Originally Posted by
flyingcrooked
I'm not sure from AC's POV it makes sense to reward FFs who pay the most for a flight. Suppose you and I both fly ABC-DEF once a month, and AC's system is sophisticated enough to know the following:
- you paid $2000 for your latitude ticket, J is $500 more, and if you don't get upgraded on this route then darnit, next time you are buying J
- I paid $1500 for my latitude ticket, flex would have been only $700, and if I don't get upgraded on this route then darnit, next time I'm buying flex
Should AC give you the upgrade? If it's about rewarding loyalty, yes. If it's about maximizing revenue, no.
I doubt AC's system is making as fine-grained predictions at that, but it's not implausible that the algorithm is tailored beyond just the cost of the fare, fare brand, and status. If the computer gets enough data, it should start making predictions like the one above, where giving the hidden upgrade to the passenger on the lower fare is more likely to maximize revenue.
The other factor that exists here is that we don't know exactly which outcome the revenue management system is actually biasing for. The notion of 'rewarding' your frequent flyers may be secondary to the entire problem.
For example, the default thinking suggests that the outcome being optimized is maximizing revenue in the J cabin. If that's the case, then allowing an upgrade when you think you could sell a seat is bad because you forego extra revenue. That is probably stereotypically where our thought pattern goes.
However, from an outcomes perspective, having a half empty J cabin that clears upgrades at the gate can be bad as well. Particularly on routes with a high Y load, doing so means that you generated less Y revenue by not upgrading those passengers in advance and selling the extra Y seats.
To some extent, this is solved through overselling Y but I believe we are seeing less of that because both the compensation cost of doing so has increased and there is less excess capacity in the system overall so it's harder to accommodate oversold pax with low incremental cost.
The real question the revenue management system has to answer is if on a probabilistic basis, the J seat at a higher price point will generate more actual revenue or the Y seat at a lower price point will. If the answer is the latter, then someone needs to be moved from Y to J to open up another seat to sell there and the question with respect to the upgrade changes from 'if' to 'who'.
My thinking is that the opaque part of the upgrade strategy is a mechanism to provide fine optimization in these types of edge cases. The error bands in the projective equations for J vs Y maybe don't yield a conclusive result so you wait to see what sells and make a decision once the revenue is in hand.