How to calculate: the four mistakes behind most wrong answers

The question behind the question

“How to calculate” on its own is one of the most searched phrases there is, which is odd for a question with no object. Nobody wants to know how to calculate in general. What the bare phrase signals is someone who knows they need a number and has not yet worked out which one.

That is worth taking seriously, because the step people skip is almost never the arithmetic. Division is not the hard part. Choosing what to divide by is.

Four families cover most of it: percentages, averages, rates that compound, and measurements of space. Each has a specific way of going wrong, and in every case the wrong answer looks perfectly reasonable.

Percentages: four different questions wearing one word

A share of a number. 20% of 80 is 16. This is the one people mean when they say percentage, and the only one that causes no trouble.

One number as a share of another. 16 out of 80 is 20%. Same numbers, different question, and the reverse of the first.

The change between two numbers. This is where it breaks. Going from 200 to 150 is a 25% fall, but recovering it takes a 33.3% rise, not 25%. Percentage changes do not reverse symmetrically, because each is measured against a different base. A 50% drop needs 100% to get back; a 90% drop needs 900%.

Reversing a percentage. If a price including 20% tax is 120, the pre-tax figure is not 96. It is 100, because you divide by 1.2 rather than subtracting 20%. Getting this backwards is the most common error in retail arithmetic.

The percentage calculator keeps these four apart deliberately rather than merging them into one box, and the percentage change calculator shows the reverse figure alongside the change, because that asymmetry is the part worth seeing.

Averages: there are three, and they are not interchangeable

The arithmetic mean — add them up, divide by how many — is right for quantities that add together. Test scores, heights, daily takings.

The geometric mean is for anything that compounds. Average a +50% year and a −50% year arithmetically and you get zero, implying you broke even. You did not: 1.5 × 0.5 is 0.75, a 25% loss. Growth rates, investment returns and inflation all belong here.

The harmonic mean is for rates over a fixed quantity. Drive one leg at 60 mph and an equal leg at 30, and your average speed is 40, not 45, because you spend twice as long on the slow leg. The harmonic mean calculator does that weighting automatically.

The three always order the same way: harmonic ≤ geometric ≤ arithmetic, equal only when every value is identical. So reaching for the arithmetic mean by default does not merely risk a wrong answer — it reliably gives one that is too high.

Separately, there is the question of which average summarises a set honestly. When the mean and the median disagree noticeably, outliers are dragging the mean and the median is the better description. The average calculator returns both, which makes the disagreement visible.

Rates: the compounding nobody plans for

Anything that grows on its own previous total compounds, and compounding runs ahead of intuition in both directions. 1,000 at 7% for ten years is 1,967, not the 1,700 that ten lots of 7% suggests.

The useful shortcut is the rule of 72: divide 72 by the rate to get the doubling time. At 8% that gives 9 years against an exact 9.006. It drifts as rates rise — at 25% it is off by about 7% — but for ordinary rates it is close enough to do in your head.

The same arithmetic runs in reverse on debt. A credit card balance feels immovable because the first payment is mostly interest: on 5,000 at 19.99%, 83 of a 150 payment goes to interest and only 67 touches the balance. That ratio improves every month, which is why progress feels glacial and then suddenly is not.

Try it on the compound interest calculator, and note what changing the compounding frequency does. A nominal 6% compounded monthly is 6.17% effective, and that gap is why quoted rates and actual returns disagree.

Space: the units, always the units

Area and volume calculations are arithmetically trivial and go wrong constantly, almost always at a unit boundary.

Measure a room in inches and multiply, and you have square inches — divide by 144, not 12, to reach square feet. Volume is worse: a cubic yard is 27 cubic feet, not 3. The conversion factor gets squared or cubed along with the measurement, so forgetting it is an error of 12× or 27×, not 12% or 27%.

The other trap is which dimension a formula wants. A cone's volume uses its vertical height while its surface area uses the slant height, and substituting one for the other produces a plausible number that is simply wrong.

For rooms and floors, the square footage calculator handles the conversion. For anything three-dimensional, the volume calculator covers six shapes and shows the litre equivalent, since that is usually the number actually wanted.

The habit that catches all four

State the units of the answer before starting. Square feet, pounds per month, miles per hour. If the arithmetic cannot produce those units, the method is wrong regardless of what the calculator returns.

Then sanity-check the magnitude. Should the answer be larger or smaller than what you started with, and roughly by how much? Most calculation errors are off by a factor of ten or a hundred rather than by a few percent, and a rough expectation catches those instantly while a misplaced decimal never announces itself.

That is genuinely the whole method. The formula is the easy half.