Two Medicare late enrollment penalties are described the same way in almost every summary: sign up late, pay more for life. They are built on different units. The Part D drug penalty counts single months and multiplies them by a premium figure that the Centers for Medicare & Medicaid Services resets every year. The Part B medical penalty counts only completed twelve-month periods and multiplies them by a much larger premium. For 2026 the two base figures are $38.99 and $202.90. What follows is what each formula does with the same gap, and why one of them costs a different amount every January while the gap behind it does not change. What Starts the Part D Count The condition is not lateness in general. The Medicare drug coverage cost page states: “You may owe a late enrollment penalty if at any time after your Initial Enrollment Period is over, there’s a period of 63 or more days in a row when you don’t have Medicare drug coverage or other creditable prescriptio...
A credit score can fall in the same month a credit card balance goes to zero. That outcome surprises people because it looks like a reward being withheld. In scoring terms it is ordinary: the score never measured the payoff. It measured a set of file characteristics, and the payoff changed several of them at once, in opposite directions.
FICO publishes the weighting of its general-purpose scoring models, and the Consumer Financial Protection Bureau publishes guidance on the two inputs that carry most of the weight; the figures below were read from those publishers' own pages in August 2026. Read together, those two sources explain most of the movement that looks arbitrary from the outside. This article walks through the five weighted categories, then works through the specific scenario — paying off a card and closing it — where the arithmetic runs against intuition.
What the five categories are, and how much each one carries
FICO states that its scoring models sort the information on a credit report into five categories, with these approximate weights for the general population:
Two features of that chart do most of the explanatory work.
First, payment history at 35 percent and amounts owed at 30 percent together account for roughly 65 percent of the model. The three remaining categories — length of credit history at 15 percent, credit mix at 10 percent, and new credit at 10 percent — split the rest. A change that touches only those three has a ceiling on how far it can move a score.
Second, FICO attaches an explicit caveat to the weights: they describe the general population, and the importance of each category can differ from one credit profile to another. A file with a single late payment and a file with a bankruptcy do not weight payment history identically. The percentages are a description of the model's overall shape, not a per-account formula that can be applied by hand.
Payment history: what it counts and how long it counts for
FICO describes this category as whether past credit accounts have been paid on time. The CFPB calls payment history the number one factor for building a strong credit score, and its guidance on recovering from missed payments is to get current and stay current rather than to take any corrective step against the record itself.
The record has a defined shelf life. Under the CFPB's summary of the federal credit reporting rules, a consumer reporting company generally may report most negative information for seven years. Bankruptcies may stay on a report for up to ten years. Lawsuits and judgments may be reported for seven years or until the governing statute of limitations expires, whichever period is longer.
Those periods are the single most useful thing to know about payment history, because they set expectations that no repayment strategy can override. A collection that is paid in full does not reset to zero; it becomes a paid collection, and the CFPB's own glossary defines a paid collection as exactly that — an account that went into collections because it was past due and was then paid. Payment changes the status. It does not change the clock.
Amounts owed: where a payoff can move the number the wrong way
FICO describes the amounts owed category as an assessment of whether a lot of available credit is being used. The CFPB's practical version of the same idea is a threshold: keep use of credit at no more than 30 percent of the total credit limit.
That is a ratio, which means it has a numerator and a denominator, and a payoff can change both. The scenario below is the ordinary case in which the second effect swamps the first.
The arithmetic, step by step
Take three cards. Card A carries a $6,000 limit and a $2,900 balance. Card B carries a $3,000 limit and a $1,500 balance. Card C carries a $2,000 limit and a $600 balance.
- Starting point. Limits total $11,000. Balances total $5,000. Overall utilization is 45.5 percent — well above the CFPB's 30 percent figure.
- Card A is paid to zero. Limits are unchanged at $11,000. Balances fall to $2,100. Utilization drops to 19.1 percent.
- Card A is then closed. Its $6,000 limit leaves the calculation. Limits fall to $5,000, balances stay at $2,100, and utilization climbs back to 42.0 percent.
The $2,900 payment removed 26 percentage points of utilization. Closing the account gave back 23 of them. The file ends the month with less debt and a ratio that is barely better than where it began, because 55 percent of the available credit left the report along with the balance.
A second layer sits underneath the overall figure. Individual account utilization is also visible on the report: Card B at $1,500 against a $3,000 limit reads as 50 percent on its own line, regardless of what the aggregate says. A file can show a comfortable overall ratio and still carry one card near its ceiling.
Reporting dates, not calendar dates
A related mismatch has nothing to do with closing anything. A credit report is a series of snapshots. The balance a scoring model sees is the balance the issuer reported on its own reporting date — not the balance on the day the score is calculated, and not the balance after the payment clears.
If an issuer reports on the statement closing date, a card with a $4,000 limit used for $1,900 of spending each month will report $1,900, or 47.5 percent, even when the statement is paid in full every cycle and no interest is ever charged. The same account, paid down to $400 before that closing date, reports 10 percent. Nothing about the household's finances differs between the two versions. Only the timing of the snapshot does. Issuers set their own reporting dates, so this depends entirely on the specific card.
The balance-carrying myth, priced out
A persistent belief holds that a card must carry a balance for the account to help a score. The CFPB addresses it directly: carrying a balance is not required, and paying cards off monthly produces the best scores while minimizing interest. Nothing in FICO's description of the amounts owed category rewards a balance that is carried rather than paid.
The cost of acting on the myth is measurable, because the Federal Reserve publishes the rate banks actually charge.
The Federal Reserve's G.19 release reports that in 2026 Q2 the average rate on credit card accounts assessed interest was 22.15 percent, against 20.94 percent across all accounts. Both figures sit roughly six percentage points above their 2021 levels — 16.45 percent and 14.60 percent respectively. On a $2,000 balance carried for a year at 22.15 percent, the interest cost for the year comes to about $443. The scoring model does not register the payment of that $443 as a positive input. It sees only whether the account was paid on time and what balance was reported.
The three smaller categories
Length of credit history (15 percent). FICO states that this category considers the age of the oldest account, the age of the newest account, and the average age of all accounts. That is three separate measurements, which is why the effect of closing an old account is not a single predictable number. FICO does not publish the retention rules for closed accounts in good standing, so a specific figure for how long a closed account keeps contributing should not be assumed.
Credit mix (10 percent). FICO describes this as the mix of credit cards, retail accounts, installment loans, finance company accounts, and mortgage loans. This is the category most affected when an installment loan — an auto loan, a personal loan — is paid off early. The loan's payment record remains on the report under the retention rules above, but the file now has one fewer open account type. At a 10 percent weight, the effect is bounded.
New credit (10 percent). FICO examines the opening of several accounts in a short period. On inquiries specifically, FICO states that inquiries remain on a credit report for two years, while FICO Scores consider only inquiries from the last 12 months, and that inquiries usually have a small impact, with many types ignored completely. FICO's public education material states that its models allow for rate shopping, but that page does not publish the exact window used to consolidate multiple applications for the same loan type, so no specific number of days is asserted here.
Checking one's own report does not create a scored inquiry. The CFPB confirms consumers are entitled to at least one free report per year from each of Equifax, Experian, and TransUnion through AnnualCreditReport.com, and notes more frequent free online access may be available.
Which score is being looked at
A score that moves unexpectedly is sometimes not the score a lender saw. The CFPB studied this in 2012, analyzing 200,000 credit files from each of the three major bureaus — 600,000 files in total — and found that roughly one in five consumers would likely receive a meaningfully different score than a creditor would, meaningfully different being defined as a gap large enough that the consumer would likely qualify for different credit offers.
The practical consequence is that a monitoring app's number and a lender's number are different measurements of the same file, produced by different models. Comparing a decline letter against a monitoring dashboard is comparing two instruments, not detecting an error.
Numbers to Re-check
| Figure | As used here | Where to verify | When it changes |
|---|---|---|---|
| FICO category weights | 35 / 30 / 15 / 10 / 10 percent | myFICO credit education, "How are FICO Scores calculated" | When FICO revises or republishes model documentation |
| Utilization guidance | No more than 30 percent of total limit | Consumer Financial Protection Bureau, Ask CFPB | On CFPB guidance updates |
| Negative information retention | Generally 7 years | Consumer Financial Protection Bureau, Ask CFPB | Only by amendment to federal credit reporting law |
| Bankruptcy retention | Up to 10 years | Consumer Financial Protection Bureau, Ask CFPB | Only by amendment to federal credit reporting law |
| Inquiry treatment | 2 years on report; 12 months scored | myFICO credit education, new credit | With new scoring model versions |
| Credit card rate, accounts assessed interest | 22.15 percent, 2026 Q2 | Federal Reserve G.19 Consumer Credit release | Quarterly, with the monthly G.19 release |
| Free report entitlement | At least one per bureau per year | Consumer Financial Protection Bureau; AnnualCreditReport.com | When the bureaus change their free-access programs |
Where This Doesn't Apply
Thin and inactive files. The weighting above assumes a file with enough history to be scored at all. A report with very few accounts, or with no recently reported activity, may not generate a score, in which case the categories have nothing to weight.
Other scoring models. The 35 / 30 / 15 / 10 / 10 split is FICO's published description of its own general-purpose models. It is not the structure of VantageScore, and it is not a description of proprietary lender models built on bureau data. A number labeled "credit score" in an app may come from a model with different inputs entirely.
Industry-specific FICO versions. Auto lenders and card issuers frequently use industry-tuned versions that run on different scales than the general-purpose score. A figure from one scale cannot be compared against a figure from another.
Files with major derogatory events. FICO states plainly that category importance varies by profile. On a file containing a bankruptcy or a recent charge-off, marginal utilization changes behave differently than they do on a clean file, and the general-population weights are a poor guide.
Decisions that are not score-driven. Manual underwriting, portfolio-specific rules, income and debt-to-income tests, and required documentation all operate independently of the score. An approval or a denial is not necessarily a statement about the number.
State law and account type. States may impose credit reporting requirements in addition to the federal rules described above. Business credit files are maintained separately from consumer files and are not governed by the consumer framework discussed here.
Concrete framework
For anyone trying to reconcile a score movement against what changed on the file:
- Separate the numerator from the denominator. Establish whether total balances fell, total limits fell, or both. A payoff paired with a closure changes both, and the direction of the ratio depends on which change was larger.
- Check the reported balance, not the current balance. The relevant figure is what the issuer reported on its reporting date. Pull the report and read the balance line rather than the account app.
- Check per-account utilization as well as the aggregate. One card near its limit is visible on its own line.
- Locate the event on the retention timeline. Most negative information runs seven years; bankruptcy runs up to ten. Anything inside that window is still being counted.
- Confirm which model produced the number. Given the CFPB's finding that about one in five consumers would receive a meaningfully different score than a creditor would, two numbers that disagree are not automatically evidence of an error on the file.
- Read the reason codes. A declined application generally comes with a statement of the specific factors that most affected the score. Those factors are model output and are more informative than any general weighting table.
The underlying point is that the weights describe a file, not a behavior. Paying a debt is a decision about interest cost and cash flow. Whether it raises a score in the same month depends on which of the five categories it touched and in which direction — and those two questions have different answers.
This article explains how the rules are written. It is not tax, legal, or insurance advice, and it does not account for any individual situation. Amounts and thresholds change; verify the current figures at the source listed above before acting.
Comments
Post a Comment