Professions that throughly understand FICO scoring?

[Aahz]

---

@iv wrote:

---

@masscredit wrote:

Thank you for the detailed reply. I thought our scores are like being graded on a test in school. A person would get a good grade if they understand what they were taught.  When it comes to credit, a person should have the knowledge of what they are supposed to do then do it ( if they can).  They would have a good score because of that.  Many people know the basics - pay bills on time and don't be over extended.  And then there are people that take that a step further.  They micromanage their accounts, watch utilization and toss in a loan here and there to maximize their scores. A lot of tests are done to figure out what works and what doesn't.

i understand your description and understand that scores are a business. Consumers usually have to pay to get them so EQ, TU and EX make money.  I don't understand why they don't give consumers greater knowledge.  Someone can give one card a lot of use and not receive a CLI for years while another card will hardly be used and receive an increase every 3-4 months   There is a deeper layer when it comes to credit.  Maybe I'm wrong but I think we should know about that like we know that our bills should be paid on time. 

---

This does seem to be a common misconception...

A scoring rubic for grading in a school (at least in theory...) is based on clear, documented goals (ideally spelled out in advance of testing) - and mastery of those goals results in good grades. (Yeah, yeah, it doesn't always work that way in all schools... but it's meant to.)

Credit scoring... isn't like that at all.

Something I posted in a different thread last month is fairly relevant here:

"Scoring models (and reporting retention) aren't designed with "fair" as a main criteria.

They are designed to be statistically predictive across lenders' customers bases.

Most of the "if I was building the model" posts I see seem to treat it like building an elementary-school grading rubric: do a/b/c to demonstrate mastery of x/y/z, and receive an "A".

But reporting/scoring isn't like that.  It's more: identify the slice of customers who will return $X profit at X% interest, the slice who will return $X profit at Y% interest, the slice who will return $X profit at Z% interest, and the slice to avoid.

The reporting/scoring doesn't care about you. Just about the statistical group you are in, and what terms to offer that group for best profit."

Notice that nothing in there provides a motivation for lenders, CRAs, or FICO to to provide specific knowledge targeted at increasing an individual's score. The only real business motivations are to decrease delinquencies and increase returns... (Some seem to think that this is some deep, dark secret conspiracy... but it's just business, and hardly a secret.)

---

If I had a bushel barrel full of kudos to hand out I'd simply slide the whole thing over to iv for this post!

I repeatedly post this same concept in numerous threads here, but that's one of the best wordings of the idea I've seen!

[grillandwinemaster]

---

@masscredit wrote:

I was thinking, it must be nice to work deep in a profession that throughly understands FICO scoring. Much deeper than what we know/try to figure out. A level that understands what it takes to maximize CLIs and scores. What would some of those professions be and will anyone here admit to being at that level?

---

A couple of occupations come to mind, 

"Mind reader, crystal ball operator, forensic scientist..."

[Anonymous]

---

@iv wrote:

---

---

This does seem to be a common misconception...

A scoring rubic for grading in a school (at least in theory...) is based on clear, documented goals (ideally spelled out in advance of testing) - and mastery of those goals results in good grades. (Yeah, yeah, it doesn't always work that way in all schools... but it's meant to.)

Credit scoring... isn't like that at all.

Something I posted in a different thread last month is fairly relevant here:

"Scoring models (and reporting retention) aren't designed with "fair" as a main criteria.

They are designed to be statistically predictive across lenders' customers bases.

Most of the "if I was building the model" posts I see seem to treat it like building an elementary-school grading rubric: do a/b/c to demonstrate mastery of x/y/z, and receive an "A".

But reporting/scoring isn't like that.  It's more: identify the slice of customers who will return $X profit at X% interest, the slice who will return $X profit at Y% interest, the slice who will return $X profit at Z% interest, and the slice to avoid.

The reporting/scoring doesn't care about you. Just about the statistical group you are in, and what terms to offer that group for best profit."

Notice that nothing in there provides a motivation for lenders, CRAs, or FICO to to provide specific knowledge targeted at increasing an individual's score. The only real business motivations are to decrease delinquencies and increase returns... (Some seem to think that this is some deep, dark secret conspiracy... but it's just business, and hardly a secret.)

---

Hmmmm, I only have basic statistical software, but I kinda want to keep data on every single change in my reports and the scores and then play with correlations.  Not that I will discover any deep dark secret, but i can pretend I know what I am doing, and create pretty charts.....

[Aahz]

---

@Anonymous wrote:

---

@iv wrote:

---

---

This does seem to be a common misconception...

A scoring rubic for grading in a school (at least in theory...) is based on clear, documented goals (ideally spelled out in advance of testing) - and mastery of those goals results in good grades. (Yeah, yeah, it doesn't always work that way in all schools... but it's meant to.)

Credit scoring... isn't like that at all.

Something I posted in a different thread last month is fairly relevant here:

"Scoring models (and reporting retention) aren't designed with "fair" as a main criteria.

They are designed to be statistically predictive across lenders' customers bases.

Most of the "if I was building the model" posts I see seem to treat it like building an elementary-school grading rubric: do a/b/c to demonstrate mastery of x/y/z, and receive an "A".

But reporting/scoring isn't like that.  It's more: identify the slice of customers who will return $X profit at X% interest, the slice who will return $X profit at Y% interest, the slice who will return $X profit at Z% interest, and the slice to avoid.

The reporting/scoring doesn't care about you. Just about the statistical group you are in, and what terms to offer that group for best profit."

Notice that nothing in there provides a motivation for lenders, CRAs, or FICO to to provide specific knowledge targeted at increasing an individual's score. The only real business motivations are to decrease delinquencies and increase returns... (Some seem to think that this is some deep, dark secret conspiracy... but it's just business, and hardly a secret.)

---

Hmmmm, I only have basic statistical software, but I kinda want to keep data on every single change in my reports and the scores and then play with correlations.  Not that I will discover any deep dark secret, but i can pretend I know what I am doing, and create pretty charts.....

---

Be sure to share your results with all of us!  

[Anonymous]

We will never really know.  Just pay your bills on time.  NEVER 1 day late.  Always pay at least twice the minimum or pif before the due date, never carry more than 30% on one card, ideally you should carry 1% on just 1 card.  Keep lowering your ut on your auto loans, student loans, mortgages and such to get at least below 60%.  And the older your AAOA is, the better.   No inq's would be ideal.  If you play by these rules, you should be able to reach above 800 before long.  Many here have proven it can be done.  Just a matter of time for some of us.  

[masscredit]

---

@Anonymous wrote:

Keep lowering your ut on your auto loans, student loans, mortgages and such to get at least below 60%.

---

I thought one of the magic numbers is under 70%?

[Revelate]

---

@masscredit wrote:

---

@Anonymous wrote:

Keep lowering your ut on your auto loans, student loans, mortgages and such to get at least below 60%.

---

I thought one of the magic numbers is under 70%?

---

I don't think the high point is nailed down yet unlike the magic number of 10%, the my little pony one hasn't been well tested since most people just pay straight down to the low number and off they go.  I know from my own data that somewhere between 82.93% and 64.67% exists said my little pony line.

Hoping SJ gets that for us actually next month or two with his auto loan paydowns.

[masscredit]

My auto loan will be reporting 69% at the end of August. We'll see what happens. 

[Anonymous]

---

@RobertEG wrote:

Regardless of occupation or knowledge of the consumer, the FICO scoring algorithms are legally protected as trade secrets, and Fair Isaac is careful not to disclose details that would enable one to fully understand and thus duplicate the algorithms. 

They are not protected by patents, and thus if known, can be freely used.

Thus, the assumption that certain parties could, based on education or profession, thoroughly understand the workings of the FICO algoritms is not the case.

---

I respecfully disagree, although I do concede your statement is in fact correct.  I am someone of belief that with enough data any problem can absolutely be solved.  So for high level mathematicians and data scientists I do believe they could just about nail it.