A close look at the mechanics of the rolling hub system reveals how to minimise connectivity loss while improving operations

Last June, American Airlines unveiled a new scheduling system at its hub at Chicago O'Hare, followed six months later by the introduction of the same programme at its main Dallas-Fort Worth (DFW) hub. The system, which spreads flights out during the day instead of arranging them in the "peaked" connecting banks used previously, aims to promote higher aircraft, employee and gate utilisation without making large schedule cuts.

Industry experts immediately started debating the benefits of a depeaked schedule, with some estimating utilisation increases sufficient to remove 11 aircraft from the system at Dallas and five in Chicago, and manpower reductions on the magnitude of 4-5%. For its part, American announced intended annual cost savings of $1 billion.

It has never been easy to communicate the economic advantages of hub-and-spoke systems and how they greatly outweigh the inconvenience caused to transfer passengers. The idea that airlines were moving away from their slavish devotion to connectivity was greeted by many observers with clear enthusiasm.

What was less clear is what American had actually done, and what impact the system would have on the goals of productivity and connectivity. This analysis looks at the mechanics of what was implemented, its consequences, the basic rules and the key requirements for the shift towards a depeaked system and its potential for application outside the USA.

Connectivity issues

To begin, it is important to recognise that a volume effect exists in the connectivity of any given network and a certain number of transfer passengers will be present no matter how the schedule is designed. Even Southwest, which does not plan its schedule to maximise connectivity, has somewhere around 30-40% of its traffic connect from other flights. It is useful then to measure how many connections are generated over and above what a random distribution of the same flights throughout the day might have been expected to yield without any planning.

This measure of the quality of network planning - hits per expected hits - will give an idea as to how high a priority American's planners made the capture of transfer traffic in their newly depeaked hubs. Analysis of how this indicator developed for the carrier in Dallas and for it and rival United Airlines in Chicago, gives an idea of the changes introduced in these two hubs.

Looking at the structure of American's schedule in DFW before and after depeaking, it is clear that the change has been profound. What was one of the most pronounced hub structures in the world has lost its chief operational characteristic, its high peaks and low troughs having given way to a much more even profile.

As a consequence, the maximum number of scheduled movements per 15min have decreased from 50 to 37, allowing for vastly better operating conditions. And these better operating conditions and the reduced ground congestion have been factored into block times, now on average 10min shorter than before.

However, these gains predictably resulted in a loss of connectivity, with the hits per expected hits decreasing from 2.14 to 1.71. This result shows that, despite the reduction, a careful structuring of the schedule still allowed American to make 71% more connections than would be delivered by a random distribution.

So, how was this effect achieved? Examination of the schedule reveals that, for flights to or from destinations east of Dallas, the wave patterns are maintained, with a certain overlap, never higher than 30min. It is only on frequencies to and from cities to the west that flights take off and land in a constant-flow pattern, irrespective of potential connecting opportunities.

Naturally, connectivity suffers in the system. Some transfer markets are no longer served - their potential passenger revenue sacrificed to the cause of better operating conditions on the ground. However, the careful segmentation and selection of which markets to serve has enabled the carrier to keep much of its traffic. Comparing the origin and destination (O&D) markets served in the summer 2001 schedule with those served in winter 2003 shows that the average passenger volume on those O&Ds removed from the schedule is less than a third of the volume of those markets which have been maintained.

Then there is the case of Chicago O' Hare, the first hub where depeaking was applied. Because American shares this hub with rival United, the example is in many ways unusual. However, as liberalisation progresses in Europe and Asia, it is possible to envisage more cases like this.

The comparison of the schedules before and after the change shows that connectivity took a more severe hit than in Dallas, from 1.89 hits per expected hit before depeaking to only 1.28. The maximum number of movements per 15min decreased from 28 to 24 and, as in Dallas, the average block times were reduced by 10min - curiously, still higher than the levels United has on common routes.

Analysis of the flight directionality is again key to understanding what has been done. As in Dallas, a bank system still clearly exists, except that now the banks are no longer omni-directional, but have a pronounced directional structure. Because almost all destinations linked with Chicago are either east or west of the city, with very few north or south, it was possible to "split" the waves and overlap them, so that inbound banks and outbound banks in the same direction are occurring simultaneously, while banks in opposite directions occur immediately after the preceding bank. This continuity of operations results in more manageable ground operations.

United's reaction was to take the opposite tack, as the carrier moved quickly to take advantage of the runway and airspace slots freed up by American's depeaking exercise by publishing a timetable significantly more aggressive, connected and "peaky" than its summer 2001 version. As a result, its connectivity quotient rose from 1.77 to 2.34.

Ironically, United was able to implement this strategy in large part because of the improved operating conditions at the airport, and the average punctuality records posted by both carriers in Chicago improved significantly thanks to an environment of significantly reduced traffic and capacity.

The result

In these days of terrorist attacks, Chapter 11 and SARS, it is extremely difficult to attribute any development in the market performance of a carrier to only one factor. However, a quick look at the official market share data from 2001 to the first months of 2003 is illustrative of how American performed relative to the competition in the period.

In Dallas, where American has scant competition, it suffered a market share loss of only 1% in the relevant timeframe. In Chicago, on the other hand, existing competition coupled with United's competitive response, have contributed to American losing around four market share points.

It seems inevitable that the introduction of rolling hubs will necessitate the loss of some connectivity and therefore some degree of market share. The benefits of the improved operating environment may or may not outweigh these costs, but it is in any case critical to wisely select which O&Ds to continue serving and the manner in which to serve them.

In general, the segmentation of the transfer markets served will be geared towards the improvement of the yield mix. That is, it will have to separate those O&Ds with higher transfer volumes and better yields (A-segment in the graphic) from those with lower transfer volumes and yields (C-segment). Additionally, the flights supporting the C-segment should have a low transfer share, and should be justifiable simply on the basis of their own local traffic. In this case, it will be possible to optimise the productivity of those flights, confident that the revenues lost due to the connections cut will be more than outweighed by the operational benefits, including the productivity enhancements.

Of course, the A-segment must be defended, and the connectivity of the supporting flights must be safeguarded. Intriguingly, as the number of banks in a depeaked environment is typically lower than one would have in a fully "peaky" environment, the productivity of the aircraft on the legs supporting the A-segment will actually decrease. While this effect will be compensated for by the increased productivity coming from the C-segment, evidence suggests that the overall productivity advantages reported may have been understood to be greater than they actually are.

To depeak or not to depeak?

The crucial relevance of the directional profile of a hub also becomes clear for those looking at the viability of introducing a rolling hub. From this point of view, a high portion of inter-continental flights is less than ideal, as the airline will want to feed these typically more profitable flights omni-directionally

The geographical location of the hub also plays a key role. A hub at the centre of a number of different markets - such as Amsterdam or Frankfurt - is a less attractive candidate for depeaking than a hub at the gateway between two distinct directional markets, such as Chicago and Dallas, both of which are key east-west conduits, but which serve to connect very little traffic to the north or south. The above table summarises some of the main factors in the decision of whether or not to implement a rolling hub system.

These factors seem to indicate that straightforward replication in a European environment of the system that American installed would be highly challenging, given the high omni-directionality of the European hubs and their high share of inter-continental flights. However, the truth lies somewhere in between. The logic behind depeaking a hub can be extremely fruitful even for carriers whose hubs do not satisfy the requirements of directionality and short-/long-haul mix to the same degree that American's did. This is particularly true when segmenting and selecting those flights to be optimised for productivity, versus those that will continue to be optimised for connectivity.

Taken to its logical conclusion, it is possible to see how such logic can help steer the system towards higher connectivity in good times, and towards productivity during rainy days. To be able to make such changes without fundamentally reshaping the entire network is something that could have enormous, sustainable benefits for European carriers, too.


Source: Airline Business