Emirates and Dubai: the anatomy of a global mega hub
In our first anniversary post in September we committed to diversifying our analyses to include more non-Qantas and global topics. This is one of our first in-depth forays into this and takes a deep dive into the operational strategy behind global connecting hubs. We’re starting with Emirates and their Dubai hub, and will follow-up with others in the coming months.
A common misconception is that EK are singularly focussed on connecting passengers between Australia and Europe. While Europe is their most important market for Australia, hubs need diverse network to be successful. End-of-the-line carriers like QF are different, utilise hub/s to distribute a traffic to a narrow range of domestic or regional destinations.
As our recent “chart of the week” post showed, approximately 90% of Australia-UAE passengers are net transit traffic. This is a huge number, but where are all these passengers coming from and going to, and just how does it work?
If not Europe, then where?
The UK is one of the most prominent markets for EK. As another recent chart of the week showed, more than two-thirds of Australia-UK traffic flew via transit points. EK is the market leader between Australia and UK, carrying 23% of the traffic according to a recent ACCC submission (12 months to May 2024). EK is aided by the diversity of their network, connecting 5 Australian cities (9 daily flights) with 8 airports in 6 UK cities (operate to Heathrow, Gatwick and Stansted in London; 19 daily flights), providing 30 one-stop city pairs (40 one-stop airport pairs).
A rough estimate is that the UK accounts for about 25% of EK’s passengers to/from Australia. So where are the other 75% going? EK hold a 34% market share to/from the rest of Europe, 56% to/from the Middle East and 27% to/from Africa.
How did we estimate this?
ABS data shows that 1.2 million passengers arrived in Australia from the UK in 2023. Applying a market share of 23% means that approximately 276,000 flew on Emirates. BITRE indicates that Emirates carried 1.1 million passengers on their flights into Australia (from UAE). 276,000 of 1.1 million is 25%.
So just how do connecting hubs work?
The Middle East’s geography generates a comparative advantage. A much claimed statistic is that two-thirds of the world’s population live within a 4 hour flight of DXB. More importantly, DXB lies at a crossroads connecting Europe, Asia and Africa with minimal backtracking. Geography is a necessary condition, but it’s not sufficient.
As a generalisation, hubs are either banked or rolling hubs. Neither is better, rather they have different attributes, advantages and disadvantages. In practice, strategies aren’t as discrete as the normative descriptions, with some operating as hybrids.
Banked: Flights are timed to arrive in a cluster (arrival banks) followed by a cluster of departures (departure bank) utilising the same aircraft. Banks undermine aircraft utilisation as flights are timed to connect to/from banks rather than maximising utilisation. They’re optimised to fit as many flights into as short a time as possible to minimise connecting times between flights, generating a trade-off:
Increasing connecting times and/or broadening the bank (longer time between first arriving and last departing flight) increases the number of one-stop city pairs and reduces risk, but makes the network less competitive.
Reducing connecting times and/or narrowing the bank requires more infrastructure (that are under utilised between banks) and increasing congestion.
Rolling: Flights are spread throughout the day, scheduled to maximise aircraft and airport utilisation. They’re cheaper to operate but results in unpredictable and long connection times, and fewer one-stop city pairs particularly on low frequency routes.
Emirates’s Dubai hub
EK operate a banked hub at DXB with 4 arrival and 4 departure banks. The banks are not uniform with the breadth and overlap, and number and distribution of flights varying. Histograms show the frequency and distribution of EK’s DXB arrivals and departures to/from in 15 minute blocks, highlighting the discrete banks. A few flights fall outside of a bank. Reasons for this include routes with high volumes of O&D traffic or flights with unique slot or curfew challenges that dictate timing. We’re also able to generalise the banks and indicative connections:
A1: 3:00am to 8:00am → D2: 6:00am to 11:00am
A2: 11:00am to 2:30pm → D3: 1:30pm to 5:00pm
A3: 6:30pm to 8:00pm → D4: 8:30pm to 11:00pm
A4: 9:00pm to 1:00am → D2: 1:00am to 4:30am
Banks are affected by unique operational idiosyncrasies. For example, EK stipulates a 75 minute minimum connecting time between flights that broadens banks compared to short haul hubs. This is affected by everything from security requirements to airport design, and aircraft turnarounds. Long haul hubs require more time to turn aircraft compared to short long haul hubs due to the higher proportion of widebody aircraft that take longer to turn than narrowbodies. This is even more severe for EK due to the high proportion of A380s that take even longer than smaller widebodies.
How does this affect routes to/from Australia?
Let’s look at EK’s Australian flights (non-stop only) to study the design of the banks, firstly the arrival banks: there are 9 daily arrivals timed into 2 banks. All 5 cities connect into A1, while only cities with multiple daily flights connect into A2 (BNE MEL SYD). SYD has 3 daily flights, however 2 connect into the same bank (A1).
A1:
EK417 SYD DXB 21:10 4:30
EK435 BNE DXB 21:00 5:10
EK413 SYD DXB 21:45 5:15
EK407 MEL DXB 22:15 5:15
EK441 ADL DXB 20:40 5:20
EK421 PER DXB 22:20 5:20
A2:
EK431 BNE DXB 1:55 10:05
EK409 MEL DXB 6:15 13:05
EK415 SYD DXB 6:00 13:20Departures from DXB to Australia follow a similar pattern. D1 gets flight to all destinations and D2 gets the 2nd daily flights to BNE, MEL and SYD. Unlike arrivals, SYD’s 3rd flight is aligned with a 3rd departure bank (D4).
D1:
EK414 DXB SYD 1:50 22:50
EK440 DXB ADL 2:00 20:50
EK430 DXB BNE 2:30 22:20
EK420 DXB PER 2:45 17:35
EK408 DXB MEL 3:00 23:20
D2:
EK406 DXB MEL 10:05 6:30
EK412 DXB SYD 10:15 7:05
EK434 DXB BNE 10:35 6:25
D4:
EK416 DXB SYD 21:30 18:20It’s evident how banking undermines aircraft utilisation. Take EK414 as an example: it arrives in SYD at 10:50pm (right before the curfew begins at 11pm) and overnights before returning to DXB at 6:00am as EK415, spending more than 7 hours on the ground. The same occurs with EK412 which spends 14 hours on the ground before returning as EK417. At first glance this must be a result of SYD’s curfew, right? When looking at MEL’s schedule it’s evident that it’s due to banking (MEL doesn’t have a curfew)! EK408 arrives in MEL at 11:20pm before departing at 6:15am as EK409, also spending about 7 hours on the ground. The same then occurs with EK406 which also spends 14 hours on the ground before returning as EK407.
ADL and PER are moot as they only have a single daily flight and the shorter flight times allow immediate turnarounds. BNE is peculiar: EK430 arrives in BNE at 10:20pm, turning relatively quickly in less than 4 hours, departing at 1:55am as EK431. This generates a trade-off as it arrives in DXB at 10:05am, nearly an hour before the start of the A2 bank at 11am. A2 connects to D1 (1:30pm to 5pm) resulting in connections as long as 7 hours. Comparatively, MEL and SYD’s longest connection gap is 4 hours.
Not all banks are created equally with the number and distribution of destinations varying. Most flights from Australia arrive into A1 connecting onto D2 that’s primarily focused on Europe (37 destinations), Africa (21), Americas (11) and Middle East (10), with limited connectivity to South Asia (4). The 2nd/3rd daily flights from BNE, MEL and SYD arrive into A2 connecting onto D3, providing the 2nd most destinations in Europe (22), same number in Middle East (10), marginally more to South Asia (7), but none to Africa and Americas.
There are no Australian flights into A3 and A4, however it’s evident why this is the case. A3 connects to D4 and provides no options to Europe, Africa and Americas, and limited to Middle East (4). It’s primarily focused on South Asia (12). A4 connect to D1 and provides ample connections to Europe (16) and Americas (8), but relatively weaker connectivity to Africa (5) and Middle East (7). It’s focus is South Asia (13). A4 would be viable and practical as a 3rd bank (would need to depart MEL/SYD between 2-6pm to connect to A4).
Let’s also analyse the countervailing flights to Australia. D1 and D2 are the strongest departure banks to Australia, with high concentrations of connections from Europe (37 and 23), Africa (12 and 14), Middle East (8 and 9) and South Asia (8 and 14). Arrivals from the Americas are fragmented compared to departure banks, making connections to Australia more challenging with only 6 and 4 destinations in D1 and D2.
The 3rd daily SYD flight connects from A3 to D4 and appears to provide limited connections from Europe (6), Middle East (8) and South Asia, and none from Africa, but it’s the strongest bank from the Americas (8). It doesn’t appear viable, however the European destinations are important ones (Amsterdam, Frankfurt, London Gatwick and Heathrow, Manchester and Paris)! D3 wouldn’t provide any advantage over D4, so it’s D4 or double up in D1 or D2.
What about flydubai?
Flydubai is another DXB based airline flying B737s to a range of destinations in Africa, Asia, Europe, and Middle East. Many mistakenly assume it’s an EK subsidiary. While they’re both owned by the government they’re operationally independent. EK codeshare on some flydudai flights, however we didn’t include these in the analysis since flydudai follow a rolling hub model! Histograms of the frequency and distribution of their DXB arrivals and departures shows this. Some flydudai flights provide useful additional network scope for EK, however these are limited as a result of the rolling hub model, and are likely opportunistic or highly targeted, but not systematic.
Conclusion
Emirates operate a banked hub in Dubai with 4 arrival and 4 departure banks of varying breadth and overlap, and number and distribution of flights. Each bank has varying regional concentrations and schedules are strategically timed. A rolling hub would result in a random distribution of flights, limiting the efficiency of the hub.
The Australian case study highlighted the challenges and trade-offs of the banked hub, specifically how it undermines aircraft utilisation to optimally time flights to/from banks. Alternatively, when aircraft utilisation is increased, the hub becomes less efficient and connecting times cumbersome. It also showed the importance of frequency, with Australian flights concentrated on 2 arrival and 2 departure banks, optimising connections to/from the focus markets of Europe, Africa, and the Middle East. Some connections to/from South Asia and the Americas are possible but not prioritised, nor are connections to/from the rest of Asia.
Each destination and region is unique, requiring unique analysis. Flights between other regions follow similar strategic considerations but result in a different outcome. Some options may generate a focus on only single connecting bank, while others may match more banks. Thus, Australia-Europe’s interactions with the banks aren’t generalisable to other regions although the general principle holds.
Each airline will design their hubs differently. For example, Qatar follows a similar business model and Doha is a comparable hub. However, Qatar’s strategy is different with a more diverse fleet including a wide range of smaller aircraft that allow them to serve more destinations than Emirates. This has significant implications with Qatar operating a smaller number of connecting banks. In the coming weeks we’ll conduct a similar analysis of Qatar and examine how their hub differs!
Thanks for reading and remember to subscribe and share!