Utilization %s

[Anonymous]

---

Logical wrote:

---

cheddar wrote:

I have experienced the effect of tiers firsthand.  I got nothing for going from 14% to 12%, but got a nice little boost for going from 12% to 9%.

---

First, I'm not saying there are not tiers, just saying it makes no sense that there are. They have probably 100's of millions of calculations on the risks of the entire range utilization. They don't have to have ranges, they know the risk at every percent.

That said, your experience does not prove tiers because a curve could do the same thing. The slope of the curve could be steeper around 10% than 13% and that would easily explain it. Plus, is it really possible to isolate one factor? I've seen my scores go up and down and not having a clue what caused it. Something did! Perhaps there are tiers, but I doubt it.

---

You make some good arguments, but I do have a couple of points to make.

I agree that my experience alone does not prove anything.  I will also concede that the vast majority of evidence for the existence of tiers is anecdotal in nature, but that evidence is all over the place.  Many, many others have reported experiences similar to mine.  After months and months of reading post after post offering anecdotal evidence of the existence of tiers, I've come to the conclusion that the existence of utilization tiers is a safe assumption, even though we have no concrete proof one way or another.  To believe that the risk factors for utilization are calculated on a more granular level is to ignore the experiences of dozens and dozens of forum members with evidence to the contrary, in spite of the fact that, yes, it could be calculated down to the individual percentage.  It just doesn't appear to be the case.

Actually, a great deal of what we "know" about FICO scoring is based on nothing more than the experiences of our forum members.  How do we "know" that TU will slap you if you have half or more than all of your open accounts reporting a balance?  Well, we know that one of the negative factors on page 2 of myFICO TU reports is "Too many accounts with balances."  The next question is then one of how many accounts with balances are too many.  The only way we have discovered an answer to this question is when a number of people reported that the magic number appears to be fewer thanhalf of all CCs and fewer than half of all open accounts.  What we "know" about utilization is no different.  We don't really know, but the vast majority of evidence tends to point in one direction.

Regarding the question of isolating one factor, I used to ask the same thing.  I used to seriously question posts where members reported that action x caused a gain/loss of y points.  That was back when I was pulling a single report and score every couple of months.  These days, though, I know all of my account details practically by heart, I pull my reports and scores frequently enough that I can isolate the changes with some degree of certainty.

For example: Here's what I know about my EX report in just the last three weeks.

Accounts with balances: 4.  FICO 730

Accounts with balances: 3.  FICO 737

Accounts with balances: 2.  FICO 741

New inquiry and new account: FICO 732

This is all in three weeks time, and all on EX.  Not much else changed from one report to the next, except the factors mentioned.  It's pretty safe to say the score changes were at least partially affected by those factors.

[Logical]

cheddar,

Perhaps I should explain my bias. I programmed databases in the past and I'm good at math so I know how easy it would be to use curves. Plus, what about the accuracy factor? A curve would be much more precise. Just for the record, it would be a curve, not a line. A curve would have different slopes at different places.

The percent of cards showing a balance can also be a curve. The slope may accelerate positively in the 40% area which would mean 50% is less good. TU probably has a programmed trigger that offers that warning when card usage hits 50%. I just checked yesterday's TC report and 76% of my TLs show a balance and my scores are trending up.

I have my own antidotal stories. I monitor TC and Credit Secure. Yes, I know they are facos, but they do show trends. The monthly WAMU scores go up and down with TC and CS. Last month one-by-one I noticed the 3 CRA scores went up several points. They happened when my bank line of credit went from 100% to 0% utilization.

First of March my average TC scores jumped 25 points and WAMU pfico jumped 25 points. I had 3 new cards in March 2007. March 2008 they stopped being "new credit" and became "old credit." Now that is a tier! I'll give you one!   July 2007 I had 7 new cards, so in 3 months Fico should go through the roof! (I hope)

The only way to know for sure is if FairIsaac releases the information on scoring. I don't think they should. I plan on doing what is in my best financial interest. If it is best for Fico, fine, if not . . .  Micromanaging Fico before a mortgage is prudent, otherwise perhaps not. Know the trends and aim that direction.

[MidnightVoice]

The other thing about a curve is that small increments would not necessarily trigger a point change - they could be too small, and the minimum point change is 1.  So a curve is in effect a stepped line, with vertical rise triggered by a change sufficient to give a one point score change

[Logical]

---

Logical wrote:

First of March my average TC scores jumped 25 points and WAMU pfico jumped 25 points. I had 3 new cards in March 2007.

---

I have to correct myself already. The new credit was in Feb 2007 not March. First of Feb 2008 TC went up but pfico went down 6. April 1 pfico just appeared and it's up 18 points. The only change is the last 2 inq went off. So, I don't know - util is high, percent of cards with a balance is up and pfico keeps going up. And tax season is over.

[haulingthescoreup]

My highly anecdotal evidence also comes from the opposite experience: how many posts have we all read from someone in tears or in a rage (or both), because they paid their balances down a thousand bucks, or something, and their scores didn't change at all? After some gentle questioning, it usually comes back that they only dropped from 64% to 52%, or from 29% to 22% (I'm completely making up numbers here).

The logical explanation is that despite the impressive reduction in both the dollar value of the balances and in the util, it didn't go over some unknown threshold that would create a score change.

And Logical, I completely hear what you're saying about the difficulty in separating out on the influences on our scores. We spend a lot of time navel-gazing on this, and we are frequently proved wrong. I just don't think that FICO or the lenders saw a need to have an exquisite fine-tuning of util calculations. Geez, the scores would be even more volatile than they already are, which is scary to imagine!

I think that when you're using statistical analysis to categorize anything, including consumers of credit, you want to minimize the number of "blobs" into which you assign them, or it becomes to unwieldy to use.

[RobertEG]

Logical, I also have a strong math and programming background, and see excactly where you are coming from.  Maybe some of what you point out may account for some of the wider swings in scoring that we all see when only a minor difference is reported.

Skip this paragraph if you hate math, but here is my point.  The FICO models were developed based upon millions of credit records, and  the assignment of risk in future default in the next couple of years based upon this statistical data for any category, such as %util.  Statistically, such a broad dataset will inevitably not be linear, which in English means the rate of change of risk at each plotted point will be different.  This, in math, is called a non-linear equation, and math has a way to calculate instantaneous rates of change.  For those mathematically inclined, it requires the simple use of calculating the derivative of the statistical equation at any point, which is its rate of change at that point, thus showing how quickly one's risk is improving or declining.  Piece of cake, and any competent programmer at FairIsaac with math knowledge. could easily do this, and incorporate this into their risk model algorithm.  It is not mathematically complex, and in may ways may be much simpler to do. 

For those I have lost with this mathematical diatribe, I will return to plain talk.  It would be easy mathematically and from a programming perspective to tweak any any all FICO category scores at any instant leve l and point in timel, but this would result in instantaneous FICO changes.  If the assumption is true, and I dont know how they program it, then FICO scores would exhibit much less daily stability, which aint cool.  However, as most seem to feel from anecdotal experiences, the algorithm may rely on "cutoff" trigger points along the curve to change its risk assessment, thus changes in FICO score.  While inherently being slightly less mathematically pure in their daily assessments, use of such trigger points would nonetheless show more short term stability at the expense of larger "bumps" once such thresholds are hit.

Sorry for all the math talk, but after all, FICO is math, and I think the opinions expressed on this topic make a lot of math sense to me.

And all I have said above has NOTHING to do with "buckets."  All of this is within only one scoring algorithm, and the way it is set up to do the match.  When different "buckets" are entered into the scoring, then an entirely different scoring algorithm is used based upon your new peer comparison group.  So I think we have both programming bumps within each bucket, compounded by scoring bumps when moving to another bucket.

Bump, bump bump!

Message Edited by RobertEG on 04-17-2008 04:20 PM

Message Edited by RobertEG on 04-17-2008 04:23 PM

[Anonymous]

---

RobertEG wrote:

.

Skip this paragraph if you hate math, but here is my point.  The FICO models were developed based upon millions of credit records, and  the assignment of risk in future default in the next couple of years based upon this statistical data for any category, such as %util.  Statistically, such a broad dataset will inevitably not be linear, which in English means the rate of change of risk at each plotted point will be different.  This, in math, is called a non-linear equation, and math has a way to calculate instantaneous rates of change.  For those mathematically inclined, it requires the simple use of calculating the derivative of the statistical equation at any point, which is its rate of change at that point, thus showing how quickly one's risk is improving or declining.  Piece of cake, and any competent programmer at FairIsaac with math knowledge. could easily do this, and incorporate this into their risk model algorithm.  It is not mathematically complex, and in may ways may be much simpler to do. 

---

RobertEG

It has been many years since I studied trigonometry and calculus...   Just wanted you to know that I DID follow your paragraph, and it DID make sense to me.

[smallfry]

Maybe a little off topic but I noticed that when I let 3 credit cards report a balance my score went down a bit. Experian dipped from 707-693. I have 6 cards in total. The 3rd card that reported only reported a $24 balance on a 40K credit limit for a recurring charge that was posted before I could pay the small amount online. So I think you really need to watch the number of accounts with a balance even if they are small amounts. I try to rotate the cards every month or so. Having any more than what I currently have would be hard for me to manage. I never want to risk an issuer closing for inactivity so I closed a number of cards I opened last year when I realized I went app crazy. Never again.

[Logical]

---

smallfry wrote:
Maybe a little off topic but I noticed that when I let 3 credit cards report a balance my score went down a bit.

---

The amount I owe on cards is steadily going up. Several card's util is 70%+. Last i looked 76$ of the TLs were reporting a balance, yet since the first of the year my facos and ficos are going up! WaMu pfico was +25 in March and +18 in April. The only change is the amount I owe rising and 3 cards had their 1st birthday in February. This is why I prefer to look at long trends, not short term manipulation. The changes could come from who knows where!

[smallfry]

---

@Logical wrote:

---

@smallfry wrote:
Maybe a little off topic but I noticed that when I let 3 credit cards report a balance my score went down a bit.

---

The amount I owe on cards is steadily going up. Several card's util is 70%+. Last i looked 76$ of the TLs were reporting a balance, yet since the first of the year my facos and ficos are going up! WaMu pfico was +25 in March and +18 in April. The only change is the amount I owe rising and 3 cards had their 1st birthday in February. This is why I prefer to look at long trends, not short term manipulation. The changes could come from who knows where!

---

I know in my case the reason the scores went down during that one month period was because I let 3 of 6 open revolving lines report a balance. The next month with exactly the same utilization % and only 2 cards reporting the scores went right back up to where they were the previous month to the number. My report doesn't change at all anymore. No new accounts no new inquiries just a month further down the line across the board. I imagine if I live to be 100 I could see 800. I started playing the game too late.