AG UT at 0.0% with 2 Cards Reporting - Fico Score Implications

[JLK93]

---

@Anonymous wrote:

Hi TT.   In your discussion of buffers, you write:

"... each of the Fico factor categories is allocated a certain total # of points that sum to 850 which each is fulfilled. Within some categories, depending on scorecard, I believe there is more than one way to reach the category limit."

That's one way that buffers might work (assuming they exist, which I believe they do).

But I think there's another way.  The five categories don't have to sum to exactly to 850.  It's possible that the categories sum to (say) 860 -- and then FICO imposes a final rule which is to render all scores in the range of 851-860 as 850.  Is there any particular reason to believe that the developers wouldn't implement the buffer this way?

Curious to hear your thoughts.

---

I believe that is what TT was saying. Or, at least that is what I thought he was saying.

[Anonymous]

---

@oilcan12 wrote:

---

@Anonymous wrote:

Hi TT.   In your discussion of buffers, you write:

"... each of the Fico factor categories is allocated a certain total # of points that sum to 850 which each is fulfilled. Within some categories, depending on scorecard, I believe there is more than one way to reach the category limit."

That's one way that buffers might work (assuming they exist, which I believe they do).

But I think there's another way.  The five categories don't have to sum to exactly to 850.  It's possible that the categories sum to (say) 860 -- and then FICO imposes a final rule which is to render all scores in the range of 851-860 as 850.  Is there any particular reason to believe that the developers wouldn't implement the buffer this way?

Curious to hear your thoughts.

---

I believe that is what TT was saying. Or, at least that is what I thought he was saying.

---

I don't think so.  Not quite.  He suggested that each category sums exactly to 850.  That means that an absolutely perfect "score" in one of the five big categories won't help offset a less than perfect score in another of the the five categories.

For example, the way TT suggested it might work, an absolutely stellar constellation of "age" related factors (age of oldest account = 31 years, age of oldest CC = 31 years, AAoA = 20 years, most recently opened account = 38 months, etc.) will not help offset a less than perfect performance in (say) Amounts Owed or Credit Mix.

The way TT described it was that only factors within a given category could offset each other.  You still needed an A in each of the five categories.

It's like having five exams in a class.  The question is whether you have to get 5 A's to get a final A (with some wiggle room within each exam score), or whether you can get an A+ in one exam and an A- on another.

The way I described was the second way.  All the sub factors from all groups add up to 860 (say) and then FICO lowers all scores above 850 back down to 850.

I of course have no idea HOW the buffering works.  I was just suggesting that the 2nd approach might be how they do it, with the 1st also being a possibility too.

[Thomas_Thumb]

---

@Anonymous wrote:

Hi TT.   In your discussion of buffers, you write:

"... each of the Fico factor categories is allocated a certain total # of points that sum to 850 which each is fulfilled. Within some categories, depending on scorecard, I believe there is more than one way to reach the category limit."

That's one way that buffers might work (assuming they exist, which I believe they do).

But I think there's another way.  The five categories don't have to sum to exactly to 850.  It's possible that the categories sum to (say) 860 -- and then FICO imposes a final rule which is to render all scores in the range of 851-860 as 850.  Is there any particular reason to believe that the developers wouldn't implement the buffer this way?

Curious to hear your thoughts.

---

CGID, Truncation of total score to 850 only is the generally accepted hypothesis for Classic FICO 08 and Classic FICO 09 buffers.

Nonetheless, the hypothesis of category score truncation/buffers can help explain variability in poster data relative to # cards reporting balances and individual card utilization on score. For example:

1) Some posters with scores in the 700s report score drops associated with an individual card above 30% utilization while others see no score change even when utilization on a card exceeds 70%.

2) Posters with scores in the 700s report a variety of behaviors with Fico 08 score associated with # (or %) cards reporting a balance.

The score sensitivity (or lack thereof) could, in some cases, be explainable by the poster with no  reported score drop having previously exceeded a point limit for that FICO category which then provides a category specific buffer. As we know, category scores encompass multiple factors each of which can contribute (or take away) from the whole.

Either way, I am pretty sure the reason my Classic Fico 08 and Classic Fico 09 scores have stayed at 850, even with shaking things up a bit, is due to buffers of some type.

[Anonymous]

Hi TT.  I definitely agree that buffers appear to exist at the top level.  (Thus your ability to have an 850 and yet also fail to meet each sub-factor perfectly, like number of cards reporting or individual utilization).

Just not clear why that would prvilege one possible "back end" explanation over another (i.e. the example I gave of the five A's being needed for an A vs. three A's, an A+, and an A-). 

But I have no especially keen interest in which is the case (or whether there is some other explanation neither of us mentioned).  The key idea is that it seems possible for a person to score an 850 while manifestly failing to score perfectly on some well known sub-factor).

PS.  I have another question for you that DOES interest me.  Actually a pair of questions.  Got a minute?

[Thomas_Thumb]

CGID,

Sure, what's on your mind? Send me a PM if you prefer.

P.S. The buffer at 850 does not favor one hypothesis over another. It is the variability in scoring associated # cards and UT% with "like profiles" at scores below 850 that suggests the possibility of category buffers.

For example, Masscredit did a test on # cards reporting from 2 of 10 up to 10 of 10 while maintaining Ag UT% quite low. He saw little or no drop in Classic Fico 08 score depending on CRA [No change in TU 08 - see below link]. That is suggestive of a category buffer. We see the same with individual card UT%. Revelate (and others) have allowed individual cards to show above 70% UT without negative impact to score while others report a drop at 50% or even 30%. Sure, scorecard assignment plays a roll but, the variability in score shift occurs with "like profiles" that probably are assigned to the same scorecard.

http://ficoforums.myfico.com/t5/Understanding-FICO-Scoring/Going-to-let-10-10-cards-report-a-balance...

I don't have a vested interest either way.

[Anonymous]

It's about charge cards.

(1) We know that, as a practical piece of guidance, if a person wants to optimize the CC balances on his profile for a big credit pull, a simple rule to give him is to let one card report a small-ish balance and all others report zero.  Would you say that there's enough evidence about charge cards and AU cards that it makes sense to tweak that with:

(a)  The one card reporting should be a true credit card, not a charge card

(b)  It should also NOT be an AU card

(c)  The balance on the card should not be so tiny that it could cause it to be treated as $0 -- practcally then at least $5.  (If I remember right all reports of this happening appear to have been amounts < $3.)

Note that I am talking about practical advice, not theoretical claims.  E.g. we know that the optimization can also be achieved with more cards reporting a balance, depending on how many total CCs the person has, and the balance on the card could possibly be much greater  -- e.g. possibly 8% total util considering all revolving lines -- etc.   But while that is true, it's also the true that it's also achievable by doing all lines zero except one with a small balance.  Thus the "practical" nature of the advice.  I find myself often needing to give practical actionable easy to remember guidance to friends.

(2)  I remember that a bunch of you tried to figure out a while back whether there might be some marginal advantage to the credit mix category in having a charge card along with your credit cards.

(a)  Were you able to answer that with some certainty?

(b)  If there is an advantage, how much does it help?

(c)  Are there charge cards out there with no annual fee?  I might get one if there was a no-AF card, but I know very little about them, except that the ones I have heard of all seem to charge one.

[Thomas_Thumb]

Charge cards are hard to categorize. I used to have a BB "real deal" charge card - no annual fee. But, they converted over to credit cards only back in 2009. Perhaps there are still some store cards that are true charge cards. Store cards do not typically have an annual fee.

The downside with a store card (charge or credit) is the negative impact on credit based insurance scores (CBIS) - at least for LexisNexis model scoring. 

AMEX charge cards are (or can be) classified differently than revolving credit cards. Do they help with mix? Not sure - my mix is classified very good (as opposed to excellent) due to too few accounts in total. I keep my AMEX and pay the grandfathered $55 annual fee primarily because it is my oldest card. It's my only card with a fee.

Regarding the one card only optimization strategy, I would not specify a specific dollar amount as it is overly restrictive and really offers no benefit. The $3 minimum does appear to be valid for avoiding a 0% utilization. That being said, I would add some buffer on the low side and advise reporting a balance between a $5 minimum and a maximum of 5% of the card's CL That would be practical and valid - IMO. Also note, for Fico 08, the card reporting needs to be a revolving credit card that is not an AU card. [AU card ok for Fico mortgage but not Fico 08 - non AU works for all Fico models]

Side notes:

1) I think minimum CL for issued cards these days is $300. so, $5/$300 = 1.7% UT and 5% of $300 = $15

2) If card CL = $20k then => $5/$20,000 = 0.025% and 5% of $20k = $1000. Is $5 enough to avoid a zero balance being reported in this case?

Alternatively you could advise a minimum of $5 or 0.1% of CL, whichever is greater . This would equate to a $20 minimum for a $20k CL account.

I am not a one card only proponent although it has been validated as optimum for Fico mortgage scores. There are some indications that Fico 08 may not be always be optimized with the one card only strategy (data from multiple posts claim higher scores with more than one card reporting) but perhaps the treatment of AU cards and charge cards is clouding the issue.

[JLK93]

If I were to use my best available data, and make my best guess, I would say that the Experian FICO 9 buffer, for my profile, is worth about 40 to 46 points.

My data suggests that FICO 9 only scores 3 inquiries. I currently have 3 inquiries on EX 08, and I am losing approximately 19 points. 

I conducted an experiment recently and let 18 cards report a balance with one card at 42%. My EX FICO 9 decreased from 850 to 846. My Experian FICO 9 auto enhanced score decreased by 15 points. My Bankcard enhanced score decreased by 20 points. That is a difference of 11 points to 16 points.

So, 19 points for inquiries plus 11 to 16 points for the decrease in FICO 9 enhanced scores suggests a 40 to 46 point buffer.

Obviously, this is just a very rough guess, but I suspect it is the best guess I will ever be able to make.

 Interesting, that this would make the unbuffered FICO 9 classic score max out at close to 900. Once again, this is a very rough guess.

[Thomas_Thumb]

Oilcan12, Thanks for the information.

Here's a summary I put together based on looking at enhanced Fico score High/Low fluctuations when Classic Fico scores held at 850. Assuming Classic Fico would have fluctuate like enhanced scores [in absense of a buffer] suggests a H - L difference could be considered a possible quantitative representation of minimum buffers for Classic Fico 08 and Classic Fico 09.  Undoubetedly the amount of Classic Fico 08 and Classic Fico 09 buffer is profile dependent.

In my case it looks like I have a minimum of 20 points worth of buffer with EQ and TU on both Fico 08 and Fico 09 Classic. With EX the minimum buffer appears to be 25 points on both Fico 08 and Fico 09 Classic - based on max point differences.

Please keep the information flowing. Respectful differences of opinion are welcome (if not encouraged).

[Thomas_Thumb]

Possible scenarios on Classic Fico 08 and Classic Fico 09 buffers:

1) Each of the 5 main Fico scoring categories has a specified amount of points. The upper bound limits, when taken together, add to over 850 - let's say 880. Total score is limited to 850 so any totals above 850 are truncated to 850.  In this case there are potentially 30 points of buffer available applied to total score.

2) Each of the 5 main scoring categories has the same scoring potential as above. When summed together the points again would come to 880. However, scores in each category are subject to truncation prior to adding together. In this case the truncated category totals add to 850.

3) Some combination of 1 & 2 above. Say truncated category totals add to 860 and then a further truncation is applied to the total if it exceeds 850

The reason scenario #1 seems less likely is that there would be a lot more profiles scoring 850 (perhaps 10% vs the current 1% to 2%). Scenario #2 requires "pointing out" in each category. However, each category has a buffer so "in category" scoring factors don't all have to be maximized. As mentioned up thread, scenario #2 also provides potential for score stabilization for sub 850 scores on a category basis. 

Scenario #3 has some merit but is more of a stretch IMO. It does offer additional score stability to 850 profiles in the total point buffer zone.