)

Analysis: Is privatising the Multi-Technology Mix the best way forward?

With almost no competition to incentivise an upgrade to its network, can NBN’s monopoly Multi-Technology Mix keep up with user demands or will it become Telstra 2.0?

Infrastructure Australia released it’s “Australian Infrastructure Plan” report today. It’s contents are not exclusively about the National Broadband Network, however, it made a number of recommendations to the Government suggesting that the NBN should be privatised in the medium term.

A bit of background: the fibre monopoly

Since the conception of the National Broadband Network, there had always been a plan and provision to sell company once the rollout is complete. With the original “full fibre” plan envisaged by the then-Labor Government, selling or not selling the fibre network at completion wouldn’t materially affect competition.

Fibre at the premises level is a natural infrastructure monopoly. Much like how water supply companies do not build competing sets of pipes for competition, fibre is effectively a pipe with limitless upgrade opportunities by simply “swapping out the equipment at each end”. It is vastly inefficient to duplicate networks for the sake of competition, and the Optus and Telstra HFC “war” in the 90s was an excellent example of how broadband infrastructure competition can fail.

There is a case for small amounts of network duplication in inner metropolitan areas. However, it becomes hard to compete (at least within the fixed-line space) with a network that’s already built to the scale that NBN would have.

Selling, or even long term (99 year) leases of the fibre network based on geographical areas would have made sense. Like water companies, telco infrastructure entities wouldn’t spend too much time duplicating each other’s network. Upgrading networks to user demands would incur minimal capital expenditure, as it’s a matter of swapping equipment at two ends. Competition will continue to exist, but only in the retail/service delivery component of the network by Internet Service Providers.

The Multi-Technology Mix

But things changed. If you haven’t noticed, the NBN is no longer rolling out fibre to the majority of premises. The Coalition Government’s policy of a Multi-Technology Mix fixed-line network creates new challenges policy makers in the future will need to deliberate carefully before pressing sell in parliament.

On the upside, a Multi-Technology Mix is an effective way to drive infrastructure competition. However, it doesn’t work when it’s being built by a single entity — especially not a Government-owned enterprise.

No incentive to upgrade, unless split up and privatised

As a monopoly, investing large sums of money on upgrading technologies like VDSL2 over copper where upgrade paths is not as simple as “swapping out equipment” presents an ongoing burden for the NBN company. As user bandwidth demand grows, many telecommunications companies would upgrade and extend their networks to meet user demand. Or else, they would risk a competitor coming in to take over their market share. That’s what happens in areas with healthy infrastructure competition — such as Hong Kong.

But like Telstra before it, NBN wouldn’t have a business case to upgrade its networks — even if user demands skyrocket. The company would stand to lose money (at least in the short-term) by investing to upgrade its network, while it would keep a steady revenue if it just maintained the network as it is and not upgrading at all! Civil costs to push fibre out further into the field to meet user demand isn’t cheap, especially when it’s done in successive truck rolls (which is what the current policy is).

We can say that the current NBN policy focuses on short-term policy objectives — lowering the short-term capital expenditure costs, while risking medium to long-term operation costs — in other words, leaving it for policymakers of the future to deal with.

Privatising the Multi-Technology Mix by its technologies would make sense — provided the necessary regulatory instruments were put in place. Consider privatisation by splitting the NBN into a FTTx, HFC and Wireless/Terrestrial entities. These three entities could lend themselves to infrastructure competition in areas where there is sufficient network overlap. In metro areas for example, failure to upgrade the copper-based VDSL2 solution used by the FTTx entity would see the threat of an adjacent HFC footprint grow to meet user demand (and vice versa).

In regional and rural areas, the viability of infrastructure competition is not immediately obvious. Improvements in fixed wireless technologies could threaten the market share of the FTTx entity, but the fixed wireless network would need to grow substantially to be of real competition. In addition, wireless and terrestrial (satellite) solutions tend to be more expensive. So is it possible to create healthy competition, or would those areas have to rely on Government subsidies to promote upgrades?

Even if competitors are identified, each of these entity would need to have equal opportunity to grow its footprint into their competitor’s existing footprint. Having equal access to things like duct and pipe, Tier 1 and Tier 2 facilities, power and potentially many more aspects is crucial before any privatisation goes ahead.

Then there’s the issue of vertical integration

Looking at the current telco market in Australia, obvious candidates with enough capital for purchasing parts of a privatised NBN would be Telstra, Optus, the TPG Group and the M2 Group. Companies like Nextgen or Vodafone could still take a share.

But all these companies I’ve listed, bar Nextgen, already hold a retail front: Telstra, Optus, TPG and Vodafone are obvious by name. M2 owns Dodo and iPrimus.

Currently, the unique nature of NBN is that it is entirely wholesale only with no vertical integration, driving substantial retail competition in the service provider front. In my opinion, this needs to stay as is. Infrastructure companies should stay structurally separated from its retail front to ensure innovation within the “service” sector and ensures equal pricing regardless of who owns the network.


Conclusion

In a privatised world without a ubiquitous network of natural monopoly (like fibre), the only way networks will meet user demand is by infrastructure competition. Infrastructure competition requires network duplication (or at least the threat of) to stimulate investment to upgrade. So the question is: do we want this duplication? Isn’t it blindingly inefficient to have networks that do the same thing being built two or three times over just to keep with user demand?

I think we’ve put ourselves in another telecommunications policy nightmare in Australia. Time and time again, successive Governments — in their short sightedness — have failed to realise the consequences of their policy decisions. The NBN was meant to solve the issue of a lack of infrastructure investment that resulted from the privatisation of Telstra. It still could, but potentially at the cost of inefficient network duplication.

Yes, I’d still argue that a Fibre to the Premises model would have been the ideal model. It would have made privatisation easier, it would have been far more elegant. But what’s done is done.

Some would argue it’s a good thing. Infrastructure competition could work well for companies who have substantial capital while also owning large amounts of existing network assets.

Currently, TPG is the disrupter in the market. The rollout of inner-city Fibre to the Basement network is a good example of what healthy infrastructure competition could look like.

So yes, in conclusion, if done correctly — splitting and selling the NBN is probably the best way forward for Australia. But as Malcolm would say, “it’s not the way we would have done it”.

What do you think?

NBN Fibre to the Node Trial at Umina Beach

nbn’s right: users probably not affected by node congestion

nbn has strongly refuted suggestions that their node’s backhaul link is already hitting capacity, and figures seem to match their assertion.

(analysis) The company responsible for building the National Broadband Network, nbn, has refuted claims that backhaul congestion is the primary cause of slow speeds experienced by users on its recently-launched Fibre to the Node network. As a matter of distinction, this is separate from the blog post I wrote 7 months ago about long-term capacity challenges faced by nbn.  In this post, we are discussing congestion issues faced by current customers.

If we examine the figures closely, there simply isn’t enough customers on each node (yet) to need to worry about backhaul congestion yet.  In raw numbers, the lucky chaps in Belmont North connected to node 10 in Belmont 7 (2BLT-07-10) had just a smidgen under 100 premises connected at the start of this month.  This is followed closely by node 5 in Belmont 5 and node 2 in Belmont 1 in the raw take-up of high 80s.

Node (ADA) ID Node premises count Currently active premises
2BLT-07-10 233 93
2BLT-05-05 247 88
2BLT-01-02 226 84
2GRK-02-11 273 83
2BLT-10-08 246 82
2BLT-07-07 210 81
2BLT-07-17 233 80

If we consider the FTTN AVC profile revealed in estimates (13% on 12/1 Mbps, 50% on 25/5 Mbps, 25% on 50/20 Mbps and 12% on 100/40 Mbps), we can consider an aggregate average downlink AVC of roughly 39 Mbps per FTTN user.  At roughly 100 per node sharing 1GE uplink fibre, even during full saturation — each user would get on average an uncontended 10 Mbps link.  That’s better than a 1:4 contention ratio — well above most, if not all, residential-grade services.

Of course, I’m not privy to NBN’s network utilisation graphs.  But I’m fairly confident that they will show in each node, there is currently plenty of buffer space.

As I’ve said in an earlier blog post, node-based congestion can really only be expected during peak times on a fully loaded node.  Without wider saturation of 4K TVs at this point in time, the entire neighborhood of 100 premises must stream at least 2 HD Netflix or YouTube video simultaneously before there starts to be congestion issues — a virtually impossible scenario.

We don’t know what may happen in the future. Perhaps when 4K becomes more mainstream, it will become an issue.  The scenario I posed was 25% of customers simultaneously streaming 4K is all it would take to congest a full loaded node.  But at least for now, it’s pretty safe to say that this particular factor is not major concern.

So if it’s not node backhaul, what is it?

It must suck to be in this situation currently.  When you have end users posting speed tests well below 10 Mbps download “consistently”, you know there is an issue somewhere. Whether it’s a sync speed issue, CVC underprovisioning, packets being dropped by nbn somewhere within the network or something else altogether — testimony given by CEO Bill Morrow at the recent Senate Estimates suggests that nbn and service providers are working through it bit by bit to diagnose.  One thing’s for sure, it’s most probably nothing to do with node backhaul congestion yet.

Frequent travellers penalised under proposed Opal fare changes

Commuters travelling more than 10 journeys per week will pay on average $255 more per year.

In a classic pre-Christmas news dump, the Independent Pricing and Regulatory Tribunal (IPART) released its draft report into the review of public transport fares in Sydney and surrounds.

The review’s aims were clear — to remove the penalty commuters currently endure when switching between different modes of transport.  In doing so, the revenue will decline and to plug this revenue hole — a raft of changes about fare caps and rewards have been introduced.

After reading the “thrilling” 106 page report, what becomes immediately apparent after reading the report is that it’s not easy for your average Joe to compare fares and see how it may affect them.

So, knowing me, you’d probably guess that I’d build some fandangled app to do it… and I did: opalcompared.com.

The rest of this blog post will be split into two main sections, for different audiences:

  • a findings (based on some 12 thousand calculations done by visitors) section
  • a technology section (on how the app was built)

Findings

During the short time since the launch of Opal Compared, it had accumulated over 12,000 weekly journey calculations.  Through this, a few interesting trends had started to emerge:

(A small note: the statistics are based on a snapshot of around 12,000 Adult Opal fare calculations made on Opal Compared up till about 27th December 2015)

Travellers with over 10 journeys per week will have the highest fare increase

Probably summed up perfectly in this chart below, the more journeys you take on a weekly basis — the higher the average fare increase.  The less journeys you take, the more you save.


Source: Opal Compared (opalcompared.com)

The point where the average crosses over is at exactly the 10 journey mark.  Commuters who travel more than 10 journeys per week will on average pay $4.90 more per week (or $254.80 per year — if you budget on an annual basis).  Those who travel less will likely pocket a healthy discount of $3.49 per week on average.

This baffles me.  The proposed Opal fares seem counter-intuitive since the proposed fare changes will disincentivise people from using public transport.

It simply doesn’t make sense to reward those commuters who contribute the least to revenue.  Shouldn’t IPART be looking on setting fare structures that reward those commuters who travel the most, encouraging more people to use more public transport thus increasing revenue?

(more…)