Actual demand randomness
While demand randomness adds more of a real world feel, it serves no useful purpose in the Sim, it denies / clouds player feedback, makes it unnecessarily more difficult for players to become better players.
Fore example if the actual demand for Wednesday is 120 pax, unless there is a global growth or contraction of the demand, the demand should be the same 120 pax the next Wednesday. If the route is served by a single airline, using the same aircraft, same time, same CI, same RI, it should be allocated the same pax on every Wednesday, unless something changes.
Players should be encouraged to become better players by experimenting and receiving feedback. Small changes, positive or negative should produce direct effect, giving player a feedback if he is doing the right thing or wrong thing. By receiving feedback, the player would learn how to play, how to run the airline better.
Unfortunately, as is, the randomness of the demand completely wipes out the feedback of small changes. Unless a player keeps spreadsheets that would have some 10 values of before the change, 10 values after the change, averaging the result, that may let the player know if the change was good or bad, assuming nothing else really changed over the period of 20 weeks.
Fat chance, no one does that, everybody is more or less flying blind, guessing what the optimal ways are, or going by some AWS urban legends.
Will it help me to fly premium seating? What if I fly premium seating and rise prices 5%? As is, no one really knows, and it is nearly impossible to find out. The noise of randomness erases the results of these small changes. Even the most experienced players don't know for sure. Newbs? Impossible for them to know, and they will not really find out by experimentation, because the experimentation necessitates removal of the randomness, and the removal of the randomness (by averaging a number of results, over a very long period of time) is just too difficult and too time consuming. Very few Newbs are really going to do this.
Ok suppose I really want to set up an experiment and do it right. It takes 20 weeks to collect 2 x 10 values, and even that may be affected by things such as an aircraft on the route going into a check. So it is really an impossibility...
Pax allocation randomness
I am not sure if the pax allocation introduces new randomness on top of randomness of demand. If it does, it shouldn't for the same reasons as above.
Actual Demand Estimate
When a player clicks on the route, he will receive the estimate for demand of the route, with certain accuracy, with randomness added into this. Sure, this exists in real life, before the route is flown demand can only be estimated.
Regardless of the above, suppose I want to find out what exactly the demand is? Nothing is stopping me from refreshing 100 times, and averaging the 100 values. Despite the levels of confidence and randomness, I WILL get the accurate demand. If each estimate needed market research and cost player $10,000, maybe there is some purpose to it. But the clicks are free, needlessly burden the server, and are just time consuming.
Why not just display 100% accurate demand of the route on the first click for the player? What is gained by not doing so?