Percentage Change: The Formula and Four Ways It Misleads

Percentage change is (new − old) / old, multiplied by 100. The old value is the base, and getting the base wrong is the most common error in the whole topic. Going from 40 to 50 is a 25 percent increase, because the base is 40. Going from 50 back to 40 is a 20 percent decrease, because the base is now 50. Same two numbers, same absolute gap of 10, two different percentages. Whenever a percentage is quoted without saying what it is a percentage of, it cannot be checked.

If a conversion rate moves from 4 percent to 5 percent, that is a rise of one percentage point and a rise of 25 percent. Both statements are true and they describe the same event. Reporting "conversion rose 25 percent" is technically correct and routinely misread as a much larger change than one point. When the quantity being measured is itself a percentage, say which unit you are using; the ambiguity is where most misleading dashboards come from.

A 50 percent fall requires a 100 percent rise to get back to where you started, because the base shrank. Halve 100 and you get 50; to return to 100 from 50 you must add 50, which is all of the new base. This asymmetry gets worse as the drop deepens: an 80 percent fall needs a 400 percent rise to recover. It is why a portfolio or a metric that oscillates by the same percentage each period drifts downward rather than staying flat.

Applying +10 percent and then −10 percent does not return the original value. Multiply by 1.10 and then by 0.90 and you get 0.99, a net loss of one percent. Chained percentage changes must be multiplied as factors, not summed. Over many periods this compounds: twelve alternating ±10 percent moves leave you about 5.7 percent below where you began, even though the changes appear to cancel.

If the old value is zero, percentage change is undefined — the denominator is zero, and no percentage describes a move from nothing to something. Report the absolute change instead. Negative bases are worse, because the sign of the result flips in ways that are almost always misread: going from −10 to −5 is an improvement, but the formula returns −50 percent. For any series that crosses zero, absolute differences are the honest presentation.

The mean of two percentage changes is not the overall percentage change unless the two bases are identical. If one region grew 50 percent from a base of 10 and another grew 5 percent from a base of 1000, the unweighted average of 27.5 percent describes nothing real; the combined growth is about 5.4 percent. Aggregate the underlying totals and compute one percentage from them, rather than averaging percentages that came from different denominators.

A percentage without its base is not a complete statement. Revenue rose 12 percent is useful only if the reader also knows whether the base was last month, last year, or a forecast. When presenting sequential changes, show both the factor chain and the final absolute value so the reader can verify the arithmetic. When the quantity is itself a percentage, label the unit as percent or percentage points every time it appears, including in chart axes and table headers.