Interpretation of "impartial retailing"

[tony_mac]

There are far more possible road routes between London and Istanbul than rail routes between A and B in the UK - any that are not in the RG can be ignored for a start.

You don't understand the differences between the two problems; I will try to explain a little.
A (weighted) shortest-path is easy to find - it's even easier for roads/rail as you have an approximate distance to the finish point.

e.g.
http://en.wikipedia.org/wiki/A*_search_algorithm

But, if you are dealing with roads then the length of travelling from A to B then B to C is the length of A to B + the length of B to C. It is also at least the (geographical) distance between A & C.

The first condition is not true with railway pricing, and the second cannot be meaningfully used.

This means that calculating a cheapest route is not easy (or polynomial-time computable), as you have to try all of the alternatives.

Obtain the possible split points using the actual stopping points of trains on permitted routes

For a reasonably large map (like map AS) this quickly gets into numbers that are not really practical to deal with, even for a computer (assuming you are allowing as many splits as are required, and not just the one that you suggest).

 

[bb21]

Well, there I would need a computer!

1. Obtain the possible split points using the actual stopping points of trains on permitted routes
2. For each split point
2.1 Find fare Aberdeen to split
2.2 Find fare split to Brighton
You now have an array of at most 2,500 pairs of fares (assuming the worst ie every GB station is a possible split point)
3. Find the cheapest through fare
4. Are any of your split fares less than this? If so, for the cheapest split, find if the fare is available and present it to the user as an option, with caveats.

A trip like Aberdeen-Brighton will be permitted via just about any route in the country apart from deep Westcountry and Wales. A comprehensive split fare check is not possible as that would include, say there are only 100 stations along all permitted routes, 100 x 99 x 98 x ... x 1 = 100! variations. Do you have any idea what size this number is? That is without taking into account other factors.

If you are only talking about checking splits along the specific route of a through train then I think there are people who are already on this case, although they're mainly for recreational purposes and not in any official capacity.

 

[DaveNewcastle]

You now have an array of at most 2,500 pairs of fares

and then, re-iterate the process, to determine every possible split for each of those pairs.
The number of possibilities multiplies exponentially.

For each of these, you'd better examine starting short and/or finishing short journeys for cheaper ticket options too.

Never mind the fare calculation, can you even work out how many component journey fragments may be involved?

 

[Greenback]

Even the short lived splitfare website had a disclaimer to state it was limited to one or two splits, if I remember correctly.

 

[hairyhandedfool]

And to illustrate a point made, the Wigan-Reading Anytime Single fare is £74.50. I know joe public can travel from Wigan North Western to Reading for under £15 on a maximum of two tickets, traveling on only two trains, at anytime, with no advance fares being used. This is done by starting and ending short.

 

[lemonic]

Also don't forget that although you may be technically able to find the cheapest tickets by splitting, the answer may not be practical as your train has to stop where the tickets are split and this may be inconvenient and could increase the journey time considerably. A split at a major station compared to a minor station which costs slightly more is probably more desirable but this may not be able to be quantified easily by a computer.

 

[OwlMan]

There are over 5,500 origins/destinations in advantix. That gives over 30 million pairs (You have to include A-B and B-A as some fares are asymetrical)
Between each pair you probably have at least 10 different fares (off peak, anytime, single, return, day etc)thats 300 million fares.
Taking a middle distance route Birmingham - Doncaster there is probably an average of 10 stations per route and probably over 10 different routes.
10 stations gives you 1.8 million different splits over ten routes that is 18 million.
Between any two stations there are probably at least a quarter of all fares valid - 75 million fares
That gives you 1.25 billion possiblities . Now multiply by 30 million to get every route that gives you 37 quadrillion (or 15 zeros). As long routes will have a far number of splits this is probably an underestimate.

Not so easy is it?

Peter

 

[MikeWh]

My contribution is to point out that Martyn Lewis of moneysavingexpert.com tried to get some programmers to come up with an application to find the cheapest split. He is not someone to give up easily. He didn't manage to do it.

 

[exile]

There are over 5,500 origins/destinations in advantix. That gives over 30 million pairs (You have to include A-B and B-A as some fares are asymetrical)
Between each pair you probably have at least 10 different fares (off peak, anytime, single, return, day etc)thats 300 million fares.
Taking a middle distance route Birmingham - Doncaster there is probably an average of 10 stations per route and probably over 10 different routes.
10 stations gives you 1.8 million different splits over ten routes that is 18 million.
Between any two stations there are probably at least a quarter of all fares valid - 75 million fares
That gives you 1.25 billion possiblities . Now multiply by 30 million to get every route that gives you 37 quadrillion (or 15 zeros). As long routes will have a far number of splits this is probably an underestimate.

Not so easy is it?

Peter

300 million fares in a single table can be looked up in a couple of milliseconds using an index. We're going to be using a computer! And I'm not suggesting we look up a route involving buying a ticket from Brighton to London Road, London Road to Preston Park and ending with Portlethen to Aberdeen. We can probably limit it to, say, a maximum of 2 legs each way (who wants to have to travel with a wallet full of 100 tickets?) as long as that's what's stated on the tin, so we're not going to have to look up trillions of permutations.

In my Warrington-Newcastle example I probably need to make a dozen searches, given that advance tickets tend to be available on, for example, direct trains Warrington-York, Leeds-Newcastle, Manchester-Newcastle, York-Newcastle. In fact, split tickets are only likely to be cheaper in this sort of scenario, where advance tickets are available for a section of the route.

If it's impossible, what's this:-

http://splityourticket.co.uk

Admittedly they don't have Aberdeen to Brighton..... but then they're having to create their own database on the fly since the industry won't let them at the real data.

 

[bb21]

If it's impossible, what's this:-

http://splityourticket.co.uk

They're not using an algorithm to work out cheaper splits. They are merely a website containing known split combinations.

 

[Greenback]

300 million fares in a single table can be looked up in a couple of milliseconds using an index. We're going to be using a computer! And I'm not suggesting we look up a route involving buying a ticket from Brighton to London Road, London Road to Preston Park and ending with Portlethen to Aberdeen. We can probably limit it to, say, a maximum of 2 legs each way (who wants to have to travel with a wallet full of 100 tickets?) as long as that's what's stated on the tin, so we're not going to have to look up trillions of permutations.

In my Warrington-Newcastle example I probably need to make a dozen searches, given that advance tickets tend to be available on, for example, direct trains Warrington-York, Leeds-Newcastle, Manchester-Newcastle, York-Newcastle. In fact, split tickets are only likely to be cheaper in this sort of scenario, where advance tickets are available for a section of the route.

If it's impossible, what's this:-

http://splityourticket.co.uk

Admittedly they don't have Aberdeen to Brighton..... but then they're having to create their own database on the fly since the industry won't let them at the real data.

Some people will happily carry 200 ticket sif they can save £2.50!

Please note the caveats on the website:

There are no first class fares listed.
The maximum realistic number of splits listed is seven journeys. For journeys over long distances (i.e. Aberdeen to Brighton) it will not be possible to break the journey into seven smaller journeys.
This website provides advice on suitable splits but does not guarantee that all the train services will call at the stations listed. It is with each customer to check the timetable and ensure that the train calls at the required stations prior to taking your journey.

Now the discussion here is whether it shoould be incumbent on the industry and it's representatives to offer splits as part of the requirement to provide the cheapest fare. At the moment this is limited to the cheapest through and direct fare between origin and destination.

If splits had to be included, it would be very difficult to provide what is definitively the cheapest possible fare as there are so many factors involved. Note also that this website, whcih appears to be pretty similar to the old splitfare one, makes no bones about the fact that that is not what it is about. In effect, it is attempting to provide an additional tool to help the intending passenger make an informed decision about whether to split or not. This is quite a different proposition to what would be expected of the rail industry!

 

[yorkie]

u

If a human being can do it (and I can), a computer can do it.

The person you were responding to knows more than most on how to tackle this particular problem! (I've met him) however if, by 'I can do it', you mean you can guarantee to find the cheapest possible fare, or combination of fares, for any given journey, then you're better than me, as I can't!

I sometimes can, perhaps even usually can, but always? No. People often PM me or tell me in person (e.g. at a forum meet) that they have found a new way to do a particular journey cheaper (that I was unaware of).

It's impossible to do it consistently.

You can have a go, you can get close, you can usually manage it, but that's not the same thing.

A computer program to manage all this would have to have knowledge of all the fares AND be tied into the timetable AND know how much extra time the person is prepared to spend to save any particular amount (which you don't always know unless presented with the options).

If you like, I can give you some particularly good examples (which I know the answers to, but you can see how you get on with) to get going with, which we can discuss next week (I will reply to your PM when I am back home)

I used to think it would not be too hard to achieve such a program. But the more I learn, the more I realise it is much, much harder than I thought it would be...

 

[exile]

Some people will happily carry 200 ticket sif they can save £2.50!

Please note the caveats on the website:

There are no first class fares listed.
The maximum realistic number of splits listed is seven journeys. For journeys over long distances (i.e. Aberdeen to Brighton) it will not be possible to break the journey into seven smaller journeys.
This website provides advice on suitable splits but does not guarantee that all the train services will call at the stations listed. It is with each customer to check the timetable and ensure that the train calls at the required stations prior to taking your journey.

Now the discussion here is whether it shoould be incumbent on the industry and it's representatives to offer splits as part of the requirement to provide the cheapest fare. At the moment this is limited to the cheapest through and direct fare between origin and destination.

If splits had to be included, it would be very difficult to provide what is definitively the cheapest possible fare as there are so many factors involved. Note also that this website, whcih appears to be pretty similar to the old splitfare one, makes no bones about the fact that that is not what it is about. In effect, it is attempting to provide an additional tool to help the intending passenger make an informed decision about whether to split or not. This is quite a different proposition to what would be expected of the rail industry!

Well, it would be a start if it offered even that! And should be able to do a lot better than this website given it has the raw data to work on.

I have to admit I hadn't considered the option of splitting a journey up into dozens or hundreds of legs in order to save a few pence - clearly this would be more or less impossible. I was thinking in terms of 2 legs, not even the 7 offered here - which in most cases will offer a substantial saving when there are advance tickets available for part of the journey but not the entire journey (which is the case for Warrington-Newcastle).
So all I'd be asking is the cheapest fare possible within those parameters.

I am, as it happens, a computer programmer. And my inside info comes from another programmer, who works for an online ticket retailer. The reason he advanced for not offering splits is not because it wasn't technically possible (given a limit to the number of legs). It really isn't a problem for a computer to search a database with 300 million fares on it, the existence of indexes means it's hardly any slower than searching one with 300! Clearly, if it had to do it millions of times to search all possible permutations from hundreds of different splits that would be a tad slow.

 

[hairyhandedfool]

But even now you are simply comparing fares, nevermind time restrictions, stopping patterns, service frequency, rangers, rovers and routeings, and I've probably missed things again.

 

[exile]

u
The person you were responding to knows more than most on how to tackle this particular problem! (I've met him) however if, by 'I can do it', you mean you can guarantee to find the cheapest possible fare, or combination of fares, for any given journey, then you're better than me, as I can't!

I sometimes can, perhaps even usually can, but always? No. People often PM me or tell me in person (e.g. at a forum meet) that they have found a new way to do a particular journey cheaper (that I was unaware of).

It's impossible to do it consistently.

You can have a go, you can get close, you can usually manage it, but that's not the same thing.

A computer program to manage all this would have to have knowledge of all the fares AND be tied into the timetable AND know how much extra time the person is prepared to spend to save any particular amount (which you don't always know unless presented with the options).

If you like, I can give you some particularly good examples (which I know the answers to, but you can see how you get on with) to get going with, which we can discuss next week (I will reply to your PM when I am back home)

I used to think it would not be too hard to achieve such a program. But the more I learn, the more I realise it is much, much harder than I thought it would be...

I will admit defeat here and now - I'm sure I couldn't tell you whether it's cheapest to get from London to Penzance by getting off at Acton, buying a ticket to Ealing Broadway and so on ad infinitum!

I recognise the difficulty now (being able to consider splitting the journey into as many legs as there are split points) but it shouldn't be necessary if a caveat is added to the "split ticket finder" eg only offer a maximum of 2 legs in each direction.

If the industry is supposed to offer you the cheapest fare - well, it's not even managing that reliably, limiting itself to through fares, given the anomalies we see on these threads every day. So maybe it is expecting too much to look for anything better!
--- old post above --- --- new post below ---

But even now you are simply comparing fares, nevermind time restrictions, stopping patterns, service frequency, rangers, rovers and routeings, and I've probably missed things again.

Maybe we need human beings for that!

 

[radamfi]

Would you suggest a similar system here, where there are no period returns, no advances and, presumably no local rangers or rovers?

If so, how would you deal with those people who currently usually buy off peak returns and stay away for a day or tow (or longer). How would you address the concerns of individual TOC's that claim they will lose revenue?

I'm not trying to have a go at you, I would love to have a simpler system based on distance. But realistically, some industry members and customer groups are going to need convincing that they will not lose out.

There is no need for period returns if all fares are singles.

Singles are only valid for one day in the Netherlands but there is no reason why that needs to be the case should a similar system be introduced here. Germany has distance based walk on fares but also has quota based Advance type fares and regional rovers, as well as PTE-type tickets. Of course, you can end up with anomalies if you mix ticket types.

Clearly if a sensible walk on fare system was introduced in the UK then some fares would go up and some would go down if it was revenue neutral.

 

[RJ]

Maybe we need human beings for that!

Agreed. Admittedly I'd offer splits including use of rangers, but only if a customer was actively "keen" to save money and asked if there was anything else I could do for them. At this point, they happily accept whatever combination you come up with, as long as it gets them home. Even then, I'd only split at points where they'd change trains anyway, or it was 100% likely that their train was calling at.

I did that because I enjoyed exploring the system. It's utterly unfeasible to expect staff to learn/know about these things and I don't think anyone in their right mind would waste labour time incorporating it as an offical topic in a training syllabus.

 

[talltim]

So we've pretty much established that the TOCs have created (and to be fair inherited) a fare structure that makes it impossible to meet their obligation to provide impartial ticketing, even with a supercomputer at every ticket window!
How about short travel? Firstly, my impression is that it would be a lot easier to calculate any savings from that than splits+limitations.
Secondly, what's the thoughts on whether ticket offices should advise on these where cheaper? Except in the case of advances, these will normally involve the same routes and time limitation as the straight A-B ticket, so be no more complicated to use for the customer (although probably harder to understand the logic of!)
An example. Until the last price rise in January, a Chesterfield-Dore ticket was cheaper than a Chesterfield-Sheffield ticket*. As you have to go via Sheffield to get to Dore anyway you might as well get the Dore ticket. If I had gone to the ticket office and asked for the cheapest ticket that would get me to Sheffield, should they have recommended the Dore ticket?

*possibly a very long standing anomoly left from the days when Dore had platforms on the MML

 

[Greenback]

There is no need for period returns if all fares are singles.

Singles are only valid for one day in the Netherlands but there is no reason why that needs to be the case should a similar system be introduced here. Germany has distance based walk on fares but also has quota based Advance type fares and regional rovers, as well as PTE-type tickets. Of course, you can end up with anomalies if you mix ticket types.

Clearly if a sensible walk on fare system was introduced in the UK then some fares would go up and some would go down if it was revenue neutral.

it;s the sensible walk up fares system that is impossible in my opinion, due to each companyw anting to maximise its revenue through yield management and market pricing.

I did that because I enjoyed exploring the system. It's utterly unfeasible to expect staff to learn/know about these things and I don't think anyone in their right mind would waste labour time incorporating it as an offical topic in a training syllabus.

I couldn't agree more.

So we've pretty much established that the TOCs have created (and to be fair inherited) a fare structure that makes it impossible to meet their obligation to provide impartial ticketing, even with a supercomputer at every ticket window!
How about short travel? Firstly, my impression is that it would be a lot easier to calculate any savings from that than splits+limitations.
Secondly, what's the thoughts on whether ticket offices should advise on these where cheaper? Except in the case of advances, these will normally involve the same routes and time limitation as the straight A-B ticket, so be no more complicated to use for the customer (although probably harder to understand the logic of!)
An example. Until the last price rise in January, a Chesterfield-Dore ticket was cheaper than a Chesterfield-Sheffield ticket*. As you have to go via Sheffield to get to Dore anyway you might as well get the Dore ticket. If I had gone to the ticket office and asked for the cheapest ticket that would get me to Sheffield, should they have recommended the Dore ticket?

*possibly a very long standing anomoly left from the days when Dore had platforms on the MML

Well, it;s not impossibel to have impartial ticketing, as long as it is limited to providing the cheapest through ticket between the stated origin and destination. To be fair, the idea of impartial retailing was devised to protect consumers from unscrupulous companies who might only sell their own tickets - imagine Virgin at Euston refusing to acknowledge LM fares.

As to the Dore example, agan, it's unfeasible to expect clerks to know every possible place where long or short tickets may save money.

 

[tony_mac]

imagine Virgin's website refusing to sell LM only fares.
Oh, wait.......

I do think that this whole debate is all rather pointless though; they have decided how much they want for e.g. Chesterfield to Sheffield - if they are told to sell it for something else they will just amend that price to what they actually wanted it to be in the first place.

 

[bb21]

I will admit defeat here and now - I'm sure I couldn't tell you whether it's cheapest to get from London to Penzance by getting off at Acton, buying a ticket to Ealing Broadway and so on ad infinitum!

I recognise the difficulty now (being able to consider splitting the journey into as many legs as there are split points) but it shouldn't be necessary if a caveat is added to the "split ticket finder" eg only offer a maximum of 2 legs in each direction.

If the industry is supposed to offer you the cheapest fare - well, it's not even managing that reliably, limiting itself to through fares, given the anomalies we see on these threads every day. So maybe it is expecting too much to look for anything better!
--- old post above --- --- new post below ---

Maybe we need human beings for that!

I think there is a difference between a comprehensive system that works out the cheapest splits logically and one that is scaled down to the barebones, only offering the most basic functions and comparing a very limited set of fares, which appears to be what you're now proposing. I think over the course of this thread the two sides of the argument have gradually drifted further apart focusing on essentially two very distinct things. I don't disagree that if you need something that is a watered down version then it is entirely possible however this would be of no use to the ticket office. What is the point of spending millions implementing a system that is only going to be doing half a job?

To implement the complete system for finding the cheapest split, I don't think it is possible. The figure I quoted does not include any functions being laid on top. Merely having a computer program looping through all these variations without doing anything else on top will take more than a few millions of years.

 

[radamfi]

it;s the sensible walk up fares system that is impossible in my opinion, due to each companyw anting to maximise its revenue through yield management and market pricing.

Fares could be set centrally with TOCs merely bidding to run a specified service. You can still have a franchised rail network without the TOCs deciding fares.

 

[MikeWh]

So we've pretty much established that the TOCs have created (and to be fair inherited) a fare structure that makes it impossible to meet their obligation to provide impartial ticketing

I'm sorry, but you are missing the point about the word impartial. It has nothing whatsoever to do with cheapness or any other consideration, other than the relationship between the seller and the owner of the products being sold. If a ticket window is impartial then it means that it must sell whatever ticket is asked for regardless of who benefits from the sale.

 

[RJ]

So we've pretty much established that the TOCs have created (and to be fair inherited) a fare structure that makes it impossible to meet their obligation to provide impartial ticketing, even with a supercomputer at every ticket window!

Not in the slightest. If you approach a clerk and specify the split ticket you want, they are supposed to sell it to you, according to the rulebook.

 

[hairyhandedfool]

So we've pretty much established that the TOCs have created (and to be fair inherited) a fare structure that makes it impossible to meet their obligation to provide impartial ticketing, even with a supercomputer at every ticket window!....

The obligation of Impartial Retailling is to offer the cheapest ticket from the origin to the destination that is valid for the journey being made (regardless of TOC), unless a passenger/customer requests a specific fare. The railway is actually quite good at that.

....How about short travel? Firstly, my impression is that it would be a lot easier to calculate any savings from that than splits+limitations....

Easier maybe, but not easy, bear in mind you'd have to account for starting short, stopping short, time restrictions, rovers, rangers............

....Secondly, what's the thoughts on whether ticket offices should advise on these where cheaper? Except in the case of advances, these will normally involve the same routes and time limitation as the straight A-B ticket, so be no more complicated to use for the customer (although probably harder to understand the logic of!)....

Well, there is no oligation to do so, but there is no reason staff can't mention it, I would have the same reservations for this as I do split ticketing though, even if it is less complex.

 

[talltim]

I am aware that impartial retailing is not (mainly) about finding the cheapest fare for a passenger, its about selling the cheapest fare regardless of who gets the money.
In my Dore example, the revenue from a Chesterfield-Sheffield ticket would get divided between EMT, XC and Northern. The lower revenue from a Chesterfield-Dore ticket would get shared between EMT, XC, Northern and fTPE. This means that an EMT ticket office not offering this option is not being impartial as it is only offering the fare which EMT makes more money from, and not the one that the customer would be more likely to take up, if they knew of it.
This is my argument with splits. I realise that is imossible to calculate them on the fly in a customer service environments, but again its not about the price the customer pays, its about who gets the money. If you don't offer the (cheaper so more tempting) split where another TOC gets all/more of the revenue, then you are not being impartual.

 

[hairyhandedfool]

I am aware that impartial retailing is not excusivly about finding thecheapest fare for a passenger, its about selling the cheapest fare regardless of who gets the money.

In my Dore example, the revenue from a Chesterfield-Sheffield ticket would get divided between EMT, XC and Northern. The lower revenue from a Chesterfield-Dore ticket woudd get shared between EMT, XC, Northern and fTPE. This means that an EMT ticket office not offering this option is not being impartial as it is only offering the fare which it makes more money from, but that the customer would be more likely to take up, if they knew of it.
This is my argument with splits. I realise that is imossible to calculate them on the fly in a customer service environments, but again its not about the price the customer pays, its about who gets the money. If you don't offer the split where another TOC gets all/more of the revenue, then you are not being impartual.

I dont think you understand how the money is split. I was, however, unaware that TPE stopped at Dore, I thought it was only Northern that stopped there.

Even if TPE do stop at Dore at some point, why should they get revenue for someone travelling from Chesterfield to Sheffield? (the fare is set by Northern)

I get that you think passengers should be offered the lower fare, no matter how impractical it is, I just don't think you realise how complex it actually is to implement that.

 

[talltim]

Just had to double check, but yes EMT, northern and TPE stop at Dore. With regards to your comment about TPE not get revenue between Chesterfield and Sheffield then no, they shouldn't. However they should get a share of the price of the Sheffield-Dore section, in the same way that XC would only get a share for part of the journey.
But I'm not saying that there is a duty to offer the cheapest ticket. I am saying that there is a duty to be impartial when selling tickets where the revenue may go to other TOCs. If you only offer the option that gives your TOC more revenue then you are not being impartial.
The fact that my example was cheaper is mainly irrelevant, excepting the fact that if offered two products that do exactly the same, people are likely to go for the cheapest.

 

[hairyhandedfool]

Being impartial is about not putting the interest of any TOC above any other.

If your arguement is that the fare is cheaper, then as I say, I can see where you are coming from, but if it is because a TOC would benefit from another fare being sold, regardless of price, then I cannot agree.

 

[talltim]

Being impartial is about not putting the interest of any TOC above any other.

And that is what I am talking about.
EMT ticket office. Choice of two tickets. One gives EMT more money and other TOCs less. EMT only suggest that ticket. Impartial?