How to Calculate Percentages: Three Formulas That Cover Everything

What a percentage actually is

A percentage is a fraction with the denominator fixed at 100 — the word comes from the Latin per centum, "by the hundred". Saying 37% means 37/100, which means 0.37. That is the whole definition, and it is worth internalising, because every percent calculation is just fraction arithmetic with the denominator pre-agreed. Wolfram MathWorld defines it in one line; the NIST style guide treats the % symbol as nothing more than shorthand for the number 0.01. Percentages are not a special kind of number — they are ordinary numbers wearing a convenient uniform.

The uniform matters because human brains compare hundredths far more easily than raw fractions. Told that one shop takes 3/40 off and another takes 2/25, most people stall; told 7.5% versus 8%, nobody does. That convenience is also where the traps live — percentages feel so familiar that people add them, average them and reverse them in ways the underlying fractions do not allow. The percentage calculator handles the three core operations; this guide covers how to recognise which one your problem is, and the handful of traps that produce almost every wrong percentage answer.

The three questions every percent problem reduces to

Strip the wording away and nearly every everyday percentage problem is one of three questions:

1. What is A% of B? — result = (A ÷ 100) × B. This is the discount question, the tip question, the "8% tax on this price" question. You know the rate and the base; you want the piece.

2. A is what percent of B? — result = (A ÷ B) × 100. This is the test-score question, the market-share question, the "what share of my income goes on rent" question. You know the piece and the whole; you want the rate.

3. What is the percentage change from A to B? — result = (B − A) ÷ |A| × 100. This is the price-rise question, the pay-rise question, the year-on-year growth question. You know before and after; you want the relative movement, signed: positive for an increase, negative for a decrease.

The skill is not the arithmetic — the percentage calculator does that instantly, and shows the formula it used. The skill is translation: hearing "the bill came to 84 and we want to leave about 15" as question 1, "42 of the 60 orders shipped on time" as question 2, and "rent went from 950 to 1,020" as question 3. One reliable tell is the word of — in percentage language, "of" almost always marks the base, the number that goes on the bottom of the fraction or gets multiplied by the rate. Find the "of" and you have found B. These identities are standard pre-algebra — see OpenStax Prealgebra for the textbook treatment — but the translation step is where real-world problems go wrong, so the rest of this guide lives there.

Worked example: one purchase, all three questions

A laptop lists at 1,250. The shop takes 18% off, then adds 8% sales tax on the discounted price. Chase it through:

The discount (question 1). 18% of 1,250 = (18 ÷ 100) × 1,250 = 225. The sale price is 1,250 − 225 = 1,025. A shortcut worth adopting: "18% off" means you pay 82%, so go straight to 0.82 × 1,250 = 1,025 in one step. The discount calculator is built around exactly this pattern.

The tax (question 1 again, new base). 8% of 1,025 = 82. Total at the till: 1,025 + 82 = 1,107. Note the base moved — the tax applies to the discounted price, not the list price. Getting the base right is most of the battle, as the next section shows.

The damage (questions 2 and 3). What percent of the list price did you actually pay? 1,107 ÷ 1,250 × 100 = 88.56%. And as a percentage change from the list price: (1,107 − 1,250) ÷ 1,250 × 100 = −11.44%. Run any of these through the percentage calculator by picking the matching mode from the dropdown — it will echo the formula back with your numbers in it, which is a good habit to check against.

One more thing hiding in this example: an 18% discount followed by 8% tax is not a net 10% discount. You paid 88.56% of list, a net reduction of 11.44% — because the two percentages acted on different bases. That observation opens the biggest trap of all.

The traps: where percentage intuition fails

Percentages of different bases never add

A 10% rise followed by another 10% rise is a 21% rise, not 20%: 1.10 × 1.10 = 1.21, because the second 10% acts on the already-raised value. Successive percentage changes multiply as growth factors — convert each change to (1 + rate), multiply them, convert back. Three annual rises of 3% compound to 1.03³ ≈ 1.0927, a 9.27% total — the same machinery that drives the compound interest calculator. Adding percentages is only legitimate when they share the exact same base at the same time: 30% of voters plus 25% of the same voters is meaningfully 55% of voters.

Up and down are not symmetric

A 20% fall followed by a 20% rise does not get you home: 100 drops to 80, and 20% of 80 is only 16, landing you at 96. The recovery needed after a loss is always bigger than the loss — after losing p%, you need a gain of p ÷ (100 − p) × 100 to break even. Lose 20%, need 25%. Lose 50%, need 100%. This is not a curiosity; it is the reason "the market fell 30% but has since risen 30%" describes a portfolio still 9% underwater (0.70 × 1.30 = 0.91).

Reversing a percentage is division, not subtraction

You know the sale price was 1,025 after 18% off — what was the original? The tempting move, adding 18% back (1,025 × 1.18 = 1,209.50), is wrong. The discount was 18% of the original, not of the sale price, so the reversal must undo the multiplication: 1,025 ÷ 0.82 = 1,250. The same logic strips tax out of a tax-inclusive price — a receipt of 120 that includes 20% tax contains a base of 120 ÷ 1.20 = 100, not 120 − 20% = 96. This is precisely how a VAT calculator works backwards from gross to net, and the VAT guide walks through why the two answers differ. Whenever you find yourself undoing a percentage with the opposite operation at the same rate, stop — you are about to be off by the square of the rate.

Percentage change, percentage difference, percentage points

Three different measures share the % sign and get conflated constantly. Percentage change is directional and uses the starting value as its base: 40 → 60 is +50%, while 60 → 40 is −33.3% — same pair of numbers, different answers, because the base switched. Percentage difference is symmetric: it divides the gap by the average of the two values, so 40 versus 60 is 20 ÷ 50 = 40% either way; use it when neither number is the natural baseline. And percentage points measure the raw gap between two things that are already percentages: a poll moving from 45% to 50% approval has risen 5 points but 11.1% in relative terms. Headlines routinely pick whichever framing sounds most dramatic — "support surged 11%" and "support rose 5 points" can describe the same poll. When the compared values are a measurement and an accepted true value, the convention is different again — that is percent error, covered in the percent error guide and its calculator.

Mental-math shortcuts that actually work

• Build from 10%. 10% is a one-place decimal shift, and most everyday rates assemble from it. A 15% tip on a 64 bill: 10% is 6.40, half that again is 3.20, total 9.60. Faster than finding your phone — though the tip calculator also splits the bill.
• Flip the problem. A% of B always equals B% of A, because (A ÷ 100) × B = (B ÷ 100) × A. 18% of 50 feels hard; 50% of 18 is instantly 9. Same number.
• Work with what you keep, not what you lose. "25% off" means "pay 75%" — one multiplication (× 0.75) instead of a percentage-then-subtraction. This habit also makes reverse percentages natural, because ÷ 0.75 is its obvious undo.
• Chain growth factors, not percentages. For successive changes, multiply the (1 + rate) factors and only convert back to a percentage at the end. +12% then −5% is 1.12 × 0.95 = 1.064 → +6.4%, with no opportunity for the adding-percentages error.
• Sanity-check against 1%, 50% and 100%. Before trusting any answer, bracket it: 1% of 1,250 is 12.50, so 18% must be a bit under 20 × 12.50 = 250. An answer of 225 passes; an answer of 22.50 or 2,250 fails immediately. Order-of- magnitude checks catch far more real mistakes than careful arithmetic does.

Common mistakes

Dividing by the wrong end of a change. Percentage change always divides by the starting value. Rent going from 950 to 1,020 is a rise of 70 ÷ 950 = 7.37%, not 70 ÷ 1,020 = 6.86%. Dividing by the new value systematically understates increases and overstates decreases — a favourite trick of anyone presenting numbers they would rather you not examine.

Averaging percentages from different-sized groups. A seller with 90% positive ratings on 10 sales and 60% on 1,000 sales does not average to 75%. The honest figure weights by group size: (9 + 600) ÷ 1,010 = 60.3%. Unweighted averages of rates are only valid when every group is the same size — which, outside textbooks, they never are.

Treating a percentage as an amount. "Which would you rather: 40% off a 50 item or 25% off a 100 item?" The smaller percentage here is the bigger saving (25 versus 20). A percentage is meaningless without its base — the ratio machinery underneath is the same as in the ratio calculator, and any time a claim quotes a rate but hides the base, the base is the first thing to go and find. Related: fraction ↔ decimal ↔ percent are three costumes for one number, and moving between them (0.375 = 37.5% = 3/8) should feel mechanical.

Confusing markup and margin. A product bought for 80 and sold for 100 carries a 25% markup (20 ÷ 80, base = cost) but a 20% margin (20 ÷ 100, base = price). Same transaction, both correct, different bases — and pricing a catalogue with the wrong one quietly erases profit. The markup calculator keeps the two straight.

When the three formulas aren't enough

The three modes cover point-in-time questions. The moment time enters — money growing at a rate, year after year — single percentage changes stop being the right tool and compounding takes over: repeated application of (1 + rate), which is the compound interest calculator's territory. Measurement and lab work have their own convention, comparing an observed value to a true value — the percent error calculator handles that. And statistical claims — margins of error, relative risk, "50% more likely" — need the base-rate context no formula can supply. For everything else, the answer is one of the three questions above, and the percentage calculator will show its working while it answers.

Frequently asked questions

The FAQ on the percentage calculator page covers the three formulas themselves, percentage points versus percentage change, values over 100%, and why zero-baseline changes are undefined. The questions below tackle what comes up in actual use — reversing discounts, why changes don't cancel, when adding percentages is legal, and the mental-math shortcuts. For neighbouring problems, see the discount calculator guide and the ratio guide.