Reading a confidence interval without being a statistician

When a study tells you "62% purchase intent," the right first question isn't "is that a lot?" It's "give or take how much?" A percentage on its own means little until you know its margin.

A result is a range, not a point

No measurement is exact. With a panel, you survey a sample, not the whole population. The true number therefore sits somewhere around the one you read. The confidence interval is that range.

In practice, "62% [58 – 66]" reads like this: the measurement lands on 62%, and the real value most likely sits between 58 and 66%. The point in the middle is the most probable estimate; the bounds tell you how far you could reasonably be off.

Why it helps you decide

The interval completely changes how you read a comparison.

• Concept A at 62% [58 – 66] versus concept B at 60% [56 – 64]: the ranges overlap heavily. Calling A the "winner" would be a stretch. The two are neck and neck.
• Concept A at 62% [60 – 64] versus concept B at 51% [49 – 53]: here the ranges don't touch. The gap is solid, you can call it.

Without the interval, you would have treated both cases the same. With it, you know when a gap deserves a decision and when caution is warranted.

What makes a range tighter

Two main levers. First, panel size: the more respondents, the tighter the range. Going from 100 to 500 respondents shrinks the margin noticeably. Second, how spread out the answers are: a divided audience yields a wider range than a unanimous one.

That's also why aiming for a huge panel "to be safe" has diminishing returns. Past a certain point, doubling the size barely tightens the margin.

The habit to keep

Never read a number without its range, and don't celebrate a one-point gap. Just ask yourself: do the intervals overlap? If yes, be careful. If no, you have a real signal.

At Panelia, every result is shown with its interval, precisely to avoid false precision. A good tool doesn't hand you a reassuring number, it hands you an honest one.