A 100-Point Overnight Drop Has Four Likely Causes

A hundred points between two pulls is worth handling methodically, and most of what gets written about it is a list. A list is not much use: you already know the candidates: a late payment, a big balance, a collection. What you do not know is which one landed on your file, and the number will not tell you, because a score is a compression of a report.

The useful move is to stop interrogating the number and look at what changed underneath it. Every plausible cause of a large single-period drop leaves a specific, findable mark. A status field that flipped. A balance that moved while a limit sat still. A limit that moved while a balance sat still. A line that was not there last time. Those marks do not look alike, which is the point: you can tell the causes apart by looking.

So this is organised by evidence rather than by cause. Four things account for most large drops, each with its own signature. A fifth possibility is not a drop at all, and it is the most common false alarm, so it gets its own section. Then the case where the mark is real and belongs to somebody else.

Then work account by account. Four fields carry almost all of the diagnostic weight:

• The payment history grid and account status, which record whether a payment was reported late and in which month.
• The reported balance, which is what the furnisher sent, not what you owe today.
• The credit limit or high balance, the denominator in every utilization figure.
• The date reported, which is often much later than the event it describes.

Then scan for a tradeline present now and absent before, and read the inquiry list.

The working rule is blunt: the field that changed names the cause. If one status field moved and nothing else did, you are not looking at a utilization problem, whatever your balances are in general. If limits moved and balances did not, you are not looking at a spending problem. That gets skipped, because the instinct when a number falls is to reach for the explanation you already feared.

What it leaves behind is narrow. One account's payment history grid has a cell for one month that changed from a clean marker to a 30. The current status may well read current again, because you paid as soon as you noticed, which is why it gets missed. The grid is dated, so it hands you the month.

Two fields confirm it. The past due amount, which read zero in the earlier pull. And the date of last payment, sitting more than thirty days after the due date it was meant to cover. What this cause does not leave behind matters as much: balances unchanged, limits unchanged, no new tradeline. If the two files are identical except one status field on one account, you have your answer.

The second suspect is a statement balance that reported at or near the limit, and its signature is the opposite of the first. Balances moved. Limits did not. No status changed and no new account appeared.

The mechanism people miss: your report shows the statement balance, not what you are carrying today. The furnisher sends whatever was outstanding when the statement cut. Pay in full three days later and the report still shows the statement figure until the next cycle.

Utilization carries no memory. Unlike a delinquency, a dated historical record, the ratio is computed from whatever balance is currently reported. When a lower balance reports, the input the model reads is lower. What that is worth depends on the file and the model.

The third suspect is a derogatory tradeline appearing for the first time, and its signature is the easiest to spot: a line that was not on the earlier report. A collection agency name you have no relationship with, an original creditor field pointing at something you do recognise, a balance, and a date of first delinquency.

That last field resolves the word overnight. The event is not new. The reporting is new. A debt that went bad in the spring and was sold in the summer can surface on an ordinary Tuesday, dated to the original delinquency rather than to the day it appeared. An old date of first delinquency beside a recent date reported is an old problem arriving late.

Before accepting the entry, check the original creditor, the amount, and the date of first delinquency. That date is where errors concentrate, and it governs when the entry has to come off.

The fourth suspect is the one that feels unfair, because you did nothing. Your limits changed. Issuers close inactive accounts and cut limits on their own initiative. The signature is the mirror image of the second suspect: limits moved, balances did not.

Take the same four cards and 12,000 dollars of total limit, with 2,400 dollars of balances reporting. That is 20 percent. The issuer then closes the unused 5,000 dollar card. Balances are still 2,400, you have spent nothing, but the limits now total 7,000, so the ratio is 2,400 divided by 7,000, about 34 percent. The figure a model reads moved roughly fourteen percentage points on a decision you were not part of. A limit decrease does the same without closing anything: same numerator, smaller denominator.

To confirm it, compare the limit column across the two pulls account by account, and check whether the tradeline is marked closed by the grantor or by the consumer.

Rule this one out before any of the four, because it is not a cause at all. It is the most common false alarm in the category: the two numbers were never comparable.

Every credit score is three things: a model, a bureau, and a date. Change any one and the number changes without anything on your file having changed.

Then the bureau. You do not have a credit file, you have three, and furnishers need not report to all of them. A tradeline sitting at two bureaus and missing at the third produces different scores from the same model on the same day. The math is identical. The data is not.

Then the date: a report from last week already describes a file that has moved.

So before treating a gap as a drop, write down the model, the bureau and the date for both numbers. If any differ, you have compared two measurements, not measured a drop. This is also where a gap of about a hundred points is least likely to mean what it appears to: a scale difference alone can produce one.

The last case is where the mark on the file is entirely real and not yours.

The signature is a tradeline you do not recognise in any form. Not a collection for a debt you vaguely remember, but an account you never opened. It travels with company: hard inquiries you did not initiate in the weeks before the account appeared, because an application always precedes an approval. Check the personal information section too. Addresses you have never lived at, an employer you never worked for, a name you have never used.

There is also a remedy distinct from an ordinary dispute. Information you identify as resulting from identity theft can be blocked from your file on the strength of an identity theft report and proof of identity, and the bureau must act within four business days. A dispute asks a furnisher to verify an item; a block takes the item out of what the bureau reports. Use the block for fraud and the dispute for errors.

Four suspects, one imposter, and one that is closer to a police matter than a credit problem.

What makes this tractable is the refusal to diagnose from the number. A hundred points is a magnitude, not a symptom. The two reports, laid side by side and compared field by field, will tell you in an evening what the number never will.

The reason to be this disciplined is that the remedies barely overlap. Nothing you do about utilization touches a delinquency record: one is a current ratio, the other a dated historical entry. Disputing an accurate entry accomplishes nothing except a letter confirming it is accurate. A wrong diagnosis does not simply cost you a month; it sends you at the wrong field entirely, and you conclude that nothing works.

One honest limit. I have deliberately not attached point values to any of the four causes, and I would be wary of anything that does. The same event on two different files does not produce the same movement, because the models are reading everything else on those files at the same time. Find the field that changed, and you know what you are dealing with.

1. Can a credit score really drop 100 points overnight?

A score can change sharply between two pulls, and the change appears all at once because scores are recalculated when a report is pulled rather than continuously. The underlying event is often not overnight at all. A collection that surfaces today may carry a date of first delinquency from months earlier. Compare the date reported with the date of first delinquency to see which it is.

2. How do I tell a late payment from a utilization spike?

They leave opposite marks. A newly reported late changes one dated cell in one account payment history grid and moves the past due amount off zero, while balances and limits stay the same. A utilization spike moves balances while limits and every status field stay the same. If nothing about any balance changed, you are not looking at utilization.

3. Why did my utilization change when I did not spend anything?

Because utilization is a ratio and the denominator can move on its own. If an issuer closes an account or cuts a limit, your balances stay put and the total limit shrinks. On a file with 2,400 dollars of balances, dropping from 12,000 dollars of limits to 7,000 takes the ratio from 20 percent to about 34 percent with no spending at all.

4. Does paying my card in full stop a utilization spike from reporting?

Not by itself, because the furnisher generally reports the statement balance. If the statement cut with a large balance on it, that is the figure sent to the bureaus even if you paid it off days later. Paying before the statement cuts is what changes the reported figure, and the effect shows up when the next cycle reports.

5. I compared two scores and they were very different. Is that a drop?

Only if both numbers came from the same model, the same bureau and the same date. Base FICO and VantageScore both run 300 to 850, but industry-specific FICO auto and bankcard versions run 250 to 900, so a scale difference alone can produce a large gap. Different bureaus also hold different data, so the same model can return different numbers on the same day.

6. What should I check first if I do not recognise an account on my report?

Check the inquiry list for hard inquiries you did not initiate in the weeks before the account appeared, since an application precedes an approval. Then check the personal information section for addresses, names or employers you do not recognise. If the account is fraudulent, place a fraud alert or a freeze and file an identity theft report.

7. Is a dispute the right tool for a fraudulent account?

A dispute asks the furnisher to verify an item, which is the right tool for an inaccuracy. For information resulting from identity theft there is a stronger remedy: on the strength of an identity theft report and proof of identity, the bureau must block the item from your file within four business days. Use the block for fraud and the dispute for errors.