Reason code/statement concordance chart

[iv]

UPDATE: version 0.5.2

Source here: https://drive.google.com/file/d/1KawXdpMM1dUJ2Qlaez_qcFHMhYjzBBUB/view?usp=sharing

New features!

Resilience Score Reason Code numbers are now displayed on FRI screen:

Numeric 0-5 Ratings are now displayed on FICO 8 Score Ingredients screen:

Scores without Reason Codes (within 50 points of model max) are now also included in CSV export

Resilience Scores are now included in CSV export (including all prior FRI scores, not just current)

Additional columns in CSV export for numeric Score Ingredients

Available for Firefox in AMO, and for Chrome in CWS. Source available in post #3.

[Anonymous]

@ivat PC about to do this...what do you mean in order by scorecard? how can you tell?

[iv]

---

@Anonymous wrote:

@ivat PC about to do this...what do you mean in order by scorecard? how can you tell?

---

Maybe that was poor phrasing?

It's well-established that for each scorecard, the criteria that assign (or remove, however you want to view it) points are scored differently (and in some scorecards, may not be present at all).

It's also established that once scoring is complete, the four criteria that were furthest from granting maximum points are then displayed as the Reason Codes for the score. IE: each scoring criteria is uniquely identified by a Reason Code, and the "worst" criteria are displayed.

I was observing that the "Positive Factors" displayed by myFICO (and other CMSes that appear to use the same texts) appear to be the flipside of that. IE: it's directly the four criteria from the same list that are closest to granting maximum points, and not just "general score guidance" inspired by the scoring rules. As displayed on myFICO, the Positive Factors have a 5xx number that maps directly to the equivalent xx Reason Code, just with the positive/negative phrasing in the text reversed.

(But the relative weightings of each criteria differ between scorecards, which is why I mentioned the order depended on scorecard.)

[Anonymous]

@iv

EQ-I-26 Number of revolving accounts You have too few or too many revolving and/or open-ended accounts.

All-32 No loan activity You have a lack of recent activity from a non-mortgage installment loan.

classic-505 Few accounts with balances You have a limited number of credit accounts with a balance.

Thats 3 years worth what i cd add!

[Anonymous]

---

@iv wrote:

---

@Anonymous wrote:

@ivat PC about to do this...what do you mean in order by scorecard? how can you tell?

---

Maybe that was poor phrasing?

It's well-established that for each scorecard, the criteria that assign (or remove, however you want to view it) points are scored differently (and in some scorecards, may not be present at all).

It's also established that once scoring is complete, the four criteria that were furthest from granting maximum points are then displayed as the Reason Codes for the score. IE: each scoring criteria is uniquely identified by a Reason Code, and the "worst" criteria are displayed.

I was observing that the "Positive Factors" displayed by myFICO (and other CMSes that appear to use the same texts) appear to be the flipside of that. IE: it's directly the four criteria from the same list that are closest to granting maximum points, and not just "general score guidance" inspired by the scoring rules. As displayed on myFICO, the Positive Factors have a 5xx number that maps directly to the equivalent xx Reason Code, just with the positive/negative phrasing in the text reversed.

(But the relative weightings of each criteria differ between scorecards, which is why I mentioned the order depended on scorecard.)

---

@ivYes agree with everything, but how can we map the relative weightings and determine which are in which cards? Will require help from those in each card. A wonderful resource it would make though.

[iv]

---

@Anonymous wrote:
@iv

EQ-I-26 Number of revolving accounts You have too few or too many revolving and/or open-ended accounts.

All-32 No loan activity You have a lack of recent activity from a non-mortgage installment loan.

classic-505 Few accounts with balances You have a limited number of credit accounts with a balance.

Thats 3 years worth what i cd add!

---

Every additional code helps! Thanks, @Anonymous !

(Find any Resilience Score codes? Make sure you update to 0.5.2, if you installed 0.4.1 earlier today.)

I find Code 32 interesting - you found it on all three?

That makes it just like Code 33 and Code 98 - according to the published docs from FICO, 32/33/98 should not appear with those meanings on TU, only EQ/EX.

• The TU version of 32 "should be" 04. (But you found TU code 32 with matching text to EQ/EX, and TU 32 shouldn't exist)
• The TU version of 33 "should be" 03. (But I found TU code 33 with matching text to EQ/EX, and TU 33 shouldn't exist)
• The TU version of 98 "should be" 97. (But I found TU code 98 matching EQ/EX, and TU code 98 "should be" a CFA code...)

I wonder if this is just myFICO collapsing the matching codes for the oddball ones, or if the "real" FICO codes actually now match up better than the published data? (And if so, will we ever actually see 03/04/97 in the wild or not?)

[iv]

---

@Anonymous wrote:
Yes agree with everything, but how can we map the relative weightings and determine which are in which cards? Will require help from those in each card. A wonderful resource it would make though.

---

Very tricky, though. Maybe not completely possible... we don't have 100% info on the exact scorecard segmentation (and it's different between versions, options, and CRAs).

Even approximating it would take data from a wide range of people, over a wide range of time... combined with specific, targeted testing.

Sadly, scorecard/segment indicators are NOT in the myFICO data... (I did look, though! )

[Anonymous]

---

@iv wrote:

---

@Anonymous wrote:
Yes agree with everything, but how can we map the relative weightings and determine which are in which cards? Will require help from those in each card. A wonderful resource it would make though.

---

Very tricky, though. Maybe not completely possible... we don't have 100% info on the exact scorecard segmentation (and it's different between versions, options, and CRAs).

Even approximating it would take data from a wide range of people, over a wide range of time... combined with specific, targeted testing.

Sadly, scorecard/segment indicators are NOT in the myFICO data... (I did look, though! )

---

@iv yes they were resilience codes, but you already had all of them. Yes segmentation is the same across CRAs for the same version, at least according to fico documentation. And I've got the majority of the segmentation determined, Except some of the specifics for dirty.

No doubt I know the new account and mature segmentation thresholds for all & thick/thin for 9, and it's probably the same for 8, but we would need to confirm for 542. 

[Curious_George2]

Here's all I can currently contribute to filling in the gaps on the chart in post #1.

For Code 02, the MF Short text is "Missed payments". The chart shows this as (All), but I have only seen it on the following: EQ5, EQ AU5, EQ BC5, and EX BC3. There is zero overlap between those bureau-versions and the ones where I have seen Code 18, which has essentially the same meaning. This make me question the (All) designation for 02. I propose a hypothesis that 02 and 18 are mutually exclusive within a given bureau-version.

For Code 06, the MF Short text is "Consumer finance accounts". 

For Code 13, the MF Short text is "Recent missed payment". I have seen this on a wide variety of bureau-versions.

For Code 18, the MF Short text is "Bad payment history". There is a typo in the myFICO Score Factor/Reason Statement (long) column for this one. It should say "You have one or more accounts showing missed payments or derogatory indicators". I have seen this on many bureau-versions -- virtually all of them other than the four listed above for Code 02.

There is a positive reason Code 513. The MF Short text is "Bills paid on time recently". The long text is "You've recently been paying your bills on time".

[Curious_George2]

---

@iv wrote:

---

(Also, just a note - this is only tested and setup for 3B reports. I don't have any 1B reports to test with, so I haven't added support for those.)

---

I'd say don't sweat it and don't bother trying. The 1B reports are in a completely different format. I think it might be the built on the platform MF was using in 2012. I'm guessing little or none of your work up to this point would be applicable to the 1B reports, and it seems like relatively few people purchase them, so the benefit would be small.