Passenger loadings UK

[splashoutradio]

Over the past few months, Network rail, as well as many of the rail operators, have been making some superb datasets publicly available. One of these (in my opinion anyway ) is passenger loading data. Theese use a variety of methods such as automatic passenger counters and ticket sales data to show an average of how many people board, alight and travel through each station and train. Unfortunately many of the TOCs apply restrictive license terms which prevent them being used for much, but Network Rail's country-wide version shows Monday to Friday averages (sorry, no Saturdays only Denton Parly) and has been made available for all to enjoy.

I've had an attempt at playing with the data, and in the off-chance my amatur attempt is of any interest to anyone, it's at www.railwaydata.co.uk/loadings/gbr

 

[bib]

Some very interesting stuff there. It's fun to be able to compare to my own experience of how busy certain trains have been when I've gotten them.
I'm quite suprised that the busiest train out of Derby appears to be the early morning 0610 XC, followed by the 0632 XC, both with about 200 people modelled to change at Tamworth onto various trains to London, whereas the direct 0626 EMR train is only modelled to pick up 12. Not sure if it's an issue with the modelling with there being 2 valid routes on a Derby-London ticket, or if there was some serious price undercutting going on at the time, given that the journey time is not that different.

https://www.railwaydata.co.uk/loadings/gbr/?UID=C80038

 

[higthomas]

Thanks, that's really interesting. Looking at Cambridge for example, surprising to see that the single busiest train for boarding/alighting is the last fast train from London arriving at 2401 with 412 passengers alighting!

Although there are some things that make me question the data, such as the fact apparently the 0002 train from KGX carries >200 passengers, whereas the 0032 carries 16! Not sure that agrees with my experience.

 

[Trainman40083]

Some very interesting stuff there. It's fun to be able to compare to my own experience of how busy certain trains have been when I've gotten them.
I'm quite suprised that the busiest train out of Derby appears to be the early morning 0610 XC, followed by the 0632 XC, both with about 200 people modelled to change at Tamworth onto various trains to London, whereas the direct 0626 EMR train is only modelled to pick up 12. Not sure if it's an issue with the modelling with there being 2 valid routes on a Derby-London ticket, or if there was some serious price undercutting going on at the time, given that the journey time is not that different.

https://www.railwaydata.co.uk/loadings/gbr/?UID=C80038

Fascinating. I don't think I have ever seen 200 people get off a XC train at Tamworth. And that data suggests more got off at Tamworth than Birmingham New Street. I could see people catching a London Midland service. Maybe they commute to Milton Keynes. All in all, IF the data is accurate, you can see why certain trains are the length (in terms of coaches) that they are.

 

[slipdigby]

Public health warning. This data isn't great outside of certain fleets on certain TOCs. Some operators still operate a large proportion of non-APC equipped vehicles, so rely upon spot manual counts (which are patchy and irregular) or derived data (e.g. known corridor demand with a generic time of day profile applied). Even data from fleets with theoretically large APC equipped fleets is often dodgy due to known problems/faults with the onboard kit.

 

[Trainman40083]

Public health warning. This data isn't great outside of certain fleets on certain TOCs. Some operators still operate a large proportion of non-APC equipped vehicles, so rely upon spot manual counts (which are patchy and irregular) or derived data (e.g. known corridor demand with a generic time of day profile applied). Even data from fleets with theoretically large APC equipped fleets is often dodgy due to known problems/faults with the onboard kit.

And I don't think Cross Country do have APC. Maybe people got off at Tamworth, because they thought the ticket inspector was coming.

 

[pokemonsuper9]

While the data looks cool, I've tested it on my commute and it's absolutely wrong (I am definitely not 0 people!)
I don't know if Northern's 150/156/158 have passenger counting so that may be why.

 

[Envoy]

When you click on the head code for a train, a line graph appears showing the loadings. Surely they should have produced a bar graph showing the loading between any two stations? With the line graph, it tales off at the ends of each journey; for example, a GWR train heading east to London reaches maximum load at Reading but according to the graph, nobody is left on this train by the time it reaches Paddington. Presumably the passengers have been jumping off as the train speeds along on its final leg?

(A line graph would be appropriate for showing temperature as it gradually changes).

 

[bib]

It does say "Passengers on Departure" as the axis title

 

[The exile]

Just looked at one of my regular trains- 07.04 Bath - Filton avoiding Bristol. Total loading figures possibly on for Mon or Fri last summer, but not the other three days. Interestingly, 25 people get on the train, but only 24 get off - has anyone checked the toilets on the Bristol turbo fleet recently?!

 

[splashoutradio]

I'd noticed there were a number of irregularities with the data, and but hadn't quite realised the extent to which they exist. Perhaps I'm just lucky that most trains around my neck of the woods are fitted with APC. I'm tempted to hide the loadings system until (hopefully) the TOCs relax the licensing terms of their data. *fingers crossed

Just looked at one of my regular trains- 07.04 Bath - Filton avoiding Bristol. Total loading figures possibly on for Mon or Fri last summer, but not the other three days. Interestingly, 25 people get on the train, but only 24 get off - has anyone checked the toilets on the Bristol turbo fleet recently?!

This is one of the (seemingly few) anomalies that can be explained away. As the figures used are averages rounded to the nearest whole number, anything between 0.5 passengers and 1.4 will show as one, and thus total through passengers can occasionally 'loose' or 'gain' a passenger.

When you click on the head code for a train, a line graph appears showing the loadings. Surely they should have produced a bar graph showing the loading between any two stations? With the line graph, it tales off at the ends of each journey; for example, a GWR train heading east to London reaches maximum load at Reading but according to the graph, nobody is left on this train by the time it reaches Paddington. Presumably the passengers have been jumping off as the train speeds along on its final leg?

(A line graph would be appropriate for showing temperature as it gradually changes).

I was tempted to omit the last station on the line graph to remove any confusion, but kept it in to show the full range of the chart (starting at 0 passengers, instead of whichever number is the lowest) and opted to include an 'on departure' caveat in the axis title. The use of a line diagram is a good idea though - will look into it - thank you

 

[slipdigby]

I'd noticed there were a number of irregularities with the data, and but hadn't quite realised the extent to which they exist. Perhaps I'm just lucky that most trains around my neck of the woods are fitted with APC. I'm tempted to hide the loadings system until (hopefully) the TOCs relax the licensing terms of their data. *fingers crossed

I'd keep it up, but just heavily caveat it. Which is really what the data owner (I presumed DfT, but you mentioned Network Rail) should have done before release - not your fault!

Keep up the good work!

 

[FenMan]

I'm impressed by the loading of the 0540 Edinburgh - London King's Cross.

Quite the feat to board 1,661 passengers at Edinburgh.

https://www.railwaydata.co.uk/loadings/gbr/?UID=C47015

 

[slipdigby]

I'm impressed by the loading of the 0540 Edinburgh - London King's Cross.

Quite the feat to board 1,661 passengers at Edinburgh.

https://www.railwaydata.co.uk/loadings/gbr/?UID=C47015

And if my assumptions are correct as to the source, this dataset has been used by parts of the industry as a relatively unquestioned source of information for quite a few years

 

[Clarence Yard]

I suspect NR’s load modelling is way off. They don’t get actual load information on a daily basis so they must be modelling it in some weird way off OND or MOIRA if these are the train figures they are producing! The ECML figures are truly bonkers.

TOCs and the DfT tend to use their own actuals for their own purposes and these are still classed as confidential to the TOC/DfT. OA loads don’t go anywhere near the DfT or NR because of the Competition Act constraints.

 

[slipdigby]

I suspect NR’s load modelling is way off. They don’t get actual load information on a daily basis so they must be modelling it in some weird way off OND or MOIRA if these are the train figures they are producing! The ECML figures are truly bonkers.

TOCs and the DfT tend to use their own actuals for their own purposes and these are still classed as confidential to the TOC/DfT. OA loads don’t go anywhere near the DfT or NR because of the Competition Act constraints.

Apologies, you're absolutely correct here and I'd totally forgotten about NR's MOIRA based loadings exercise which this data clearly is. My apologies. So interesting academic exercise, but effectively meaningless outside of the original context.

To be honest I was a bit surprised that the Green Book counts had broken cover! However I had heard that there was talk of reconsidering when non-revenue datasets become non-commercially confidential so foolishly assumed that this was the case.

Cheers!

 

[Magdalia]

I've had an attempt at playing with the data, and in the off-chance my amatur attempt is of any interest to anyone, it's at www.railwaydata.co.uk/loadings/gbr

Thanks for the work you have done so far. It doesn't look amateur to me, the presentation is good.

When you click on the head code for a train, a line graph appears showing the loadings. Surely they should have produced a bar graph showing the loading between any two stations?

I do agree with this point, bars showing the loading between each station would be better.

And is it possible to link to rostered train formations and then show the total number of seats on the train?

There is also a typo where average is shown as avaerage.

Theese use a variety of methods such as automatic passenger counters

Are you able to say which trains have these?

I know about class 700s because loading information is displayed in the train. What is in the data broadly matches my experience on Cambridge-Brighton trains that I use frequently.

But what about class 720? I have looked at 1807 Liverpool Street-Ely quite a few times and it usually looks more heavily loaded at Liverpool Street than what the data says.

I'm guessing that class 387s on the Kings Cross-Kings Lynn/Ely trains don't have APC, is that right?

I don't think Cross Country do have APC.

At Cambridge the data for the Birmingham-Stansted trains looks the most dodgy, especially in the peaks, where the Cross Country trains appear to be more heavily loaded than the Great Northern trains that have more than twice the capacity.

 

[Vexed]

I'm guessing that class 387s on the Kings Cross-Kings Lynn/Ely trains don't have APC, is that right?

I think they might, you can see per-service loading for basically all of GTR on the brands respective websites. The wording on the site implies on-train measurement.

https://www.greatnorthernrail.com/travel-information/onboard-travel/find-a-quieter-train

Then if you can see a colour coded timetable if you clock though. It takes a bit of time to load...

 

[Magdalia]

I think they might, you can see per-service loading for basically all of GTR on the brands respective websites.

That sort of traffic light data existed 15-20 years ago when I was commuting with First Capital Connect on class 365 and class 317. It doesn't need APC.

 

[DanNCL]

I'm impressed by the loading of the 0540 Edinburgh - London King's Cross.

Quite the feat to board 1,661 passengers at Edinburgh.

https://www.railwaydata.co.uk/loadings/gbr/?UID=C47015

I’d be surprised if the true figure was much more than 100 on that specific service. It never seems to be busy when I see it in Newcastle and it’s hardly a surprise when LNER charge so much for it.

 

[Horizon22]

I've looked few a of the journeys and stations near me and ones I've historically used and some of this is scarcely believable - apparently there are exactly "0" boarders at a station in the early morning that I regularly see at least 50 people get on at, and for some reason there are stations where apparently 80% of the entire train is alighting at one metro stop but not the equally busy station 3 minutes down the line.

OP, you've got a good site (would be nice if searching by the 3-letter CRS code worked though!) and it's just the limitations of the data provided.

 

[splashoutradio]

Thanks for the feedback, everyone. Really appreciate it!

The data comes from NRs Strategic Rail Analysis Model (or SRAM for short ), although I'm looking to replace this by more reliable TOC data as it becomes available with suitable licensing.

The line graphs have been replaced by bar charts, with the terminal station removed, to hopefully remove any confusion.

would be nice if searching by the 3-letter CRS code worked though!

Great idea! Will get on to this.

 

[Geeves]

I believe the only trains at Northern with APC equipment are the 195s and 331s so any routes using those would be a bit more reliable (if it's working or turned on). Just from my work eye ball I would say the figures for Victoria look a bit low but probably not far off.

 

[The Ham]

Again, nothing to do with the OP, the data for at least some SWR services looks off, with some peak hour services having around 200 people get of at London. Historically I would have guessed that many got off at Farnborough (as often the bridge from the London bound platform was almost full of people as there were so many students (colleges and private schools).

One option the OP could consider (not sure it's possible to program this way) to show loading at the terminus stations could be to use the alighting data for the last column (although, if even possible, that does make creating the graph more complex, so can understand why that might be a level of complexity which isn't worth the effort)

 

[Killingworth]

Firstly to say what a pleasure it was to discover these statistics by accident last month when looking for something else on the Railway Data website. I was diverted for far too long from what I'd been intending to research.

My initial delight is now tempered with considerable alarm if they are being relied upon to determine service levels and stopping patterns.

They say they're averages for weekdays during the summer timetable. Is that for every day during the currency of the timetable? What are the start and finish days for the averages?

My local station is Dore & Totley on the Hope Valley, Manchester-Sheffield line with 3 operators so this link brings up some interesting data. https://www.railwaydata.co.uk/loadi...YklX3sjr93ByS69e_w_aem_2g_LHPfNBniEFB_ySlDVVg

Yesterday* I caught TPE's 1B66 at 8.04 for Sheffield along with about 12-15 others, well short of the 25 suggested by the data. For a 6 car train it would be difficult to say if 67 were travelling through but it wouldn't have been over 100. However that was on a Monday.

I could have caught Northern's slightly earlier 2S70 leaving for Sheffield at 7.57 (arriving at 7.50). According to the data it loads an average of 3 with 28 travelling through. WRONG! It's a very popular commuting train into Sheffield from the Hope Valley, the previous train having been 2S90 that calls at Dore at 6.50 and the next not until 2S72 calling at Dore at 8.55. (Until December 2022 there was another popular Northern departure calling at Dore at 8.24 - axed as a peripheral change resulting from the Manchester recast.)

Commuter numbers have certainly fallen a lot since Covid but I'd estimate the 7.57 still picks up at least 25 at Dore, far more than the 3 shown in these statistics. On Tuesday 7th January 2021 a local head count recorded 51 boarding that train with another 29 on the following TPE service only 7 minutes later. There were 54 boarding the Northern 8.24 and a further 21 on EMRs 8.28.

Northern use 195s that are supposed to be able to count passenger numbers. If their figures are being relied upon I hate to think how reliable those from other rolling stock may be.

There's lies, damn lies, and statistics. Don't rely too heavily on these.

*I'll be catching TPE's 1B66 again on Thursday at 8.04 - unless I get there early enough for the Northern 7.57. (It provides a nice warm waiting room for 5 minutes on a platform with no covered seating)

 

[The exile]

Northern use 195s that are supposed to be able to count passenger numbers. If their figures are being relied upon I hate to think how reliable those from other rolling stock may be

How do they do the “count”? If it’s by weight it must be a very poor indicator of “churn” at a station (how does it distinguish between “none off, 1 on” and “ninety-nine off and 100 on’?

 

[driverd]

As with all the railway data, there's certain limitations, and substantially more reasons to be sceptical. For instance, 1Y11, the CS between Edinburgh and Fort William, carries more passengers than 1S25 (EUS-INV), which lived experience tells everyone who's used that service, simply isn't correct.

From lived experience, myself and a few other colleagues I've discussed with also question the accuracy of the station pairs journey data that's published (but that's another conversation).

 

[Northern158170]

For anyone looking at Harrogate line data, I’d say multiply the figures by just under 2 for the Harrogate to York bit. I’ve just done my own rudimentary counts for departing passengers at Knaresborough and Starbeck and they are about double what the dataset suggests

 

[Route115?]

Is there a methodology published anywhere? During my time working for the railway I've used train loading data (which doesn't record the number joining and alighting at each station), manual counts and MOIRA data (which is based on profiles which may not be totally accurate and do not account for train specific tickets purchased in advance) - all methods have their problems and limitations. I believe there is also a method of counting using infrared beams to look at those joining and alighting.

I did a doodle looking at Avanti & LNER plus open access operators that serve Kings X. For the latter I compared the total number of those joining and alighting at each station for ALL operators with ORR station usage data which isn't available by TOC. I would be a bit wary - the loading data looks at a "perfect" weekday and industrial action and disruption will affect passenger journeys, loads may be lower on Sunday, those transfering between two trains of the same operator will be excluded and so on, so my uplifting of ORR data may be too high resulting in a smaller difference between the two sources. Beware! That said some of the data looks rather dodgy, particulary Lumo.

Lumo S/B	Total	1S91_	1S93_	1S95_	1S97_	1S99_
Kings Cross		05:48	10:45	12:18	14:48	20:27
Stevenage		06:10		12:41
Newcastle		08:42	13:47	15:17	17:47	23:08
Morpeth		08:56	14:01	15:31	18:01	23:23
Edinburgh		10:07	15:11	16:41	19:11	00:43
Boarding
Kings Cross	2,782​	130​	198​	281​	199​	1,974​
Stevenage	98​	31​		67​
Newcastle	397​	138​	48​	86​	81​	44​
Morpeth	77​	37​	17​	11​	11​	1​
	3,354​	336​	263​	445​	291​	2,019​
Alighting
Newcastle	518​	53​	51​	127​	57​	230​
Morpeth	174​	9​	12​	42​	68​	43​
Edinburgh	2,663​	273​	200​	276​	167​	1,747​
	3,355​	335​	263​	445​	292​	2,020​
Lumo N/B		1E80_	1E82_	1E84_	1E86_	1E88_
Edinburgh		06:23	09:24	11:24	16:13	19:58
Morpeth		07:38	10:38	12:34	17:26	21:13
Newcastle		07:57	10:55	12:51	17:52	21:35
Stevenage				15:21	20:20	00:26
Kings Cross		10:46	13:47	15:44	20:44	01:05
Boarding
Edinburgh	741​	67​	207​	219​	114​	134​
Morpeth	167​	64​	38​	49​	15​	1​
Newcastle	388​	123​	99​	69​	58​	39​
	1,296​	254​	344​	337​	187​	174​
Alighting
Morpeth	70​	1​	12​	22​	11​	24​
Newcastle	225​	35​	32​	75​	40​	43​
Stevenage	96​			29​	29​	38​
Kings Cross	906​	218​	300​	212​	107​	69​
	1,297​	254​	344​	338​	187​	174​

Here is some data in case you don't have excel. I have my doubts about 1S99, it could be rather cosy on board! I'll have another look if I get the time - which won't be for a while I'm afraid.

 

[Route115?]

Having a bit of spare time I repeated the exercise for the Great Eastern lines out of Liverpool St which are attached. I've only included services to London plus some shuttles. The results for Romford look unlikely. I suspect that they are based on MOIRA profiles which will assume that passengers use faster trains regardless of there being space available. I have compared stations used primarily by the services shown (so excluded the like of Norwich) with the ORR station usage data for 2023/4, and allowing for the fact the data is for a summer 24 weekday, they look reasonable.

== Doublepost prevention - post automatically merged: 27 Apr 2025 ==

I've also added the results for LM services out of Euston. Load is the total load over the section of track (an arc enquiry). Red is where a train is routed on the main line and not via Northampton.