Chapter 2: The credit score machine
Lena is 41, two years past a hard divorce, and her credit score is 548. The number reads like a verdict on her character, and that reading is wrong in a useful way. A credit score is a prediction, nothing more: an estimate of how likely you are to fall 90 days behind on a payment within the next couple of years. Lenders use it to decide whether to say yes and what interest rate to charge. It does not know about the divorce, the lawyer bills, or the months she held everything together on one income. It only reads the file, and files can be changed.
This chapter opens the machine. You will see the five inputs and their exact weights, the things the formula ignores, what each score band actually unlocks, and which levers move fast versus slow. By the end, Lena's 548 will look less like a judgment and more like a parts list.
Five factors, five behaviors
The dominant score in American lending is FICO, which runs from 300 to 850. FICO publishes the recipe as five weighted ingredients, and each one maps to a plain behavior.
- Payment history, 35%. Did you pay at least the minimum, on time, on every account, every month. This is the largest slice because past delinquency predicts future delinquency better than anything else. A single payment reported 30 days late can undo years of quiet consistency.
- Amounts owed, 30%. Mostly utilization: your reported card balances divided by your limits, measured per card and across all cards. Owing $300 on a $1,000 limit reads as 30%. Lower reads as safer, and the formula has no memory here; it looks at the latest snapshot.
- Length of credit history, 15%. The age of your oldest account and the average age of all of them. Time in the system, slowly earned.
- Credit mix, 10%. Experience with both revolving accounts (cards) and installment loans (car, student, personal). A small slice, never worth borrowing just to fill.
- New credit, 10%. Recent applications, each one a hard inquiry, and freshly opened accounts. A burst of new credit reads as strain.
Notice what the weights imply. Payment history plus amounts owed equals 65% of the score. You can hold a strong score while the other three factors are mediocre, and you cannot hold one while either of the big two is broken.
What the formula ignores
Your income is not in the score. Neither is your rent, your savings balance, your job title, or your degree. A nurse earning $48,000 with clean payment history will outscore an executive earning $400,000 who pays late. Checking your own score is also harmless: that is a soft inquiry, invisible to the formula, no matter how often you look.
Then there is the most expensive myth in consumer finance: the idea that you should carry a balance so the card "shows activity" and feeds your score.
Interest buys zero points
The formula sees the balance your statement reports and whether you paid on time. There is no field for interest paid, so carrying a balance cannot earn you anything. Carrying $500 at 26.4% costs $11 a month, $132 a year, and produces the identical score you would get by using the card and paying the full statement by the due date. Activity registers either way. The interest is a pure donation to the bank.
The bands, and what they gate
Lenders do not read scores as a smooth line. They read bands, and crossing a band boundary changes real prices.
| Band | Range | What it typically means in practice |
|---|---|---|
| Poor | 300–579 | Frequent denials; security deposits on utilities and phones; secured cards as the main way in |
| Fair | 580–669 | Approvals appear, at the highest APRs; subprime auto pricing; extra landlord scrutiny |
| Good | 670–739 | Mainstream card and loan approvals at reasonable rates |
| Very good | 740–799 | Strong approvals, the best card offers, sharply better auto rates |
| Exceptional | 800–850 | Pricing power nearly everywhere a score is read |
The bands also explain why two people can have such different experiences applying for the same card. An approval engine is not weighing your story; it is checking which band your number falls in and pricing accordingly. That is discouraging when you are at 548 and clarifying once you accept it, because bands are crossable. Moving from fair to good is worth more in real dollars than almost any coupon-clipping a household can do, and the rest of this guide is a map of that crossing.
Scores get read far beyond banks. Landlords screen with them. Utilities and phone carriers size deposits with them. Auto lenders price by band, and the spread between a fair-band and very-good-band car loan can run thousands of dollars over a single loan's life. Mortgage pricing has its own tiers, which the Finvest Home Buying Guide covers in detail.
One paragraph on the other brand you will meet. VantageScore is a competing model built jointly by the three credit bureaus, scored on the same 300 to 850 range, and it is what many free apps and bank dashboards show you. It uses the same raw file but groups and weights the ingredients differently, so it routinely lands 10 to 30 points away from your FICO. Most lenders still pull FICO when money is on the line. Track one brand's trend over time, and do not panic over gaps between brands; the behaviors that raise one raise both.
BNPL joins the file
For years, buy-now-pay-later plans were invisible to all of this. A pay-in-four plan created no tradeline, no utilization, no history. That ended in June 2025, when FICO announced FICO Score 10 BNPL and Score 10 T BNPL, the first scores to fold BNPL accounts into the formula. As lenders and bureaus adopt them, those plans stop being a side door and start being credit, with everything that implies. Paid on time, they can add positive history for people with thin files. Paid late, they damage a score the same way a late card payment does. The mechanics matter for stackers especially: someone juggling several plans at once now shows several open obligations, and the formula reads a cluster of fresh accounts the way it reads any other burst of new credit, as strain.
Tasha sat down with her banking app and counted the BNPL plans she had stopped tracking: sneakers, a coffee table, concert tickets, a winter coat. Four plans, $610 still owed, autopays scattered across the month. None of them appeared anywhere on her credit report, which once felt like a feature. Under the new scores they surface, which means her real borrowing picture, three cards plus four plans, is becoming visible to the same lenders who price her future car loan. She wrote all four into the debt list she started in Chapter 1, because a debt the formula can see is a debt the plan has to include.
Fast levers, slow levers
Here is the most practical fact in the whole machine: the five factors move at wildly different speeds.
Utilization is a snapshot with no memory. Your card reports one balance per month, usually on the statement date, and the score reacts to the newest photo. Pay a 70% utilization card down before the statement closes and the improvement can show up within one or two cycles, 30 to 60 days. Payment history is the opposite: a permanent record in which each on-time month is one brick, and a 30-day late stays on the report for seven years, fading in weight the entire time but never moving faster because you want it to. New-credit dings sit in between, with the score impact of a hard inquiry fading within months.
Now decompose Lena's 548 with the weights in hand.
A parts list, not a verdict
Payment history (35%): one charged-off card from 2024 and two collections. This is the crater, and only time and clean months refill it. Amounts owed (30%): her one open card reports $1,380 against a $1,500 limit, 92% utilization, and this is fully fixable within two statement cycles. Length (15%): her oldest account is 12 years old and intact. Mix and new credit (20%): three applications in two desperate months left inquiries that are already fading. Her fastest lever is the $1,380 balance; her slow lever is the calendar. Chapter 3 hands her a third lever: the report itself contains an error worth real points.
Play with the two big levers yourself and watch which one snaps back quickly. The simulator is illustrative, not a quote of your own score:
Two behaviors control 65% of your score: pay every account on time every month, and keep reported balances under 30% of limits, under 10% when a major application is coming. Everything else is tuning.
Where people go wrong
- Carrying a balance "for the score." It buys nothing and costs real interest, $132 a year on a $500 balance at 26.4%. Use the card, pay the statement in full.
- Closing the oldest card after paying it off. Closing shrinks total limits, which raises utilization today, and eventually trims history length. Pay it off and let it sit.
- Applying for several cards in one bad month. Each hard inquiry is small, but a cluster of them at 10% of the score, layered on new accounts, reads as strain exactly when you need approval odds.
- Panicking over the gap between a free-app score and a lender's FICO. Different brands, same file. Watch the trend, not the brand.
Key takeaways
- A credit score predicts the odds of falling 90 days behind. It measures the file, not your character or income.
- The FICO recipe: payment history 35%, amounts owed 30%, length 15%, mix 10%, new credit 10%. The first two are the whole game.
- Carrying a balance earns zero points and costs real interest. The formula has no field for interest paid.
- Bands gate real prices: roughly 670 starts "good," 740 "very good," 800 "exceptional," and crossing a band changes approvals, APRs, and deposits.
- BNPL plans entered FICO scoring with Score 10 BNPL in June 2025. Plans that were invisible now help or hurt like any account.
- Utilization moves a score in 30 to 60 days. Payment history takes months and years. Pull the fast lever first, then protect the slow one.
Sources: myFICO, what's in your credit score · FICO, scores incorporating buy-now-pay-later data · CFPB, credit reports and scores · Finvest Home Buying Guide