Credit Scoring techniques revealed

Ritchie74

never-in-doubt wrote: »

Ritchie - hey up matey..... Were your ears burning lol..

Have you moved then, like the new area or miss the old?

Interesting reply.... you stayed within the legal parameters with that... you still fail to amaze me with your knowledge in this field but you never cross the line!

I think we need to get you drunk, spike your drinks and then you'll spill the beans... hehe

Are you coming to the big meet in manc? We can share the train hehe: Who wants to meet up?

lol I'll pass thanks all the same, how very random.

Yeah all moved an settled, missing the liveliness of 'home' but still making the odd night out there. Love where I live now, but can be too peaceful sometimes.

I've been lurking a lot here but tend to only comment on new or different topics. You have the patience of a saint the amount you have to repeat yourself.

Thing with scorecards, as you know, there isn't a 'one-suits-all', even within one institution, there may be many different application scorecards (for diffent products, customer age etc). They are also changed often as populations change/ recessions hit etc etc.
Even if I told you one (which I can't) it wouldn't really help anyone as it would only be true for that product, for that bank, at that moment in time.

Things are set to get tougher soon as the powers that be are enforcing financial institutions to ensure affordability is key in decision making.

Also with the risk of the ban on automated limit increases, the typical model of 'low and grow' limit assignment will be out. So how will banks respond? Keep limits low all the way through a customers life cycle? Give higher limits at app but raise acceptence cut-offs? Rely on customer requesting increases (do they know they can, do they read their statements to see they are eligible). Interesting times.

Ritchie74

unidentified wrote: »

And what is the difference between the report from EX/EQ/CC that banks get tha we get?

Different lenders report to different bureaus, therefore the data each holds can differ from the next.

General info on bureau scores that financial institutions buy:

Scores are proprietory info, just as customers do not know how banks credit score them, banks do not know how bureaus create their scores. Different scores can be bought - new business scores (for application), behavioural scores and indebtedness scores (for existing customer treatments).

As already mentioned, banks create their own application scorecard, however they may choose to feed in the bureau score too.

Many banks also create their own behavioural score. This is usually more predictive (of a customer going 'bad') than the bureau score as it uses internal data and banks know it is relevent to their customers. The bureau score may include, say, mortgage and mobile phone information, and this info may not be relevent to a banks business model for credit cards, for example if a customer is late paying their mobile bill, is that indicative of them defaulting on their CC? That said, a common thing for banks to do is to cross tabulate both internal and external scores by decile (or other) and look for correlation. They are generally highly correlated.

Another thing to think about when looking at behavioral and bureau scores is that they may look at changes in attributes. For eg bal now:bal 6 months ago ratio etc, and not just the flat data as at today.

PROLIANT

Doing some further "related" research today, I have found a website that is in its early days of partnering with several Lenders including;

Black Horse
MoneyWay
GE

You basically fill in an application which is scored using Equifax (Soft search) and you will be given an indication of who will lend to you and at what rate.

The URL for this site is; https://www.http://www.thelendingwizard.com
Read the 'Wise Words' link on the site with reagrd to the ethos of the site and the vision.

hax_2

Here's another link for a credit scoring model called Vantage Score:

http://www.vantagescore.com/about/vantagescore_model

I have to give credit to Cash Flow for this link.

Ritchie74

As izools said, demographics will only play a small part in any scorecard as it's predictiveness is low.

Consider the following hypothetical and simplistic scorecard outcomes:

If you score high on every attribute but you live in an area where >90% people are bankrupt (extreme eg) then you will be accepted.

If a scorecard descriminates against anybody missing more than 2 consecutive payments and you fall into this category but score highly in all other areas you will be accepted.

If a scorecard descriminates against anybody having more than 4 searchesin the past 6 months and you fall into this category but score highly in all other areas you will be accepted.

have missed 2 consecutive payments, made more than 4 searches in the last 6 months and live in the low risk area then you are accepted.

However!
If you have missed 2 consecutive payments, made more than 4 searches in the last 6 months and live in the high risk area then you are declined.

Demographics alone won't mark you out as high risk.

Ritchie74

I'm starting to bore even myself now.
How a scorecard 'looks' in practice...(again, all made up as I sit here in front of 'cops with Cameras')

Constant: 183

total number of revolving cards:
0-1=0
2-4=+8
>5 =-12

etc etc

if score (constant + all attributed scores) >= 280 the accept.
if 260<= score <= 279 then refer.
if score < 260 then decline.

Further...
If score > 340 then Apr= 9.9%, limit = £2000
else if score > 310 then Apr= 14.9%, limit = £3000
else if score > 280 then Apr= 19.9%, limit = £500

unidentified

total number of revolving cards:
0-1=0
2-4=+8
>5 =-12

Interesting. Misread this at first, but I assume it's correct that more than 5 cards give a minus score.

Further...
If score > 340 then Apr= 9.9%, limit = £2000
else if score > 310 then Apr= 14.9%, limit = £3000
else if score > 280 then Apr= 19.9%, limit = £500

This is more interesting though. Don't know if this is just theoretical or actual figures, but it would seem that someone with higher score is being discriminated against?

They have a lower APR yes, but a much lower credit limit than someone with lower credit score? Odd.

Things are set to get tougher soon as the powers that be are enforcing financial institutions to ensure affordability is key in decision making.

Also with the risk of the ban on automated limit increases, the typical model of 'low and grow' limit assignment will be out. So how will banks respond? Keep limits low all the way through a customers life cycle? Give higher limits at app but raise acceptence cut-offs? Rely on customer requesting increases (do they know they can, do they read their statements to see they are eligible). Interesting times.

I don't think an 'automated' limit increase is a big deal, unless you mean by how you are 'expecting' a certain limit on a new card?

If a customer wants a limit increase they usually ask anyway.

All the other signs though point to the fact that I can't wait to escape the credit game. It's all just becoming too much.

Ritchie74

Hi unidentified,

the scorecard was just a made up example. But yes you could be negatively scored if you have many cards, this would depend on the bad rate of existing customers with more than x cards. If the bad rate is greater than the average bad rate of the total population the score would be negative, if lower it would be positive.

Re limit assignment, this could be based on not just risk, but also usage. If customers with high scores tend to spend less on credit, then lower credit limits could be assigned.

On certain books I have seen, for 2 populations with the same limit, the lower risk (ie higher scoring) group have a lower utilisation.
If after a period of time a lower risk customer wanted a higher limit, they would have little problem getting an increase. Lenders don't like to give limits that won't be used as they have to reserve monies to cover the limit. It's not discrimination, it's a logical business model based on stats.

Again this is just an example. Depending on the analysis the reverse could be true.

Do you think, for eg, a low risk customer opening a store card to take advantage of a first spend discount needs a high limit (they probably won't use the card again) and given they are credit savvy, if they did use it, would they rack up a huge balance at >25% APR?

However a low risk customer taking out a prestigious CC may utilise a high limit. Different populations = different business models.
You are basically 'fit' to a profile and 'treated' in a way that stats show that population should be 'treated'.

This is the reason that even if I could I would not show you a real scorecard. They vary between banks/products/populations and moment in time.

A further example, Bank A might conduct analysis that shows customers with >£25,000 total balance across all cards is less likely to go bad than the whole population (feasible as other banks have trusted this customer with high limits and they have not gone delinquent).
Bank B might conduct analysis that shows their customers with >£25,000 total balance across all cards is more likely to go bad than the whole population (feasible as the customer is highly indebted and still requesting more credit).

unidentified

Do you think, for eg, a low risk customer opening a store card to take advantage of a first spend discount needs a high limit (they probably won't use the card again) and given they are credit savvy, if they did use it, would they rack up a huge balance at >25% APR?

Very true, though I always had the personal feeling that storecards only ever benefitted the lenders - as anyone who is money savvy would simply use it for the deal(s) or discount and then cancel it.

I don't know how the lenders make their money from these cards but it must be from those people who spend in store without thinking.

Ritchie74

I don't think the majority of borrowers are savvy- maybe it should be compulsory to join this site ;-) I suppose some people can only get storecards (and mobiles) as a form of credit and if they like clothes shopping (certainly most holders are female) and like a certain store then there is a customer. The certainly do not seem to be price (APR) sensitive, as there are millions of storecard cardholders out there revolving on high APRs.
But yes, lower risk, savvy cardholders 'hit and run' - hence my point as to why one strategy may be to assign this type of customer a lower limit and give a higher one to a customer likely to use it.
This may also explain some limit decrease strategies for existing customers. It's too costly to have under-utilised limits.