Percentage Points vs Percent: The Mistake That Trips PMs Up on Experiment Results
If completion goes from 40% to 42%, how much did it go up?
Two answers are both correct, and they mean different things: up 2 points, or up 5 percent. Confusing them is the single most common way smart PMs sound confused when talking about experiment results, in interviews and in real readout meetings. The word "percent" does double duty: it is the unit of the metric and the unit of the change. Your ear cannot tell which one someone means, so you have to say it explicitly.
This guide gives you the distinction, the habit that prevents the mistake, and worked examples you can reuse.
The distinction, in one paragraph
Percentage points measure the arithmetic difference: subtract the old rate from the new rate. 42% minus 40% is 2 points.
Percent (relative change) measures the difference relative to where you started: divide the point change by the starting rate. 2 divided by 40 is 0.05, or 5%.
Both numbers describe the same result. Points tell you the absolute size of the move. Percent tells you the move relative to the baseline. A 2-point move on a 40% base is a 5% lift. The same 2-point move on a 4% base is a 50% lift. Same absolute change, wildly different relative change. That is why you need both.
The habit: say it both ways
Every time you report a result on a percentage metric, state it in points and in relative terms, in the same sentence. "Booking completion went from 20% to 23%, up 3 points, a 15% relative lift." It takes five extra words and it removes an entire category of confusion: nobody has to guess which unit you meant, and nobody can misquote you later as having claimed the bigger number.
This habit also protects you in the other direction. When someone else reports a result, your first question should be: "Points or percent?" If a stakeholder says "conversion is up 20%," you do not know whether they mean 20 points (enormous) or 20% relative (meaningful but ordinary) until you ask. In an interview, asking that question out loud is a signal by itself: it shows you know the trap exists.
Worked examples
All scenarios below are fictional, with round numbers chosen to make the arithmetic visible.
Example 1: booking completion
A travel booking app tests showing the full price up front (taxes and fees included) versus showing the base fare first with fees at checkout. After 4 weeks on a large sample, searches turning into a completed booking: 23% in test, 20% in control.
- Points: 23 minus 20 = 3 points.
- Relative: 3 divided by 20 = 15%.
- Say it both ways: "Booking completion rose 3 points, from 20% to 23%, a 15% relative lift."
Notice what happens if you only say "up 15%." A listener could reasonably hear "up 15 points," which would mean 35%, a result nearly five times larger than what actually happened. The both-ways sentence makes that misreading impossible.
Example 2: revenue per booking moves the other way
Same test. Average revenue per completed booking: $186 in test, $192 in control.
- Absolute: 186 minus 192 = negative $6.
- Relative: negative 6 divided by 192 = negative 3.1%.
- Say it both ways: "Revenue per booking fell $6, from $192 to $186, down about 3%."
Now the full picture is visible: bookings up 3 points, revenue per booking down about 3%. Whether the test is a win depends on the total, which is the next calculation a PM should do, not a vibe. (Roughly: 23 bookings at $186 is $4,278 per 100 searches; 20 bookings at $192 is $3,840. The test wins on total revenue even though revenue per booking fell. That is a coexistence worth stating: both facts are true at once.)
Example 3: the small-base trap
A feature lifts a rare conversion from 1% to 2%.
- Points: 1 point.
- Relative: 1 divided by 1 = 100%.
- Say it both ways: "Conversion doubled, from 1% to 2%, up 1 point."
"Doubled" is true and it is also the kind of sentence that gets a PM in trouble, because it sounds enormous while the absolute move is 1 point. The both-ways habit forces the honest version into the same sentence. In an interview, the candidate who says "doubled, but that is 1 point off a small base" sounds more trustworthy than the candidate who just says "doubled."
Example 4: the shrinking-base trap
Churn falls from 10% to 8%.
- Points: 2 points.
- Relative: 2 divided by 10 = 20%.
- Say it both ways: "Churn fell 2 points, from 10% to 8%, a 20% relative reduction."
Again, "churn down 20%" alone invites the misreading "down 20 points," which would mean negative churn, which is nonsense. Both ways, every time.
Three traps beyond the basic confusion
Trap 1: averaging percentages. If conversion is 10% in one segment and 30% in another, the overall rate is not 20% unless the segments are equal size. Weight by the denominator. This is the same composition problem that makes segment-level analysis matter, and it bites hardest when segment sizes shift between test and control.
Trap 2: points and percent in the same sentence without labels. "Conversion rose 3%, or 15% in relative terms" is not a both-ways sentence, it is a riddle. Label every number: "3 points (15% relative)." Labels are the whole habit.
Trap 3: rounding away the result. A move from 40.1% to 40.4% is 0.3 points. Rounded to whole numbers, both are "40%," and the result vanishes. When the base is large and the move is small, carry a decimal. The alternative is reporting "no change" for a real, statistically significant result.
How this shows up in interviews
Interviewers test this in two ways. The direct way: they hand you numbers and ask what happened. The indirect way: they listen to how you report numbers in a case answer, and they notice whether you label your units.
For the direct version, the answer format is mechanical:
- State both rates with their labels ("control 20%, test 23%").
- State the point change ("up 3 points").
- State the relative change ("a 15% relative lift").
- State the implication in one sentence ("total revenue per 100 searches rises, because the booking gain outweighs the revenue-per-booking decline").
For the indirect version, the habit is the answer. A candidate who says "up 3 points, 15% relative" throughout a case sounds precise without sounding pedantic, because the habit reads as fluency, not as showing off.
Practice: five reps
Compute points and relative change for each, then write the both-ways sentence.
- Activation 25% to 30%.
- Retention 60% to 57%.
- Click-through 2% to 2.5%.
- Trial conversion 12% to 15%.
- Support contact rate 8% to 6%.
(Answers: 1. 5 points, 20% relative. 2. Negative 3 points, negative 5% relative. 3. 0.5 points, 25% relative. 4. 3 points, 25% relative. 5. Negative 2 points, negative 25% relative.)
If any of those felt slow, that is the point. The translation is a small alphabet, and reps are what make it automatic. Five reps a day for a week, and the both-ways sentence becomes the default, not a performance.