How To Calculate Percent Decrease

Percent decrease answers one question: how large was the fall, measured against the level it fell from. The calculation has three moves. Subtract the new value from the original to get the absolute drop, divide that drop by the original, then multiply by 100 to turn the ratio into a percent. Only the division ever causes disagreement, because that is where the answer gets anchored to a base. Taking 120 down to 90, the drop is 30, the base is 120, and the decrease is 25.00%. Multiplying by 100 is the one step that can never be the source of a wrong answer: it is bookkeeping that shifts the decimal point two places.

The original value belongs in the denominator because it is the only figure capable of giving the drop a scale. A fall of 30 is a quarter of a quantity of 120 and a rounding error in one a thousand times larger, and the division handles that difference automatically. Choosing the smaller number because it feels like the floor, or the newer number because it happens to be the one in hand, breaks the meaning of the result even though the arithmetic still runs. The rule is chronological: the denominator is whatever stood there before the change, whichever of the two values happens to be larger. Writing both figures down with their dates beside them, before a calculator is touched, removes most of that risk.

Dividing by the new value is the mistake worth naming explicitly, because it always inflates the answer and always looks plausible. Taking 120 down to 90 and computing 30 / 90 gives 33.3333%, which is not the decrease at all but the increase required to climb back from 90 to 120. The two figures answer different questions. One asks how much was lost relative to where the quantity began; the other asks how much must be gained relative to what is left. Since the new value is smaller after any fall, using it as the divisor systematically overstates the drop, and the overstatement grows as the fall gets deeper. One tell is quick to apply: the true decrease can never exceed the figure produced by the wrong divisor, so an unexpectedly generous drop deserves a second look.

That same mechanism makes the two directions permanently unequal. Climbing from 80 to 100 is a 25.00% increase, while falling from 100 back to 80 is a 20.00% decrease. The identical movement of 20 units is being measured against two different bases, and the base for the return leg is the endpoint of the outward one. Nothing is inconsistent; a percentage simply has no meaning until you state which value it is taken of. This is why quoting a decrease without naming its reference figure invites an argument that no amount of recomputation will settle, and why one bare percentage cannot establish who gained and who lost.

Sequential moves compound, which is the practical cost of that asymmetry. Cut 120 by 25% and you are left holding 90. Raise that 90 by 25% and you arrive at 112.5, because the rise acts on 90 rather than on 120. The shortfall is 7.5 units, still 6.25% below where the sequence began. Equal and opposite percentage moves therefore never restore the starting value, and the damage grows with the size of the move: small swings are nearly harmless, large ones leave a gap that is hard to recover. Anyone planning a recovery should compute the required rise against the reduced figure first, and expect it to exceed the fall it is meant to undo.

The multiplicative view is often the clearer way to carry the same information. A 25.00% decrease leaves a multiplier of 0.75 behind, which is another way of saying that 90 is 0.75 of 120. Framing it this way makes chaining trivial: successive reductions multiply, so one fall followed by another is the product of the two multipliers, never their sum. It also hands you a free check on the answer, because the decrease and the remaining share must add to 1. A result that fails that test has been divided by the wrong figure. The habit also makes stacked reductions easy to audit, because the product of the multipliers is the only number worth quoting.

Stacked discounts are the everyday version of the same arithmetic, and they are the reason a combined saving can look generous without being quite what it appears. A 20% reduction followed by a 10% reduction multiplies 0.80 by 0.90 and lands on 0.72, a total decrease of 28% rather than the 30% that adding the two figures suggests. The shortfall is not a rounding slip: the second discount is applied to an already reduced price, so it removes less in absolute terms than it would have done on the original. Any chain of reductions behaves this way, and reversing the order of the two cuts makes no difference to the product. Publishing the combined multiplier next to the headline percentage is the clearest way to keep both sides of the claim honest.

A decrease is bounded in a way an increase is not. The deepest fall expressible without negative values is 100%, reached when the new value is zero, so 120 down to 0 is a 100% decrease and beyond that point the percentage stops carrying information. Increases have no ceiling at all, since a value may double, triple or grow by any multiple. That asymmetry of range is worth remembering whenever a figure looks suspiciously large: a reported decrease above 100% almost always means a negative original value or a division by the wrong base, and should be treated as a prompt to re-check the inputs rather than as a dramatic result to publish.

Two edge cases deserve a pause. When the original value is zero there is no base to divide by, and the percent decrease is undefined rather than infinite, so report the absolute movement and describe it in words. When the original is negative the ratio flips sign and reads backwards, presenting an improvement as a decline, which is worse than useless in a report. Finally, keep percent and percentage points apart: a rate falling from 25% to 20% has dropped by 20% of itself, which is a different statement from the plain subtraction of the two rates, and the two descriptions are routinely confused in headlines. Both statements can be true of the same event at once, so an honest report names which of the two it is using.