Percentage Change vs Percentage Difference

The short answer

A percentage is a division, and the only hard part of a percentage is deciding what goes on the bottom. That number is the base, and it is the whole argument.

Take 100 and 150. The gap between them is 50, whichever way round you write it. But:

• 100 → 150 is a 50 per cent increase, because the base is 100.
• 150 → 100 is a 33.33 per cent decrease, because the base is 150.
• The percentage difference between them is 40 per cent, because that measure uses their average, 125, as the base.

One gap, three correct answers, and the difference between them is entirely the denominator. Our percentage change calculator divides by the old value and labels the direction; the percentage calculator answers the two plain questions, what is X% of Y and X is what % of Y; the CAGR calculator is the one you want when the change happened over several years rather than in one step. Everything runs in your browser.

What follows is the rule behind each of them, worked by hand so you can check a figure rather than trust one. It is arithmetic, not advice.

One pair of numbers, three right answers

The clearest way to see the three measures is to apply all of them to a pair where all three are legal: two percentages, say a figure that moved from 10 per cent to 15 per cent.

Every row is right. They answer different questions.

Percentage change assumes one of the two numbers came first. It has a direction, and the earlier value is the base. Budget against actual uses the budget. Year on year uses last year. Before and after uses before. If you can say which number came first, this is your measure.

Percentage difference assumes neither came first. Two suppliers' quotes, two instruments reading the same tank, two branches in the same month — there is no "from" and no "to", so using either number as the base would privilege it arbitrarily. Dividing by the average treats them alike, which is why the answer is 40 per cent whichever order you write them in. Percentage change is not symmetric; percentage difference is. That is the point of it.

Percentage points is not a percentage at all. It is the plain arithmetic gap between two percentages, and it needs no base because nothing is being divided. It exists because the alternative sentence is ambiguous to the point of uselessness.

Percentage points: the word that ends the argument

Here is the sentence that causes the trouble: "conversion went up by 50 per cent."

If the conversion rate moved from 2 per cent to 3 per cent, that sentence is true — the relative change is +50 per cent. It is also how you would describe a move from 2 per cent to 52 per cent, which is a different event entirely. Both are "up by 50", and a reader cannot tell which you mean.

The convention that fixes it is one word:

• "Up 1 percentage point" — 2 per cent to 3 per cent. The gap between two percentages.
• "Up 50 per cent" — 2 per cent to 3 per cent, relative to the 2. The same move, measured against its base.
• "Up 50 percentage points" — 2 per cent to 52 per cent. Unambiguous, and obviously a different claim.

So say points for the gap and per cent for the relative change, and never the bare "per cent" when both of your numbers are already percentages. This is not pedantry about language; it is the only way to write the sentence so that one reading is possible.

The asymmetry runs the other way too. A single percentage point is enormous on a small base and trivial on a large one. One point added to a 2 per cent conversion rate is half the business again. One point added to a 95 per cent uptime figure is barely a rounding adjustment on the same arithmetic. A report that quotes only points hides the first; a report that quotes only relative change hides the second. Quote both, which costs you four words.

Converting a number into a percentage is a different job

Worth separating out, because it gets filed under the same heading and it is not the same operation. Measuring a change needs two values and a decision about which is the base. Expressing a single quantity as a percentage needs one division and no decision at all — the whole is the base, and it is handed to you by the question.

All of these are the second row of the percentage calculator — X is what % of Y — and the first row, what is X% of Y, is the same identity run backwards. If you are converting a percentage online rather than measuring a movement, that tool is the one you want, and the conversion is exact: there is no base to choose and therefore nothing to get wrong except which number is the whole.

The one trap in this group is that percentages of percentages do not add. Two successive discounts of 20 and 10 per cent leave 0.8 × 0.9 = 72 per cent of the original price, so the total discount is 28 per cent, not 30. Each percentage is taken on a different base, and the second base is already smaller.

The asymmetry that costs real money

Because percentage change uses the starting value as the base, a fall and the rise that undoes it are never the same percentage. The fall shrinks the base, so the recovery is measured against a smaller number and has to be bigger.

A holding that is down 50 per cent is not halfway home. It needs to double. A holding down 90 per cent needs a tenfold recovery, which is why "it only has to go back up 90 per cent" is one of the more expensive sentences in investing.

The formula for that column, if you want it: a fall of d (as a decimal) needs a rise of d ÷ (1 − d). And the same asymmetry is why averaging percentage changes misleads. Up 50 per cent then down 50 per cent "averages" zero and leaves you with 0.75 of what you started with — a 25 per cent loss. The arithmetic mean of the two changes is meaningless because each was taken on a different base.

When the base is negative

This is where most calculators, and most spreadsheets, quietly produce the wrong answer — not a wrong magnitude, a wrong direction.

The textbook formula is (new − old) ÷ old × 100. If the old value is negative, that denominator is negative, and the sign of the whole result flips. A business whose quarterly loss narrowed from −200 to −100 has unambiguously improved. Put those numbers through the signed formula and you get −50 per cent: "a 50 per cent decrease", which reads as bad news about good news.

The fix is to divide by the absolute value of the old value, which is what the formula in the table above does and what our tool computes:

The last row is the reason this is a safe default rather than a special case: for positive bases the two formulas are the same formula. Dividing by the absolute value changes nothing you already trusted, and it stops the sign flipping on the one case where you most need the direction to be right.

Two honest limits on it, though.

The direction word follows the arithmetic, not your P&L. Our calculator says "Increase" whenever the value went up, and in a loss line a value going up is the improvement. That is consistent and it is right, but the label is describing the number, not the business. In a report, write the sentence yourself: "the loss narrowed by 50 per cent" says in five words what no signed percentage can.

When the two values straddle zero, consider not using a percentage at all. A result moving from −50 to 150 is a 400 per cent change on a base of 50, and that figure is arithmetically impeccable and almost impossible to interpret. The swing is 200. Report the 200. Percentages compare magnitudes, and a magnitude that passed through zero on the way has no stable scale to be compared against.

When the base is zero

There is no percentage change from zero. Not a large one — none. Division by zero is undefined, so the question has no answer, and our tool asks for a non-zero old value rather than inventing one. A spreadsheet will hand you #DIV/0! for the same reason.

This matters more than it sounds, because "from zero" is exactly the situation people most want a growth percentage for. Five sign-ups in a month that previously had none is a real and reportable event. It is "five sign-ups", or "the first five". It is not +500 per cent, and it is not infinite growth. If your dashboard shows a percentage there, something in it is guessing.

The usable rule: when the base is zero or near enough to it that one unit swings the figure by hundreds of per cent, report the absolute numbers and let the reader do the comparison.

When the change took several years

One more base problem, and it is the one that produces the most confident wrong numbers in a pitch deck. Percentage change measures a single step from one value to another. It does not care how long that step took, and it must not be divided by the number of years to get an annual rate.

An investment that went from ₹1,00,000 to ₹2,50,000 over five years has grown 150 per cent in total. Divide that by five and you get "30 per cent a year", which is wrong, because it ignores that each year's growth compounds on the last. The honest annual figure is the compound annual growth rate: (final ÷ initial) raised to the power of 1 ÷ years, minus one. Here that is 20.11 per cent a year, nearly ten points below the naive answer.

That is what the CAGR calculator computes, and it is the only form in which two returns over different periods can be compared. "Doubled in five years" is 14.87 per cent a year. "Up 60 per cent in three years" is 16.96 per cent a year. Side by side as headline figures the first sounds better; annualised, the second one is.

CAGR also fixes the averaging trap from earlier. Up 50 per cent then down 50 per cent has a CAGR of −13.4 per cent a year over the two years, which correctly describes money that shrank to 0.75 of itself. The arithmetic average of the two changes, zero, describes nothing.

What CAGR cannot tell you is how rough the ride was. A smooth 12 per cent a year and a wild sequence of years ending at the same value have identical CAGR; the measure compresses the journey into the destination, on purpose. Treat it as a comparison tool, not a description of the experience.

What the calculators assume, exactly

Worth being precise, because a vague answer here is the thing you came to this page to escape.

The percentage change calculator takes a from and a to, applies (new − old) ÷ |old| × 100, and shows the signed percentage rounded to two decimal places alongside the absolute change. It requires a non-zero old value, for the reason above. The word beside the figure — Increase or Decrease — is chosen by the sign of the result, so read it as a statement about the number and not about your accounts.

The percentage calculator runs both plain forms side by side and rounds to four decimal places, which is enough to show you that 34 of 120 is 28.3333 per cent rather than a tidy 28.3 — useful when a set of shares has to add to a hundred.

The CAGR calculator needs a starting value, an ending value and a number of years, all greater than zero. It returns nothing for a negative or zero start, and that is deliberate rather than a gap: a fractional power of a negative ratio is not a growth rate, and a tool that printed one anyway would be making something up. If your series starts below zero, CAGR is the wrong instrument.

None of the three sends your numbers anywhere. They compute in the browser, which is the general rule for the whole calculator set.

Start here

1. Decide whether one of your two numbers came first. If yes, you want percentage change, and the earlier one is the base. If no, you want percentage difference, and the base is their average. Settling this takes a second and removes most of the ways the figure can be wrong.
2. If both numbers are already percentages, write "points" for the gap. Then give the relative change as well if it is material. Two measures, one sentence, no ambiguity.
3. Check the sign of your base before trusting a direction. Any series that contains a negative — profit, net cash, net change — needs the absolute-value denominator, or an improvement will show up as a decline.
4. If the base is zero, report the absolute numbers. There is no percentage there to report, and a dashboard that shows one is guessing.
5. If the change spans more than one period, annualise it. Total change divided by the number of years is not an annual rate. Use CAGR, and compare only annualised figures with other annualised figures.
6. Say the base out loud in the sentence. "Up 50 per cent on last year" cannot be misread. "Up 50 per cent" can.

BitsSolved builds web, mobile and SaaS products, and the reason this site has a calculator directory on it is that most of these tools began as something a client's product needed to get right. If you are shipping anything that shows a user a percentage — a dashboard delta, a growth figure, a discount, a conversion rate — the lesson here is the one worth carrying into the spec rather than discovering in a bug report: the formula is three characters of arithmetic, and every defect lives in the choice of denominator. Write down which value is the base, what happens when it is zero, and what happens when it is negative, before anyone writes the function.