Fresh Air/Availability of Advance Fares

[modernrail]

I have been on a few journeys recently where no advance fares were available yes there were a lot of empty seats. Particularly GWR and Avanti. Today I was the only person sat in a carriage from Totnes to London. It seems like an incredible waste of capacity.

Is there any particular consistency as to how TOCs work out how many advance fares to offer or are they all using individual methodologies? Are they getting less bothered about offering good value advance fares to eat up excess capacity since the change to franchising arrangements that means they are not on the hook for revenue?

 

[Tazi Hupefi]

10 people paying £20 or one person paying £200 results in the same revenue yield for the TOC - with the added benefit that the person paying the premium £200 gets a nice, quiet train and is more likely to pay that again.

Ultimately, the TOC revenue managers now mostly have a target yield £ per train - how a TOC slices that up depends on so many factors, both societal and economic (competition etc). Flooding a specific train with cheap Advances works in some instances, but it's often far more nuanced than that. AI is starting to be introduced to modelling and assessing demand dynamically - e.g. a concert/festival pops up somewhere on a future date - the AI solutions now automatically spot it and potentially identifies that a different yield profile needs to be applied to services heading that way etc, which then adjusts what products are made available, and at what price point.

Of course, some trains will simply never be in demand - and they only run because the DfT says so. There's plenty where you could make them free and it still wouldn't have a huge impact on demand.

 

[modernrail]

10 people paying £20 or one person paying £200 results in the same revenue yield for the TOC - with the added benefit that the person paying the premium £200 gets a nice, quiet train and is more likely to pay that again.

Ultimately, the TOC revenue managers now mostly have a target yield £ per train - how a TOC slices that up depends on so many factors, both societal and economic (competition etc). Flooding a specific train with cheap Advances works in some instances, but it's often far more nuanced than that. AI is starting to be introduced to modelling and assessing demand dynamically - e.g. a concert/festival pops up somewhere on a future date - the AI solutions now automatically spot it and potentially identifies that a different yield profile needs to be applied to services heading that way etc, which then adjusts what products are made available, and at what price point.

Of course, some trains will simply never be in demand - and they only run because the DfT says so. There's plenty where you could make them free and it still wouldn't have a huge impact on demand.

There is of course the option of say 8 people paying £60 which is clearly better for passengers and railway.

Interesting on the AI. I suppose it was bound to creep in and encourage even less predictability in fares.

 

[Haywain]

Interesting on the AI. I suppose it was bound to creep in and encourage even less predictability in fares.

I don't know about AI but LNER have been using a similar system to budget airlines for many years.

 

[Tazi Hupefi]

The main revenue management systems in the industry are currently:

Moving people forward | Sqills

www.sqills.com

Extends the inventory features of S3 Passenger with the capacity to automatically intervene per departure based on analysis of the incoming booking curve. Offers a powerful set of filter and actions based business rules that can modify bucket level booking limits or raise warnings to revenue analysts

Ventoux – Mistral Data

www.mistral-data.com

A Revenue Management System designed specifically for the needs of both London centric and regional routes, it sets the price for advance purchase tickets to maximise revenue obtained from the available capacity whilst limiting overcrowding on busy segments.

I doubt many revenue managers even know why some trains are priced the way they are - the "algorithms" presumably lead the way, and as long as the yield is good / meets the revenue target - I doubt they're going to intervene and override whatever the system is suggesting as a strategy for a particular service.

Both systems do provide warnings though when unusual booking patterns are spotted - perhaps the system has not reacted to an event or a competitor slashing prices etc.

 

[nwales58]

Even assuming the only objective is to maximise yield without causing overcrowding, I suspect there is not enough experienced mind-power around in every TOC nowadays to achieve it. We have more TOCs than low-cost airlines, remember.

It's easy to think you, or the expensive software you have bought to replace people, are maximising yield but it needs regular systematic experiments on quotas, and that needs expensive staff, to check the elasiticities and be sure.

From the obvous mismatches between capacity and availability of advance fares, which is only part of the story as I cannot see the price elasiticites, my impression is that on Cross Country, Tfw, and probably Avanti, someone set up a formula a few years ago and mostly the buckets are set automatically despite new evidence.

Apologies to anyone still working for a TOC on this who actually knows what they are up to and has the time and resources to do it!

It is screaming out for a cross-TOC approach, one unit for inter-city routes, one for regional, and no-one allowed to pursue Northern's 'set a low price and hope it's increasing total industry revenue because we haven't a clue' approach.

== Doublepost prevention - post automatically merged: 16 Jul 2024 ==

PS any mention of AI nowadays is marketing hype, probably distracting you to pay more for the algorithms than you did last year. Ask the seller to explain how it was trained and prove it achieves a well-defined objective. They can't.

 

[Tazi Hupefi]

Even assuming the only objective is to maximise yield without causing overcrowding, I suspect there is not enough experienced mind-power around in every TOC nowadays to achieve it. We have more TOCs than low-cost airlines, remember.

It's easy to think you, or the expensive software you have bought to replace people, are maximising yield but it needs regular systematic experiments on quotas, and that needs expensive staff, to check the elasiticities and be sure.

From the obvous mismatches between capacity and availability of advance fares, which is only part of the story as I cannot see the price elasiticites, my impression is that on Cross Country, Tfw, and probably Avanti, someone set up a formula a few years ago and mostly the buckets are set automatically despite new evidence.

Apologies to anyone still working for a TOC on this who actually knows what they are up to and has the time and resources to do it!

It is screaming out for a cross-TOC approach, one unit for inter-city routes, one for regional, and no-one allowed to pursue Northern's 'set a low price and hope it's increasing total industry revenue because we haven't a clue' approach.

== Doublepost prevention - post automatically merged: 16 Jul 2024 ==

PS any mention of AI nowadays is marketing hype, probably distracting you to pay more for the algorithms than you did last year. Ask the seller to explain how it was trained and prove it achieves a well-defined objective. They can't.

I think you're a bit too cynical on the AI front - it's now fairly cheap and accessible for even "normal" people, and fairly straightforward to integrate into revenue management for the purposes I outlined - looking for events, competitors pricing and deeper analysis etc that would be time consuming for a human, and I'm quite sure it does occasionally spot things that a human would probably have missed. It is still being applied in a fairly limited way, but there's certainly a fair amount of interventions that are solely down to automated algorithms (call it AI, pattern matching etc, or whatever you prefer).

Given commercial rail datasets are a fairly well known quantity, and the data is reasonably reliable - I'm fairly sure that AI being trained on it would perform quite well.

There is a lot of technology that you wouldn't even spot - Avanti has clever and extremely reliable new cameras that detect and report seat occupancy throughout a trip, so you know exactly how a carriage was occupied/used. The scan data from e-tickets and smartcards is also another extremely valuable source, and presumably why RMT union want the Avanti staff to be paid extra for scanning tickets. These are all GPS and time stamped too.

Network Rail has AI overlaid on their station CCTV that analyses your emotions and expressions and uses it to analyse various scenarios/events.

 

[ac6000cw]

I have been on a few journeys recently where no advance fares were available yes there were a lot of empty seats. Particularly GWR and Avanti. Today I was the only person sat in a carriage from Totnes to London. It seems like an incredible waste of capacity.

Of course, some trains will simply never be in demand - and they only run because the DfT says so. There's plenty where you could make them free and it still wouldn't have a huge impact on demand.

Of course there also good reasons why some trains run very lightly loaded for operational reasons - e.g. to get them to somewhere to form another service or to the correct place for overnight maintenance, as part of each train's movement/service 'diagram' (which might stretch over multiple days).

If there's prospects of generating enough revenue to justify paying the cost of a conductor/guard/TM for lightly-loaded service, then that might be better than running it as a driver-only empty stock train, but there's no point in offering discounted fares if it doesn't generate extra revenue overall for the train (and doing that might actually generate less revenue).

 

[Watershed]

Of course there also good reasons why some trains run very lightly loaded for operational reasons - e.g. to get them to somewhere to form another service or to the correct place for overnight maintenance, as part of each train's movement/service 'diagram' (which might stretch over multiple days).

If there's prospects of generating enough revenue to justify paying the cost of a conductor/guard/TM for lightly-loaded service, then that might be better than running it as a driver-only empty stock train, but there's no point in offering discounted fares if it doesn't generate extra revenue overall for the train (and doing that might actually generate less revenue).

In general, since Avanti now has driver and guard depots in all the same places, it would rarely be worthwhile for a train to run ECS when it could run in passenger service. Frustratingly Avanti do still do so on some occasions (e.g. for some trains that terminate at Crewe but then run ECS to Manchester, Liverpool or Wolverhampton).

 

[nwales58]

I think you're a bit too cynical on the AI front ...

There is a lot of technology that you wouldn't even spot ...

I butt in on these discussions in the hope, naively, that fewer mistakes could be made in the future than in my working life.

Now long retired but 35 years experience in and out of transport demand forecasting and modelling here. I have seen the same cycle repeatedly: appreciation that existing methods are poor, new technology (maths, models or computing power) will improve things. Rising optimism. Then increasing evidence from post-opening evaluations that the new methods are poor. The Gartner hype cycle.

The current optimism about AI (some/much of which is not AI, as you say), following Big Data 10-ish years ago, reminds me of the micro-simulation boom (pretty visualisations sell well even when the network model has blunders). And demand estimation from novel sources. And the integrated models boom decades ago.

If you can't evaluate what a system is doing against clear objectives it's pouring money into salesmen's bonuses on faith. I worked in consultancies, I've seen both sides. Validating a genuine AI system seems difficult, and will become impossible if you lose the real intelligence to update it's training.

Are we trying to maximise revenue? If so, with the fall in inelastic business traffic, we might be experimenting with *reducing* some of the 1 pound per mile open singles (to see whether more people trade up for flexibilty) as well as checking the toe and heel of the yield curve.

Maybe (insert intercity TOC) are doing mostly well (fresh air is unavoidable at times), but maybe not (pricing onto overcrowded services).

But we'll probably waste money on shiny new toys, hope over experience, as we have done for decades. Some of it may be useful but we may not know which bits.

 

[Cloud Strife]

Interesting on the AI. I suppose it was bound to creep in and encourage even less predictability in fares.

AI is only as good as the humans designing it, and being totally honest, many of these systems are still incredibly dumb and not doing anything particularly special. In fact, they really can't be called AI at all.

A friend is the CEO of a company that has some truly frightening software, as it uses cameras and other technology to predict when a loss might occur in retail. The system isn't quite there yet, but in testing, they're trying to make correct predictions involving theft. Essentially, the system is trying to spot suspicious behaviour and sending alerts to security before the person has even taken any goods, and once the goods are taken (and hidden), the system automatically detects this and tracks the person so security can be waiting at the exit. The system already has them recorded and confirmed to be in the process of stealing, meaning security can detain them while knowing for sure that they're not risking charges of unlawful detention.

 

[alistairlees]

I butt in on these discussions in the hope, naively, that fewer mistakes could be made in the future than in my working life.

Now long retired but 35 years experience in and out of transport demand forecasting and modelling here. I have seen the same cycle repeatedly: appreciation that existing methods are poor, new technology (maths, models or computing power) will improve things. Rising optimism. Then increasing evidence from post-opening evaluations that the new methods are poor. The Gartner hype cycle.

The current optimism about AI (some/much of which is not AI, as you say), following Big Data 10-ish years ago, reminds me of the micro-simulation boom (pretty visualisations sell well even when the network model has blunders). And demand estimation from novel sources. And the integrated models boom decades ago.

If you can't evaluate what a system is doing against clear objectives it's pouring money into salesmen's bonuses on faith. I worked in consultancies, I've seen both sides. Validating a genuine AI system seems difficult, and will become impossible if you lose the real intelligence to update it's training.

Are we trying to maximise revenue? If so, with the fall in inelastic business traffic, we might be experimenting with *reducing* some of the 1 pound per mile open singles (to see whether more people trade up for flexibilty) as well as checking the toe and heel of the yield curve.

Maybe (insert intercity TOC) are doing mostly well (fresh air is unavoidable at times), but maybe not (pricing onto overcrowded services).

But we'll probably waste money on shiny new toys, hope over experience, as we have done for decades. Some of it may be useful but we may not know which bits.

Yep. AI is somewhat overhyped here and is not a substitute for a strategy or a close unerstanding of customer demand and opportunities. But it is the buzz of the moment, so everyone will want to be seen to be using it (and spending money on it).

 

[Krokodil]

Even assuming the only objective is to maximise yield without causing overcrowding

TfW seems to have a policy of throwing out dirt-cheap advances like smarties, even on services routinely reported as "full, standing & leaving pax behind".