BA's upgrade "formula"
Nov 12, 2018, 5:47 am
#31
Join Date: Mar 2018
Programs: BA
Posts: 55
I am not sure how I have been upgraded. Economy seems quite full on my outbound, Customer care upgraded me when I enquired about a cash upgrade or my booking on website is now only showing ghost booking but with no ticket number. Even my originial ticket number is coming up as invalid. Thank you all and apologies if I am repeating myself.
Nov 12, 2018, 5:56 am
#32
FlyerTalk Evangelist
Join Date: Aug 2002
Location: London
Programs: Mucci. Nothing else matters.
Posts: 38,644
I am not sure how I have been upgraded. Economy seems quite full on my outbound, Customer care upgraded me when I enquired about a cash upgrade or my booking on website is now only showing ghost booking but with no ticket number. Even my originial ticket number is coming up as invalid.
If you were just asking about one, and you did not pay for an upgrade and you decided not to proceed with one, then you should read this thread: Flight upgrade. There's a decent chance that you have not been upgraded at all, and there are some messes in your booking that need to be sorted out. If you simply enquired about an upgrade and you did not proceed with it, there is very little chance that you have been upgraded for free.
When you say "original ticket number", do you mean the 13-digit number starting 125?
Nov 12, 2018, 6:33 am
#34
FlyerTalk Evangelist
Join Date: Nov 2011
Location: Brighton. UK
Programs: BA Gold / VS /IHG Diamond & Ambassador
Posts: 14,194
I'm pretty sure lap infants remove you from the list. When our youngest was below 2 and was a lap infant, we have been on overbooked flights in WTP where others have gone from WT to CW and we have remained in WTP, despite all 3 seated passengers being gold (2 adults and 1 child). I would recommend splitting the party with a lap infant as I subjectively feel that your chances of getting the passengers not linked to the lap infant upgraded will be higher, however slim that may be.
is that an ultimate barrier? No but it would certainly count against an opUp.
Nov 12, 2018, 6:38 am
#35
FlyerTalk Evangelist
Join Date: Aug 2002
Location: London
Programs: Mucci. Nothing else matters.
Posts: 38,644
The experience related on the other thread would, I think, lead me to call anyway.
Nov 12, 2018, 6:42 am
#36
Join Date: Nov 2017
Programs: BAEC Gold, IHG Spire, Inc Amb
Posts: 86
Hi all,
This is a fascinating discussion which keeps on cropping up here. None of us knows for sure the algorithmic approach that BA use for DUT, but as a mathematician who builds models like this for all kinds of industries (OK, these days I have people to build the models for me, but you get my point), here's how I would approach it.
Imagine you (as Speedbird) have over-booked in P class and need to move someone to J. How do you decide? Well first you need to have a decision criteria. An example could be "achieve future J class sale". In order to achieve that you need to upgrade the flyer with the highest likelihood of buying a J class ticket in the future if you upgrade them today. You can find that person by having a mathematical model that looks at *all* the people who you've upgraded in the past year, seeing which of those went on to buy a J fare and how quickly. Once you know that piece of information you then feed in all the data that you knew about them at the time of the first upgrade (i.e. the complimentary one) which would be things like CIV score (and it's underlying components), velocity of travel (how often, how long, classes etc), age, other demographics etc and see how that all correlates to the probability of booking a J class fare after an upgrade.
There's some subtlety in doing this - it could be a simple linear regression, if you had lots of input data items it could be a machine learning tree based model - but essentially what you end up with is everyone in the P cabin getting a score and you upgrade the one with the highest score.
-------
Some quick thoughts / answers to questions
"I got upgraded once but then not again in the following year. I know I was the highest rated person in the P cabin because I was the only one to reserve my seat before 24 hours out. Why wasn't I upgraded?"
A good model is going to include variables that look at the probability of buying a J fare if you've been upgraded once, twice, three times etc. Simply put, one or two upgrades might make you more likely to buy a J fare but a few more upgrades don't increase that probability. The model takes that into account so your propensity to buy a J fare isn't going to be affected by upgrading you *again*
"I've been on quite a few flights where Blue members have been upgraded. I chatted to them and this was their first BA flight in years. Why did they get upgraded?"
The problem with implementing a model like this is that once you start executing it you upgrade those with a high score and that skews future calibrations of the model. For this reason you want to upgrade a proportion of your upgrades on a totally random basis (random to fit your demographics on flights) so that when you recalibrate your model you have an accurate representation of *all* BA customers, not just the ones who travel the most.
------
Hope that helps. In practice it can get more complex. It's possible to optimise the decision across profit, future sales, customer satisfaction etc.
Of course, it could be that BA is running a much less sophisticated operation than this. If that's the case, they need to apply some more sophisticated analytics (and I could help with that!!! :-) )
This is a fascinating discussion which keeps on cropping up here. None of us knows for sure the algorithmic approach that BA use for DUT, but as a mathematician who builds models like this for all kinds of industries (OK, these days I have people to build the models for me, but you get my point), here's how I would approach it.
Imagine you (as Speedbird) have over-booked in P class and need to move someone to J. How do you decide? Well first you need to have a decision criteria. An example could be "achieve future J class sale". In order to achieve that you need to upgrade the flyer with the highest likelihood of buying a J class ticket in the future if you upgrade them today. You can find that person by having a mathematical model that looks at *all* the people who you've upgraded in the past year, seeing which of those went on to buy a J fare and how quickly. Once you know that piece of information you then feed in all the data that you knew about them at the time of the first upgrade (i.e. the complimentary one) which would be things like CIV score (and it's underlying components), velocity of travel (how often, how long, classes etc), age, other demographics etc and see how that all correlates to the probability of booking a J class fare after an upgrade.
There's some subtlety in doing this - it could be a simple linear regression, if you had lots of input data items it could be a machine learning tree based model - but essentially what you end up with is everyone in the P cabin getting a score and you upgrade the one with the highest score.
-------
Some quick thoughts / answers to questions
"I got upgraded once but then not again in the following year. I know I was the highest rated person in the P cabin because I was the only one to reserve my seat before 24 hours out. Why wasn't I upgraded?"
A good model is going to include variables that look at the probability of buying a J fare if you've been upgraded once, twice, three times etc. Simply put, one or two upgrades might make you more likely to buy a J fare but a few more upgrades don't increase that probability. The model takes that into account so your propensity to buy a J fare isn't going to be affected by upgrading you *again*
"I've been on quite a few flights where Blue members have been upgraded. I chatted to them and this was their first BA flight in years. Why did they get upgraded?"
The problem with implementing a model like this is that once you start executing it you upgrade those with a high score and that skews future calibrations of the model. For this reason you want to upgrade a proportion of your upgrades on a totally random basis (random to fit your demographics on flights) so that when you recalibrate your model you have an accurate representation of *all* BA customers, not just the ones who travel the most.
------
Hope that helps. In practice it can get more complex. It's possible to optimise the decision across profit, future sales, customer satisfaction etc.
Of course, it could be that BA is running a much less sophisticated operation than this. If that's the case, they need to apply some more sophisticated analytics (and I could help with that!!! :-) )
Nov 12, 2018, 7:10 am
#38
Join Date: Nov 2017
Programs: BAEC Gold, IHG Spire, Inc Amb
Posts: 86
Nov 12, 2018, 7:55 am
#39
FlyerTalk Evangelist
Join Date: Nov 2011
Location: Brighton. UK
Programs: BA Gold / VS /IHG Diamond & Ambassador
Posts: 14,194
I think the comment was based on BAs IT not being the most sophisticated system in the world and that BA have invested in things like the DUT process than more basic functionality (and correcting spelling errors!)
still it’s better than giving an opup based on how a person is dressed or their birthday or anniversary etc!
still it’s better than giving an opup based on how a person is dressed or their birthday or anniversary etc!
Nov 12, 2018, 9:11 am
#41
Join Date: Nov 2017
Programs: BAEC Gold, IHG Spire, Inc Amb
Posts: 86
It's very possible that there is a threshold for the upgrade score that if no-one reaches they switch to looking for the highest CIV score. If they were really clever they'd be optimising between propensity for future upgrade and customer satisfaction.
Nov 12, 2018, 9:22 am
#42
Join Date: Mar 2016
Location: Plymouth, UK
Programs: BAEC Gold
Posts: 1,159
So, as a couple of Golds who always pay to fly CW (and indeed CE) why are we never given a free taste of First, or even FLub? Don’t they want to try and squeeze even more cash from our pockets? Or do they think I won’t live long enough to make it worth bothering? 
Or do we have a record of flying AA too often?
Or do we have a record of flying AA too often?
Nov 12, 2018, 10:16 am
#43
FlyerTalk Evangelist
Join Date: Mar 2010
Location: JER
Programs: BA Gold/OWE, several MUCCI, and assorted Pensions!
Posts: 32,145
snaxmuppet ... exactly our POV. The benefits don’t justify the extra cost, partly because we’re not champagne experts! For those fixated with that and similar aspects (including fawning CC) I’me sure they have a totally different view. 
All we know is that we won’t fly sub-J any more, but are broadly contect with what we pay for!
All we know is that we won’t fly sub-J any more, but are broadly contect with what we pay for!