Systems Developments and opportunities for performance Robert Freeman, 3 July 2014 2-Dec-18

Agenda System opportunities Sub-threshold Train Location Services

System related performance opportunities Traffic management Timetable Advisory System (DAS, C-DAS) PPRP Phase 2 – Timetable Change Asset Reliability – criticality tool

Sub-threshold Delay 2-Dec-18

Sub-threshold or the Right Time Railway Just looking at PPM gives a very distorted picture of the performance issue – it is like looking at the top of an iceberg. 5% of our PPM failures (so around 0.5% PPM) suffered no attritubed delay. A train 5 minutes out of path could be over 7 miles away from where we planned it to be. Timetable headways tend to be less than 3 minutes. In a high capacity railway trains running 1 or 2 minutes late are likely to cause reactionary delay as they begin to conflict with other timetable paths. 2-Dec-18 So why subthreshold …. Remind people though that PPM and RT are well correlated – ie improve one will tend to improve the other

Sub-threshold 2-Dec-18 Underlying – and to an extent underpinning – the performance challenges has been an increase in subthreshold delay. Whereas there are clear links into direct causation – most obviously increased passengers, autumn and other effects where common cause problems present below threshold delay, the causes of this increase are probably also representative of the wider saturation of the network and drive for efficiency and marginal improvements. The following chart provides the national trend:

KEY THINGS The following slides are taken from a variety of presentations given by the National Performance Analysis Team since 2011. The slides use examples from a number of different operators. We have removed the date of when the work was done and we know that some of the issues highlighted have now been resolved. The slides though give a good illustration of the type of analysis that is possible with the right data and the right imagination and the importance of this work to understand the railway as a system. 2-Dec-18

Washing lines and block charts Understanding lateness: Performance is not just at the end of a train’s journey Success is keeping a train in path along its whole journey Every train path behaves slightly differently WASHING LINE CHARTS – show the performance of a train along its route BLOCK CHARTS – compare the performance of a group of trains along a route 2-Dec-18 Introduce what a washing line and block chart is … we have historically looked at subthreshold by using TRUST lateness data

Performance along the line of route 1L48 Swansea - Paddington Right Time and time-to-2 performance drops early in its journey. Performance sharply falls on the approach to Slough. 2-Dec-18 One train over many days … How that train is rarely running in path anywhere on its journey (i.e. as soon as it starts its journey it seems to drift out of path)

Performance along the line of route (T5) is the 1L48 similar to it’s peers? 1L48: Swansea to Paddington compared to other Swansea to London trains. For some reason 1L48 loses much more performance around Bridgend than other trains….. It then for 30% of days it is running more than 5 minutes late through South Wales. Performance for all Swansea-Paddington trains drops on the approach to Slough especially in the peak. 2-Dec-18 To understand it is sensible to compare to other trains … in tis case 1L48 is the only Swansea-London train losing as much time in Wales … But all trains do seem to lose time around Swindon (where it goes orange) and slough (where it goes red)

Timetable tweaks – an example TT changes & effect This train had the following changes made: Left WGC 1min earlier Dwell time at Hatfield and Haringay 1min longer SRT time shifted into dwell time at Potters Bar Train arrival 1min later The effect: 10% improvement in Time-to-5 v WTT 30% improvement in T2 and RT the train keeps to RT for a greater proportion of its journey so passengers see a much more punctual train service Similar changes were made throughout the Service Group and PPM in the GN Inner is improved…

GN Inners Distribution Following the rules changes to the GN Inners the distribution of trains arriving at destination has improved notably. The following graph compares terminating lateness of trains in the service group before (blue) and after (orange) the timetable changes. Much improved distribution post-TT changes. Not just a shift towards 0 but a tightening. This means that generally the service group is less variable than before. Suggests a much more systemic improvement. 2-Dec-18 To go after your RT/PPM graph slide (GN inner headcode improvement) if you feel it is relevant. The service group as a whole has benefitted from the TPR improvements made in the May TT change. The shift towards 0 is excellent for PPM and RT scores (FCC could do with a good news story there) but the tightening of the distribution is going to be felt a lot by the passengers whose service is suddenly very punctual rather than an unknown quantity. The tightening may also suggest a more systemic improvement of trains in this service group thanks to the TPR changes. All data is WTT so will not account for extra “buffer” times at termination. The improvement in this service group is en-route.

Subthreshold Mapping from PSS Currently using Data Visualisation tools and PSS data to understand locations where subthreshold delay occurs. Using lateness rather than delay to point towards locations on the network where trains often lose time Benefits of the PSS approach to viewing subthreshold are that the data is readily available to the industry and that it gives access to years worth of historical train runs. 2-Dec-18 ATW data for P1312 to P1402 Size and darkness of spot mean a greater % of trains are delayed by less than 3 minutes at location This much data is rarely available from other systems (notwithstanding GPS for some operators) and presents subthreshold in a common language to usual performance. Key to this graph is the difference it shows to an attributed delay map. Attribution mapping tends to confirm the “well we already knew that was an issue” locations, subthreshold mapping can confirm the “I thought that might be an issue” and is aiming to provide new locations for our interest. Understanding where subthreshold delay occurs for a route/operator is a precursor to understanding why. Would struggle to do this without Tableau so not sure how you would feel about some name dropping when describing this map.

Variance Using more accurate and more comprehensive data sources such as CCF and GPS is allowing subthreshold delay to be understood at a much deeper level than is possible from PSS. Demonstrating operational variance is possible with these sources (for example, how long a station stop takes or how variable headways are). This is an important part of understanding subthreshold impact on the normal running of our train service. Displaying the variability in running highlights many factors often lost in PSS data and attributed minutes Here it is possible to quickly identify problem services and “at risk” services, those that are running behind a highly variable service. It gives a different view of the timetable to that which is on a planner’s screen – white space between trains is no longer white! Tools like this are beginning to use subthreshold data to shape the operational plan for CP5 and improve the connection to performance across the industry. 2-Dec-18 Variance graph – time from left to right along bottom, from 10:00 to 11:00 am. Top-bottom: WCML south going down from Euston. 28 days of data and colours are different train services (1H, 1J etc) At each location the variability is shown by having a dot for each day, current date is the larger dot. More variable trains (highlight light blue down the centre) are more of a risk to those around them. A larger picture shows some colours appearing to overlap others, which would cause delay. Important to highlight the difference in some trains having a tight distribution (maybe a link to GN inners idea?) as these have little chance of crossover into other paths. More of those trains would see a much less disrupted service. (dark green train second to right is best used to highlight a “good” train) Tools like this are already in limited use in NPAT and are being used to directly influence the operational plan (through PPRP stage 2).

Hindley to Daisy Hill SRT Measured Variation Plan 3 mins 30 sec’s Actual 4 mins 10 sec’s + 40 sec’s Taking the combination of multiple runs (from 10’s to 1000’s) to measure the normal time taken between two points, compared to the SRT

Measuring the actual impact of speed restrictions. In this example a 10 MPH speed was imposed, and trains stopped and cautioned, the speed was then ‘boarded’ and increased to 20 MPH, with the actual impact of these actions being properly understood The stop and caution put 8 mins into each train, the boarded 20, 1 minute

Understanding plan variation VS SRT The above examples show how train planning rules create subthreshold delay, Hexham to Haydon stop to pass and stop to stop having the same SRT

Train Location Services 2-Dec-18

Presentation Title: View > Header & Footer 2-Dec-18

Presentation Title: View > Header & Footer 2-Dec-18

Presentation Title: Insert > Header & Footer 2-Dec-18

Overview 2-Dec-18

GPS/OTDR Download fitment Timescales 2014/15 2015/16 2016/17 2017/18 2018/19 CP6 Funding GPS/OTDR Download fitment D’base Build Analytics Full integration Commercial design Transparency Benefit Exploitation

Questions / Open discussion 2-Dec-18