Measuring rail reliability

Tuesday's report in The Straits Times on how Singapore's rail system fared last year must have left many readers and commuters somewhat flummoxed.

The article reported on two points: First, that the number of major MRT breakdowns had risen from 2014. Second, that the average distance clocked between overall breakdowns had increased.

The latter suggests longer intervals between breakdowns and thus a lower frequency of overall glitches.

Theoretically, it is not incorrect to interpret this as an improvement in reliability.

This was precisely what the Land Transport Authority (LTA) led off with in its latest statistics on rail reliability. The data on major breakdowns was right at the bottom of its three-page report.

But to impute reliability with the distance measure is not completely correct either. Intuitively, it is hard for anyone to reconcile an improvement in reliability when major breakdowns have spiked.

To be precise, there were 14 major MRT disruptions last year - up from 10 in 2014. As for the LRT system, the figure rose from four to 15. Hence, in total, the number of major rail breakdowns more than doubled to 29.

This was in spite of all the efforts and hundreds of millions of dollars expended after 2011, when two massive breakdowns resulted in a high-profile public inquiry and a renewal of SMRT's top management.

Among the 14 breakdowns last year was an unprecedented one in July, when both the North-South and East-West lines became inoperable during the evening peak period.

The incident affected nearly half a million commuters.

Thus, average distance between disruptions is at best a rough barometer, and it should not be the ultimate measure of reliability.

This is because disruptions are not uniform. Some affect a few hundred people for a few minutes, others affect hundreds of thousands over several hours.

In theory, you could have 300,000km between incidents - which is Hong Kong's gold standard - or even 500,000km, but that does not necessarily mean you have a reliable system.

You can achieve those numbers by eliminating all the smaller faults, so that the trains run longer before any breakdown.

It could gain you a few thousand more kilometres travelled before the next breakdown.

So trains could travel 300,000km or more before a breakdown.

But the system might still have 20 to 30 major breakdowns a year.

Would that be a reliable system?

You can argue that it is, if you are looking at it purely from an engineering angle, which was what SMRT, LTA and the Ministry of Transport seem to suggest.

But to the commuter, 20 to 30 major breakdowns a year translate to almost once every fortnight.

And it matters not if the breakdown was triggered by bad design, poor maintenance or external factors.

Yet, measuring just the number of major breakdowns alone as an indicator of reliability would not be right either, because as the rail network expands, and the number of trains running rises, the probability of incidents would also rise.

Statistically, that is almost inescapable: A bigger network is likely to have more breakdowns.

To make a measurement of reliability more robust - and indeed, more reflective of what commuters feel on the ground - we need something more encompassing.

For example, not all breakdowns are equal. So one way to make the measure more robust is to take into account how long it took to get the system running again, as well as how many passengers are affected.

We could, for example, measure downtime as a percentage of hours operated, weighted against the percentage of passengers affected in a disruption.

Downtime would include stoppages as well as delays.

This is to capture service degradation which commuters are keenly aware of, but which may not show up in the current measurement.

The percentage of passengers affected would account for the severity of each breakdown - how long it lasts and whether it is during the peak or non-peak period, for instance.

It would not be too difficult to gauge the number of passengers affected accurately, thanks to information captured by the ez-link card.

If we want to go further, we could even measure the reliability of escalators, lifts, fare gates, lighting and air-conditioning.

That would reflect the overall commuter experience more completely.

Besides system reliability, the other thing that affects commuter experience significantly is speed.

MRT trains do not seem to be travelling as fast as they did before 2011. Speed restrictions were first imposed right after the December 2011 breakdowns because investigators suspected certain stretches of the network were problematic.

Then, there were glitches when half-height platform screen doors were installed in 2012, when electronic communication between trains and doors could not sync. After that, a sleeper replacement programme, starting with the North-South Line, and now ongoing with the East-West Line, brought train speeds down.

The East-West Line's sleeper replacement is expected to be completed by the end of this year.

Until then, commuters will continue to experience slower trains.

After sleeper replacement, there will be resignalling and third-rail replacement, both of which will take a few more years to complete.

By the time the whole rail renewal programme is completed around 2019, commuters would have tolerated eight years of speed degradation.

Low train speed not only frustrates commuters, it affects the efficiency of the entire network.

To be fair, the system has improved in various other areas since 2011. The trains are not so crowded, for instance.

This is because more trains have been injected into the network, and more will follow.

The easing may also be owing to an ongoing incentive scheme which persuades people to travel before the morning peak period.

There are also more bus alternatives today than in 2011.

Steps are now being put in place to get people to cycle or to use motorised mobility devices.

This is all well and good, but the rail system remains the backbone of Singapore's public transport system. It currently accounts for 44 per cent of all public transport trips.

By 2030, when the network's length doubles to 360km, that could well rise to 65 to 70 per cent.

By then, a major breakdown will likely affect substantially more people.

Yes, those affected might have more alternatives than today because of the expanded network, but they will still feel the impact of a disruption because their journey times will stretch.

So, it is imperative that we honestly assess the system to get to the root of the problem.

LTA needs to find a measure of the network's reliability that is credible to commuters, and which measures the system's health holistically.

The system is not just the physical infrastructure and the rolling stock.

It includes the builders, operators and regulators - in short, the people.

Only then can we cure the cancer (the major breakdowns) and not just the colds (the minor delays).

This article was first published on April 7, 2016.
Get a copy of The Straits Times or go to for more stories.