Does "# of Accts Reporting a Balance" include installments or just revolvers when score is impacted?

[Trudy]

---

@Anonymous wrote:
Trudy Also remember that the algorithm can and is tweaked by each CRA, so there could be a line at TUwhere there is not at other bureaus or it could only be on certain scorecards at certain CRAs.

---

Totally agree, TU is that CB and most of its scores that respond different than the other bureau's to most manners of “known” thresholds...and some.  As a result I’m not convinced there are absolute thresholds.?.?

The struggle continues, otherwise we’d have to find something else to do.  

[Anonymous]

I guess it all depends on how you define the word absolute.

I think the major utilization thresholds are well known, tested, and established. That doesn’t mean all of them are known. For instance, @CassieCard found one at 4% that as far as I know, was previously unknown.

As far as number of accounts ago, I think they relied on the conventional wisdom that was already established for a while and it looks like it needs further testing.

You have so many different versions and flavors that it would be inconceivable to try to document everything for all of them. And it would just cost too much money for any bureau other than Experian, at least in my opinion and coming out of my wallet, although I did splurge a bit for my AOYA testing. I planned that for a year, so I had to pay the piper. But I was specifically looking at older scores as well, so I kind of had to.

[Anonymous]

A couple of points.

EX is my bullet proof bureau when it comes to number of accounts with a balance.  At the time I was at 847 when I did my testing there and I stayed at 847 in going from 2 accounts (1 revolver and 1 installment loan) with a balance to 9 accounts (8 revolvers and 1 installment loan) with a balance.  Since I was at 847 to start, it eliminates buffer from the equation.

BM, regarding number of accounts with a balance, I think the key word here is number, meaning that actual number in addition to percentage may play a role.  No way to know for sure of course.  Let's say that 33% is a threshold on EQ percentage wise.  That of course would be 1 out of 3 accounts.  33% however could also be 5 out of 15 accounts, but this person would have 4 more accounts (5 verses 1) with a balance.  If there were a threshold in terms of number of accounts at (say) 3 or 4, such a profile could take a number of accounts penalty that has nothing to do with a percentage of accounts.  If that's indeed a "thing" it could give a false-positive as far as a [percentage] threshold below 33% on that bureau.  Just something to consider. 

[Anonymous]

Very interesting observation BBS. That’s why I’ve been waiting for your input and looking forward to it.

Let me ponder the implications because that is theoretically possible.

[Anonymous]

BBS I wonder if you were capped at 847? We know there would’ve been a penalty for having that many accounts with a balance.

if you couldn’t see it, the only reason could be the buffer and you being capped at a certain score; otherwise how do you explain the fact that the penalty didn’t apply to you? Would love to hear any theories you’ve came up with, because we know that would impose a penalty. why else could you not see it?

[Anonymous]

I don't believe I was capped, or can't see a way that I would have been.  My score moved from 847 in August 2018 to 850 in September 2018.  At the time my age of accounts factors moved up 1 month to AAoA 6 years 8 months, AoYA 16 months and AoOA 17 years 4 months.  Other testing I did around that time in going from AoYA of 11 months to 12 months took me from 834 to 847.  Going from AZEO to AZ (no revolving credit use) resulted in a score drop from 847 to 825.  I had zero scoreable inquiries on EX throughout the duration of my testing.

[Anonymous]

---

@Anonymous wrote:
Very interesting observation BBS. That’s why I’ve been waiting for your input and looking forward to it.

Let me ponder the implications because that is theoretically possible.

---

I just feel the word "number" found in negative reason statements is significant.  If whoever wrote the algorithm was thinking purely from a percentage standpoint, I would expect the reason statement to read "percentage" of accounts...  We of course do know that percentage does indeed matter, but simply based on the word choice I'm inclined to believe that raw number very well could play a role.

[Anonymous]

BBS That’s very interesting, so why would you not receive a penalty for an increase in number of accounts with a balance? Now that’s a mystery and I’d love to know the answer. And yes from what you said, it doesn’t seem that you were capped.

And as I told you I’ve been running a deficit of sleep. Actually one of the things Rev and I were working on was determining if it is a number or a percentage. I wish we would’ve finished all that and got definitive answers.

When I get time and go back over the data, I may be able to determine an answer. We devised methods to determine whether it was number or percentage.

Likewise we were trying to determine the percentage thresholds, assuming it was percentages, which it appeared to be and I believe he also believed it to be.

But you bring up an interesting point that there could be a dual-mechanism that still brings number into play despite percentage. That’s very good thinking on your part that I had not even considered. That’s the beauty of collective wisdom! 😉

The problem is, it could be different between models and it could be different between scorecards. I tend to believe it would probably be the same across models, but there could be a difference across scorecards. On a thick file a small percentage could still be a large number of cards, depending on the profile, so there could be a fail-safe as you hypothesize. Definitely something to think about.

[Anonymous]

---

@Anonymous wrote:
And as I told you I’ve been running a deficit of sleep. Actually one of the things Rev and I were working on was determining if it is a number or a percentage.

---

I hear you.  For now, until further notice, I'm going to go with "both." 

I think utilization is another factor where "both" can come into play depending on scoring model.  By both I mean utilization percentage as well all know it, but also raw dollars at times.  For example, I've been at AZEO in the past with that card being at (say) 35% reported utilization.  Then I've moved to AZEO on a different card, with a similar utilization percentage reported and seen a score change.  The only difference would be dollars.  All major bank / non-AU cards of course.  I remember one month I reported something like $9k on my ~$24k at the time Citi DC as my AZEO card where prevously I had a $3500 balance or so on a Chase Amazon card with a $10k limit.  Both of these percentages were similar and within the same threshold range, but with the higher dollars reported I recall a ding that I could only chalk up to actual dollars.  I'm quite sure VS looks at raw dollars perhaps even moreso/often than Fico, too.

[Anonymous]

BBS Very Good point about utilization and there’s even some documentation that implies that and I’m glad to hear that your testing confirmed that, as I’ve read other threads that seemed to indicate that as well.

I definitely cannot rule out your hypothesis. However I don’t think it would serve a purpose on thin files, only thick files. The percentage would do the trick but as previously stated, on a thick file a small percentage could still be a large number of cards. So something like that would definitely make sense there.

I think we would need to compare some thin and thick profiles with testing to try to make any determinations. Meanwhile, like you said, we cannot rule it out.

I have found a 20% threshold though. The thing about that threshold is this, I tested it with varying number of revolvers from around 14 or 15 up to now 20. So I’m kinda sure that threshold is there independent of a potential dual mechanism.