Dev Log Week 2025-10: Ratings

[martin]

My work on Individual Travel Requests continued this week, including some changes that are likely going to be a bit controversial. As such, please make sure to let leave a comment below to share your thoughts on this!

So…individual travel requests are just that: Rather than having just three (or four, if you count cargo) static sets of rules for how passengers rate flights, the demand on a given airport pair is split into a whole bunch of individual requests. Each such request has an individual set of preferences and aversions (among other things, but more on those in a future issue). Think of these as the weight a passenger gives to a certain attribute of a given connection, like how important good seats are to them or how much they dislike connections with a lot of stops.

And while these requests are still generated according to pre-defined “passenger types”, those passenger types merely define the ranges into which the concrete preference/aversion values will fall. When a request is generated, we pseudo-randomly pick a value from the given range. Example: While a “no-frills eco passenger” type might generally have a high price sensitivity, one request might have a slightly higher one than another.

This has two main consequences.

The first one is the one for which we’re doing this in the first place: Demand in AirlineSim will become a lot more segmented, meaning airlines won’t necessarily compete for one and the same type of passenger anymore. Instead, they can and have to specialise to some degree. Certain market segments will be larger than others, of course. And in the future, their distribution might differ from place to place. But all in all, this should make it a tad harder to just build a single, massive airline that can soak up 100% of an area’s demand.

The second consequence is the one I expect to be a bit of a hard sell to people who’ve been playing AirlineSim for the past decade or so: With individual travel requests, it becomes rather pointless to show product ratings. Every request will value each attribute of a connection in a different way whenever said connection has been generated. Therefore, the ratings we currently show on flight info pages, in scheduling or in ORS results paint a false picture.

For now, I have implemented two changes:

I have removed all ratings from ORS results. Instead, the results are sorted in so-called “neutral display order”, which is also how the ORS picks a subset of connections for a travel request before rating them by its individual preferences:

On flight info pages (and wherever we currently display detailed product rating matrices) I merely removed the totals (overall product and price performance ratio):

But technically, the latter is only a partial solution: The current ratings are based on the old rating system, which takes into account weights specific to the respective service class as well as the flight distance. The new rating system works in a completely different way, though. Short of not displaying anything at all here (as in: just drop these rating matrices altogether) the only sensible thing I can think of would be absolute product values. So as an example: “On a scale from worst seat to best seat, this is where the offered seat falls”. But I am unsure about whether this would be useful at all.

Please let me know what you think!

[NoodleDoodle]

I agree with removing the ratings as they become obsolete with individual travel requests.

I don’t know if it is technically possible but it could be very helpful if you can create an individual passenger. E.g. there is a slider for every aspect that passengers take into consideration and you can select values for each different aspect. The ORS then could show which flight fits best for that passenger type.

Additionally i think it coudl be beneficial if you can filter the ORS by different airlines to compare yourself with your competitor and not having to search through all the pages to find where your competitor ranks. Maybe you could even show what seats and service they are offering etc but i don’t know how much that goes into the “company secrets”.

[Pounce2KGaming]

So for me loosing the ratings wouldn’t be the worst thing in the world however two thoughts spring to mind. Firstly I still think factors such as staff mood & aircraft ratings should be kept & still have an impact as part of your brand recognition since these factors can be more universal. I mean people will for the most part want to fly on newer planes & you get people who will exclusively prefer airline A over airline B due to xyz.

Secondly I think it is important to have a feature to allow us to quantify this data in some way, even if somewhat generalised so we can understand & monitor where traffic is coming from. Side note more in depth analysis data could be tied into the staffing concept changes if desired. Like a research & development/marketing department. I am sure the airlines probably collect as much booking data as they can through their websites cookies & probably have a good spectrum of booking habits over a longer term period.

I mean it could also be that a generalised overview of what kind of seats & services different classes of passengers on average would expect could also help us to identify roughly where on the market we may want to experiment with & anticipating the potential impact of pricing brackets on those outcomes as well. Yes I know the pricing minefield that Im sure we are all too familiar with

Also additionally would this also require updating how interlining currently works since its using a more simplified method of A to B & transit to C. Or would it already calculate the requests on passenger types & transit requirements as a prerequisite before seeing if an interlining connection is then valid?

Your onboard service is cheap in the image btw XDD

[anderm]

I have mixed feelings about the “segmented” market.
Since there is some significant effort to make the system more realistic, I would expect those segments overlap to some extent.

Let’s say there is an individual travel request from MUC to JFK in F, with no availability in F. Since the passenger still needs to get to JFK some way, he will normally look for C availability with a decent offering, C with sub-par offering and then he could choose to fly via FRA in F or a good C product than flying Y.

On the other end, for airlines with an offering similar to ryanair and spirit, their cliente would certainly pick a discounted full service Y than the sub par offering of low cost, the full service airline Y customer though could even change his itinerary than flying a low cost Y.

I believe that this feature needs to be launched together with the booking classes project in order to make more sense. Martin, if you have some time, it would be great explaining a bit more the logic behind it, yet from what you wrote us so far, I fear that due to the static allocation, there will be travel requests at any time, even at airports with multiple airlines, that will remain unsatisfied.

[Sky1995]

I as a long term player, welcome these changes. The current system was very advanced when implemented all those years ago, but it made a cap on the realism.

We have, in real life, Low cost carriers, Full service carriers and Charter/holiday passengers - in AS, its impractical to build a low cost carrier due to the current ratings. On older servers people are running 130 seat A321s just for ratings and better seats.

[martin]

@NoodleDoodle

it could be very helpful if you can create an individual passenger. E.g. there is a slider for every aspect that passengers take into consideration and you can select values for each different aspect.

I have considered this (and still am), but I am not sure whether I want to go that far as it would make the whole rating system quite transparent. But more filters are definitely on the list (including specific airlines), which should already go a long way towards figuring out the competition.

@NoodleDoodle

Maybe you could even show what seats and service they are offering etc but i don’t know how much that goes into the “company secrets”.

In my mind, the ORS (or rather: DS) interface will look a lot more like, say, Google Flights, in that it presents results in a similar way and at a similar level of detail. This commonly includes stuff like seats, in-flight entertainment options, the general level of food/beverage service etc.

@Pounce2KGaming

I still think factors such as staff mood & aircraft ratings should be kept & still have an impact

They aren’t going anywhere. The first iteration will factor them in in a somewhat simplistic way, but the goal is to add more criteria, not removing existing ones. That said, some aspects might fit better with the Brand and Image ideas on the roadmap.

@Pounce2KGaming

I think it is important to have a feature to allow us to quantify this data in some way, even if somewhat generalised so we can understand & monitor where traffic is coming from

This likely won’t be a part of Individual Travel Requests as such, but ideas in the general direction are part of the whole “DS subject area” (example).

@Pounce2KGaming

more in depth analysis data could be tied into the staffing concept changes if desired. Like a research & development/marketing department.

@Pounce2KGaming

Also additionally would this also require updating how interlining currently works

Not really. Individual Travel Requests already employ the DS internally, but the raw connections are still generated on a per-airport basis (hence the “bulk” in "Individual Travel Request Generation (bulk)") by the good old ORS. So in principle, IL works just as it used to. What happens then is that the DS sorts these raw connections in “neutral order”, then generates so-called “availabilities” for them, which it finally matches to fares. Availabilities are currently based on the 3 fixed service classes we have rather than dynamic inventories . The fares are “faked” by summing up the per-leg prices. Anyway, at the end of that process you have availabilities with fares in neutral order, which are then individually rated and sorted for each travel request. The best match(es) is/are then booked for that particular request. Does that answer your question?

@anderm

I would expect those segments overlap to some extent.

They will definitely overlap. It’s just that within a given market segment (say, “business travel”), there are a bunch of sub-segments with different general preferences. Just because you travel for business doesn’t mean you have unlimited budget, so there might be a general business travel type that values a lower price over traditional “business class perks” and as such would happily fly Economy. At the other end of the spectrum there might be an “executive type” who really values comfort and speed above anything else and will happily fly First Class if available. But both will book whatever business class flight is available (and they can afford) if the market doesn’t offer a lot of choice.

@anderm

I believe that this feature needs to be launched together with the booking classes project in order to make more sense.

Yes and no. Booking classes and Individual Travel Requests are different sides of the same coin, I guess. But they don’t necessarily depend on each other. The (current) service attributes of a flight are very much physical (think seats and service profiles) while booking classes are merely a “virtual” concept that allows more fine-grained control over capacities and fares. The latter - fares - is actually going to have a lot more impact here as it will allow to introduce more “virtual” attributes like, for example, rules for cancellations or baggage allowances which are often the critical (and currently missing) factors deciding whether a traveller will book a given itinerary or not. Long story short, Individual Travel Requests will work just fine with our current set of product attributes. Ideally, it’ll end up leading to more different types of airlines competing for different market segments.

[NoodleDoodle]

@martin

In my mind, the ORS (or rather: DS) interface will look a lot more like, say, Google Flights, in that it presents results in a similar way and at a similar level of detail. This commonly includes stuff like seats, in-flight entertainment options, the general level of food/beverage service etc.

You could compromise by just creating sliders for the “google-factors”. That would create a realistic search for a flight while not making the system too transparent.

[anderm]

@martin

They will definitely overlap. It’s just that within a given market segment (say, “business travel”), there are a bunch of sub-segments with different general preferences. Just because you travel for business doesn’t mean you have unlimited budget, so there might be a general business travel type that values a lower price over traditional “business class perks” and as such would happily fly Economy. At the other end of the spectrum there might be an “executive type” who really values comfort and speed above anything else and will happily fly First Class if available. But both will book whatever business class flight is available (and they can afford) if the market doesn’t offer a lot of choice.

Great. Will this extend to the lower end of the market as well?
A price sensitive passenger that flies low cost mostly, in case there is no low cost airline availability, will choose the lowest fare Y available, even if this is a full service airline?

And will those variables for the passengers be airport specific or will they be the same for all airports as a static allocation?

[martin]

@anderm

A price sensitive passenger that flies low cost mostly, in case there is no low cost airline availability, will choose the lowest fare Y available, even if this is a full service airline?

Up to a point, yes. Each request will have an individual budget. Any itinerary that exceeds this budget will be out, but the remaining ones will be rated according to the request’s preferences, no matter the “type” of the airline.

@anderm

And will those variables for the passengers be airport specific or will they be the same for all airports as a static allocation?

The passenger type distribution is global. What I can imagine for the future is an airport-specific attribute that determines whether the destination is tourism- or business-heavy, or something along those lines.

[FarSmallDuck]

@martin

With individual travel requests, it becomes rather pointless to show product ratings. Every request will value each attribute of a connection in a different way whenever said connection has been generated

Well, the legacy product rating will now indicate how appealing it is to passengers that want a bit of everything, so called “Comprehensive passengers”

This ORS system that you took a screenshot of is handy for budget and speed-oriented airllines, but not for quality oriented and brand oriented ones, since it doesn’t show statistics relevant for such airlines like average seat quality, IFS quality…
I also hope I’ll be able to sort by these different parameters

[martin]

To evaluate “quality”, I think filters will help. As long as we’re talking about physical properties of quality, that is. Think seat pitch, availability of meals, IFE system etc.

All the “soft” properties, like brand, image, perceived quality (like customer service) etc are fuzzy by their very nature. In reality, it would be hard to find a list of connections sorted by “staff friendliness”. As such, I don’t want to over-simplify things that aren’t simple in reality either. Just like in the real world, getting a grasp of these product attributes will likely require dedicated features in the realm of marketing and market research.

[JimmyMuff]

Generally I like the direction the ideas should drive the game.
What I don’t get out of it: How do you think you can enforce that airlines have to specialize?
Do I have to choose what kind of airline I want to found (and therefore limit the seat selection for example)? Or is there a company image (business airline, low cost carrier) that strongly impact the demand (not only the route specific quality and pricing)?
I assume that without further changes a massive airline will still soak most of the demand, just by having aircrafts with different cabin configurations and different pricing (or by founding a subsidiary for that).

[martin]

@JimmyMuff

What I don’t get out of it: How do you think you can enforce that airlines have to specialize?

That’s a really good question

It won’t be easy. But at the end of the day, my plan is for this to be the combined outcome of all the little bits and pieces on the roadmap. Some choices should just be mutually exclusive, others will be more subtle. The overall goal is always to to look at how such things work in reality and then figure out why this is the case.

For example: Network carriers all over the globe have tried, to varying success, to build their own LCC subsidiaries to compete with the new kids on the block. Why were they successful in some cases? Why not in others?

My prime example (due to proximity) is “Lufthansa vs. Ryanair”: Those simply are two completely different business, in almost all aspects. One could never do what the other one does and vice versa. That’s sort of the ideal.

[FarSmallDuck]

While I don’t expect a perfectly accurate rating for the soft qualities, I still think we should provide the airlines who care about it a rough idea about their status