Utilization 9% --> 8%; FICO 8 +7

[SouthJamaica]

One account dropped from 15% to zero, causing Experian's rounded overall utilization percentage to drop from 9% to 8%, and number of accounts with balances to drop from 10/32 to 9/32.

Result:

FICO 8                   +7

Bankcard 8           +8

Auto 8                   +8

FICO 2                 +-0

[Revelate]

Mind doing the math as to what your non-rounded percentages were to call it 2 decimal places?

[SouthJamaica]

---

@Revelate wrote:

Mind doing the math as to what your non-rounded percentages were to call it 2 decimal places?

---

I don't use the actual percentages in these because (a) I found that EX was using its rounded percentages rather than actual percentages and (b) EX is always offbase a little in its computation because it continually treats a closed account of mine as though it were open.

But the actual percentages based on EX data were 8.95% (which EX rounded up to 9%) and 8.14% (which EX rounded down to 8%).

[Anonymous]

---

@SouthJamaica wrote:

---

---

I don't use the actual percentages in these because (a) I found that EX was using its rounded percentages rather than actual percentages and (b) EX is always offbase a little in its computation because it continually treats a closed account of mine as though it were open.

But the actual percentages based on EX data were 8.95% (which EX rounded up to 9%) and 8.14% (which EX rounded down to 8%).

---

I've been wondering about this for a while, and this is just more evidence to suggest that half-even rounding, or 'Banker's Rounding' is what is being used for utilization percentages.

With that method of rounding, using what is commonly called 'precision 0' in back end financial software, 8.50 becomes 8.0, 8.51 would be 9.0.  A 9.50 would be 10.0.

Of course, there might be presentation view rounding - what we see - that is separate from the actual real percentage used as an input to the FICO algorithm, which might be rounding to 2 or 4 decimal places. I'm just playing it safe and never going above 8.50 utilization. 

[SouthJamaica]

---

@Anonymous wrote:

---

@SouthJamaica wrote:

---

---

I don't use the actual percentages in these because (a) I found that EX was using its rounded percentages rather than actual percentages and (b) EX is always offbase a little in its computation because it continually treats a closed account of mine as though it were open.

But the actual percentages based on EX data were 8.95% (which EX rounded up to 9%) and 8.14% (which EX rounded down to 8%).

---

I've been wondering about this for a while, and this is just more evidence to suggest that half-even rounding, or 'Banker's Rounding' is what is being used for utilization percentages.

With that method of rounding, using what is commonly called 'precision 0' in back end financial software, 8.50 becomes 8.0, 8.51 would be 9.0.  A 9.50 would be 10.0.

Of course, there might be presentation view rounding - what we see - that is separate from the actual real percentage used as an input to the FICO algorithm, which might be rounding to 2 or 4 decimal places. I'm just playing it safe and never going above 8.50 utilization. 

---

Yes I do think that what you refer to as "half-even rounding" is what they are using. Having this daily service from Experian is giving me lots of new insights which would have been undetectable to me a few months ago.

[iv]

---

@Anonymous wrote:

I've been wondering about this for a while, and this is just more evidence to suggest that half-even rounding, or 'Banker's Rounding' is what is being used for utilization percentages.

With that method of rounding, using what is commonly called 'precision 0' in back end financial software, 8.50 becomes 8.0, 8.51 would be 9.0.  A 9.50 would be 10.0.

---

Except that there is absolutely no reason for that rounding method to be used in a scoring model.

The issues that it is meant to address don't exist here, and I've seen no compelling evidence to show that FICO scoring models actually round at all (vs truncation/threshold).

---

@Anonymous wrote:

Of course, there might be presentation view rounding - what we see - that is separate from the actual real percentage used as an input to the FICO algorithm, which might be rounding to 2 or 4 decimal places. 

---

Not "might be". There is clearly zero relationship between how any particular end-user display chooses to round, not round, or as can frequently happen, actually display different data entirely (AU vs non-AU, etc), compared to the real scoring models.

---

@Anonymous wrote:
I'm just playing it safe and never going above 8.50 utilization.

---

Nothing wrong with that - couldn't hurt to aim for even lower, if you were concerned about how other, less well-known models might react, too.

[NRB525]

The obvious factor here is an individual card crossed the 10% threshhold, on the way to zero. That is the most likely reason for the score improvement.

[Anonymous]

---

@iv wrote:

---

@Anonymous wrote:

With that method of rounding, using what is commonly called 'precision 0' in back end financial software, 8.50 becomes 8.0, 8.51 would be 9.0.  A 9.50 would be 10.0.

---

Except that there is absolutely no reason for that rounding method to be used in a scoring model.

The issues that it is meant to address don't exist here, and I've seen no compelling evidence to show that FICO scoring models actually round at all (vs truncation/threshold).

This confuses me a little, because so many posts say to stay under 8.99% , which I assume is playing it safe to avoid possible rounding and crossing into the [9, 29) utilization interval. (Or [9,19) according to some other posts.)

As of today, my score has been calculated based on a 0.47% real utilization using 40/8500. Most people say that gets rounded up to 1% utilization according to FICO. So if my next statement balances show 179.90 and 584.68 for balances on my 2 cards with a combined credit limit of 8500, did I just cross into that [9,29) interval with this new 8.995% aggregate utilization?

---

---

@iv wrote:

@Anonymous wrote:

Of course, there might be presentation view rounding - what we see - that is separate from the actual real percentage used as an input to the FICO algorithm, which might be rounding to 2 or 4 decimal places. 

---

Not "might be". There is clearly zero relationship between how any particular end-user display chooses to round, not round, or as can frequently happen, actually display different data entirely (AU vs non-AU, etc), compared to the real scoring models.

Yes, this much is certain. I can see how one part of Experian's report shows my 0.47% real utilization as 0% on one section and 1% on another. So obviously the report value should never be used when trying to test utilization intervals for score changes.

---

@iv wrote:

---

@Anonymous wrote:
I'm just playing it safe and never going above 8.50 utilization.

---

Nothing wrong with that - couldn't hurt to aim for even lower, if you were concerned about how other, less well-known models might react, too.

And here I thought everything in the [1,9) interval was treated the same. I saw a post where someone thought 5% was a threshold. In about 10 days, my 2 cards will be reporting 6.75% and 7.00% statement balances and I'll post the changes to 32 different scores. Then next month it will be 6/6 and then 5/5 after that. It seems that the general consensus is that I shouldn't see very much of a score change from now through April, when I get that 5/5 report.

[SouthJamaica]

I just reported above that EX treated 8.95% as 9% and 8.14% as 8%, and that the change from 9 to 8% triggered a 7-point increase in FiCO 8.

[SouthJamaica]

In my profile, changes in individual card utilization alone, other than changes crossing 30%, never affect my FICO 8 score, and even many of the 30% crossings have no effect.