We are operating more trains and moving more passengers and freight than ever before and our asset performance is at an all time high. We are reducing the number of incidents that affect train performance as well as our operating costs while undertaking record levels of renewals and we have a large and expanding programme to increase capacity.

So why the headlines about “fines for poor performance”?

Well, even though train performance continues to improve, we are not going to meet all the punctuality targets set by ORR back in 2008 for the current five year Control Period (CP4). Perhaps that’s not surprising on a more congested network when many of the targets did not anticipate the growth we have seen. However, we must do, and are doing, all we can to meet our own revised internal targets and to get as close as we possibly can as an industry to those set for us for CP4.

Punctuality is clearly critical to passenger satisfaction, which is why we are quite pleased with the progress we have actually made in improving the Public Performance Measure (PPM) and Cancellation and Significant Lateness (CaSL) measures, recognising there is yet more we can do. But also important to passenger satisfaction are frequency, capacity, journey time, stopping patterns, interconnections, consistency, evening and weekend availability and the access we need to do essential maintenance and renewals. And cost.

As this article will illustrate, many of these are now valued as highly by our customers as further increases in punctuality much beyond what we are currently achieving. That is partly why we have different targets in the joint performance plans we have developed with our customers. The debate has moved on. We certainly need a different approach for CP5, 2014-2019.

There are four main parts to running a railway. It all starts with the timetable which is planned and updated twice a year to accommodate new services and to reflect what we have learned in analysing delays with current performance. Then we need to have the infrastructure in the best possible condition and minimise external events such as cable theft and fatalities and mitigate the impact of more extreme weather. Third is the seamless operation of the timetable by our signallers and controllers together with drivers and station staff to keep to the plan. And then we all need to respond quickly when something more serious happens – and keep doing all of this relentlessly, day in day out.

Timetable structure

We have two. They are closely related but have important differences. And, if either of them is substandard, then everything that follows is much harder.

The GBPRT (Great Britain Passenger Rail Timetable) is what passengers see and expect and what PPM is measured against. It is also the basis for the delay compensation mechanism between Network Rail and the train operators.

The Working Timetable (WTT) is the internal timetable or operating plan and reflects the actual engineering and operating characteristics of the infrastructure and trains. It is built up from many component parts (e.g. linespeed, train characteristics, stopping patterns) to create overall journey times. Trains are actually operated to this timetable.

Distributed within the journey times that make up the WTT are some small additional time allowances, typically a minute here or there, to help manage each train and the overall network on the day. These might reflect operating conflicts at junctions or stations, or they might take account of the likelihood of a Temporary Speed Restriction (TSR) after a planned track renewal or because of poor track quality or other reasons such as the sighting at a level crossing. There are further allowances between the WTT and the GBPRT (known as “Public Timetable differentials”) which typically reflect the commercial needs of operators. Their size and location are at the Operators’ discretion. It follows that journey times in the GBPRT will invariably be longer than in the WTT with the planned differences being fixed at given locations although the actual differences may occur elsewhere.

Developing and refining the WTT is a skilful and complex task.

What is delay?

The average attributed delay for the network is less than two minutes per train. That may seem surprisingly low but makes sense in the round. On a typical weekday we operate around 22,000 trains and Network Rail causes roughly 22,000 minutes of attributed delay, and the TOCs cause half as much again. So, after allowing for freight, around 1.5 average delay minutes a train seems about right. But of course that is against the GBPRT, which already had the few minutes of allowances against the WTT.

A further complication is that we measure two types of delay – delays (and cancellations) caused by an incident and the delay attributed to (or cancellation of) an individual train. They are related but rather different.

For an incident (e.g. a points failure, fatality or TSR) we record the cumulative impact on all the trains affected, each one by a different amount. And some of the delays may be incurred a long way from the location of the incident and much later – a few hundred miles away and several hours later is quite possible.

The attributed delay to a train is specific to that train along its journey. It may be “primary” delay, because the train is directly and immediately affected by an incident on the route, or “reactionary” delay when the delay is transferred from a different incident elsewhere by other trains. Over the last few years we have seen a massive growth in reactionary delay, reflecting the more congested network.

Measuring delays

We are measured against the Public Performance Measure (PPM) of each train and the aggregate number of delay minutes we cause. PPM has different timebands depending on the nature of the train service.

The industry’s systems firstly record how each train is running compared to its WTT timetable at regular points along its journey. When the degree of lateness in a recording section (known as a “Trust” section) is three minutes or more of real delay, the system asks for the cause of such delays to be captured through the delay attribution process. Delays of less than three minutes (“sub threshold” delays) are normally not attributed. There are roughly as many delay minutes recorded that are above as below the threshold.

An individual train’s lateness against the GBPRT at certain monitoring points is also recorded – usually where it makes intermediate stops and of course at the end of the journey. And then what is called “average lateness” is computed. This is a blend of all the lateness recorded for all trains run in that service group, against the GBPRT timings at monitoring points along the route, weighted by expected traffic loadings.

The very important next step is the allocation between Network Rail and the TOC of average lateness of trains against the GBPRT which is based on the Trust delays for incidents on the day as picked up from delays against the WTT. It is this attributed average lateness that is used for the delay compensation calculations.

Analysing delays

We spend a lot of time on this so it is worth reviewing some of the tools we use. The arrival bell curve is a fundamental measurement tool for analysis.

The aim is clearly a steep curve with a low tail and as many trains as possible before either the 5 or 10 minutes “late” used to measure PPM. Poor regulation, extended dwell times, loss of time in running and TSRs are some of the many causes of more trains arriving beyond the cut-off point. When PPM slips below a target, a relatively small percentage actually moves to just the other side of the cut-off point.

A second useful tool is to look at the accumulation of delays along a route for a particular train. An example of these “washing line” graphs charts 95 separate runs of Virgin’s 14.15 Manchester – London and shows the percentage of trains that are within different time bands (left hand scale) compared with the right time path and where the allowances are (right hand scale in minutes).

The difference between average lateness and PPM

The average lateness is designed to measure, as accurately as possible, the average lateness of a person disembarking from that train. For this reason it measures the lateness of each train not just at its terminus, but also at the key locations on its journey. Each location is given a weighting based on the proportion of people disembarking at that station – a London to Aberdeen train would be measured at Peterborough, Doncaster, York, Newcastle, Edinburgh and Dundee and the average lateness, and therefore the delay compensation cost, will depend much more on the performance at these intermediate stations than the lateness of the train arriving at Aberdeen. Note also that average lateness is an absolute measure so every minute of lateness counts – the threshold for inclusion is “on time” not the PPM bandwidth.

This compares to PPM which only considers how late the train is at destination. For many trains the number of people getting off at the final destination might be relatively low. Thus, where allowances are back-end loaded in the timetabled journey, a train may have poor punctuality along the route but could end up with reasonable PPM.

Delay compensation

Once the systems have calculated the types and levels of delay a financial value is then ascribed. This is known as “Schedule 8” within the track access agreements. The industry’s performance regime is designed to make sure that the TOC has no windfall gain or cost from unexpected Network Rail performance and so it is based on a link between performance and user-ship and therefore TOC revenue. It should be neutral during this control period if performance targets are met. There is a netting off each month and (previously at least!) Network Rail typically pays money during Winter and Autumn when performance tends to be poor and receives money during the rest of the year.

Overall this creates a situation where it might appear that four aspects of the performance regime are now rather more stacked against Network Rail:

• It is responsible for external events (including weather, cable theft etc) and these have not reduced in incidence;

• Network Rail tends to be responsible for bigger incidents which are themselves getting larger. They have a more pronounced increase in delay per incident (DPI) from a tighter network and more stretched industry resources. Also the type of incident attributed to Network Rail usually shuts more of the network – it can be easier to work round a failed train than a fatality or signalling fault;

• Its benchmarks are getting tighter and the TOCs are not;

• As the allowances get squeezed between the GBPRT and the WTT a payout is more likely if the regime is not recalibrated.

What has changed in the last few years?

Given this background it is worth looking at what has changed since the targets were set before the start of Control Period 4. The table above shows the variance in 2011/12 against the CP4 assumptions for that year.

Firstly it is worth noting that, although the targets are being missed, PPM overall has increased by around 1.5% in CP4 and that the growth in train miles (originally planned and additional) is now 8.4%. Making the broad assumption that the changes in train miles, trains run and passenger loadings are similar this indicates that the absolute number of trains or passengers delivered within PPM over the last couple of years has increased by around 10% or nearly 2000 trains a day. The industry is not getting credit for this.

The significant implication is that the network is simply busier, making it more difficult to keep to time after an incident. And the other factors mentioned in the opening section are now more valued by operators than further increases in punctuality alone.

A perhaps surprising number of trains are still arriving on time or early. But the arrivals bell curve graph has shifted to the right by an average of about one minute in the last 12 months. The important detail is where the line crosses the +10 minute mark. Network Rail is now engaged in a massive effort to try and move what is already quite a flat tail by a very small amount (down and to the left) to raise the area under the curve (up to the +10 minute cut off) from around 88% to 92% by the end of the control period. This indicates how we are now dealing with very small margins.


I am sure there are many more mature readers who are now thinking

a) this is all far more complicated that it used to be or need be and

b) if you have to understand all this then how can you have time to run the railway and

c) simply focussing on the precision of a right time railway with a good timetable, professional signalling and excellent asset condition will make everything else come right.

And they would be correct.

So let me assure everyone that we will continue to do all we can to drive up performance. And of course there is a strong counterargument to what I have set out above. With a few minutes in the hour between the WTT and the GBPRT and a further 5 minutes for a time to 5 PPM service there is a lot of scope to get it right. This is the perfectly reasonable argument of Passenger Focus and others – the industry should not be claiming that five minutes late is “on time” hence we have also started to report “right time”.

But the continued success of the railway has moved us all to a position in which focussing solely on a couple of performance metrics is necessary but no longer sufficient. Especially when these targets were set four years ago and given the growth we have seen since. There are other important factors to balance in running the network. We have important trade offs to make between performance, capacity and cost.


In this article, Robin Gisby has explained the challenges faced by Network Rail in monitoring and measuring punctuality performance against set targets.

Next month, the rail engineer will look at the steps being taken from an engineering point of view to minimise those delays.