What is a Confidence Interval? Definition, Examples & Best Practices
A confidence interval in statistics gives a range around a sample estimate, such as a survey percentage, mean score or conversion rate. Instead of saying that 42% of respondents prefer a product, a researcher might report 42% with a 95% confidence interval from 39% to 45%.
The confidence level describes the long-run reliability of the method. A 95 percent confidence interval does not mean there is a 95% probability that the true value is inside one finished interval. It means that if the same sampling process were repeated many times, about 95% of the intervals produced by that method would contain the true population value.
In survey research, confidence intervals are used for proportions, means, brand tracking results, customer satisfaction scores, product testing and biostatistics. Wider confidence limits usually mean less precision. Larger samples, lower variation and simpler sampling designs usually produce narrower intervals.
Why Confidence Interval Matters
The modern confidence interval is usually credited to Jerzy Neyman. Neyman (1937) set out the frequentist theory of interval estimation, where the procedure has a known long-run coverage rate. Student, the pen name of William Sealy Gosset, introduced the t distribution in 1908, which became central to confidence intervals for means when population variance is unknown. Cochran (1977) later gave applied survey researchers practical methods for sampling error, sample size and interval estimation.
For a 95 confidence level under a normal approximation, the lower and upper confidence limits are usually calculated as: estimate plus or minus 1.96 multiplied by the standard error. The Office for National Statistics states this exact 95% confidence interval approach in its guidance on survey uncertainty. Pew Research Center reported in April 2025 that a survey with 5,410 respondents had a margin of sampling error of plus or minus 1.6 percentage points, while also reporting a cumulative response rate of 3%.
A simple worked example is: survey estimate 60%, standard error 2%. The 95% confidence interval is 60% plus or minus 1.96 multiplied by 2%, which gives 60% plus or minus 3.92%. The confidence limits are 56.08% and 63.92%.
The weakness is interpretation. Morey et al. (2016) argued that confidence intervals are often misread as probability statements about the realised interval. They also do not fix poor sampling, biased question wording, nonresponse or unrepresentative panels.
How to Calculate a Confidence Interval
The method depends on what is being estimated. For a survey proportion, such as the share of customers who prefer a new packaging design, a common large-sample formula is:
Confidence interval for a proportion = p plus or minus z multiplied by the square root of p(1 minus p) divided by n.
In this formula, p is the sample proportion, n is the sample size and z is the critical value for the confidence level. For a 95 confidence level, z is usually 1.96. For a 99 confidence interval, z is about 2.576, which makes the interval wider.
Worked example: a packaging concept test surveys 1,000 category buyers. 620 prefer Concept A, so p = 0.62. The standard error is the square root of 0.62 multiplied by 0.38 divided by 1,000, which is about 0.0153. The 95% margin of error is 1.96 multiplied by 0.0153, which is about 0.030. The 95 percent confidence interval is 62.0% plus or minus 3.0 percentage points, or 59.0% to 65.0%.
For a confidence interval for mean, the usual formula is: sample mean plus or minus critical value multiplied by standard error of the mean. If the sample is small and the population standard deviation is unknown, researchers usually use the t distribution rather than the normal z value.
How Confidence Interval Is Used in Practice
A UK consumer app team surveys 1,200 active users after testing a redesigned onboarding flow. The survey finds that 684 users, or 57%, say the new flow is easier to complete than the old version.
Using a 95% confidence interval for a proportion, the standard error is about 1.43 percentage points. The 95% margin of error is about 2.8 percentage points, giving confidence limits of 54.2% to 59.8%.
The product team compares this with its success threshold of 55%. Since the lower confidence limit is 54.2%, the result is positive but not comfortably above the threshold. Instead of rolling the design to all users, the team launches to 25% of traffic, adds two more weeks of data collection and checks whether the interval narrows above 55%. The decision is cautious because the interval shows uncertainty around the estimate.
Sources
Cochran, W. G. (1977). Sampling Techniques. John Wiley & Sons.
Morey, R. D., Hoekstra, R., Rouder, J. N., Lee, M. D. and Wagenmakers, E. J. (2016). “The Fallacy of Placing Confidence in Confidence Intervals.” Psychonomic Bulletin & Review.
Neyman, J. (1937). “Outline of a Theory of Statistical Estimation Based on the Classical Theory of Probability.” Philosophical Transactions of the Royal Society of London.
Office for National Statistics (accessed 2026). “Uncertainty and How We Measure It for Our Surveys.” Office for National Statistics.
Pew Research Center (2025). “U.S. Public and AI Experts Methodology.” Pew Research Center.
Student (1908). “The Probable Error of a Mean.” Biometrika.
National Institute of Standards and Technology (accessed 2026). “What Are Confidence Intervals?” NIST/SEMATECH e-Handbook of Statistical Methods.
Frequently Asked Questions
A confidence interval is a range around a sample estimate that shows statistical uncertainty. In survey research, it is often used to estimate a population percentage or mean from a sample. A result of 48% with a 95% confidence interval of 45% to 51% means the estimate is not a single exact number.
Confidence limits are the lower and upper ends of a confidence interval. If a survey result is 62% with a 95% confidence interval from 59% to 65%, then 59% is the lower confidence limit and 65% is the upper confidence limit. They show the plausible range produced by the sampling method.
A confidence interval is the range of values, such as 40% to 46%. A confidence level is the long-run coverage rate attached to the method, such as 95% or 99%. A higher confidence level usually creates a wider interval because the method is designed to capture the true value more often.
A 95% confidence interval means that if the same sampling process were repeated many times, about 95% of the intervals created would contain the true population value. It does not mean there is a 95% probability that the true value is inside the one interval already calculated.
A confidence interval for a proportion estimates the range around a percentage. For example, if 400 out of 1,000 respondents prefer a product, the sample proportion is 40%. A 95% confidence interval adds sampling uncertainty around that 40%, helping researchers avoid treating the sample result as exact.
A confidence interval for a mean estimates the range around an average, such as average satisfaction score, average spend or average task time. It uses the sample mean, variation in the data, sample size and a critical value. If the sample is small, researchers commonly use the t distribution.
A 99 confidence interval is wider because it is designed to capture the true population value more often across repeated samples. To do that, the formula uses a larger critical value. The trade-off is lower precision. Researchers usually choose 95% for standard reporting and 99% when uncertainty needs stricter treatment.
A confidence interval calculator can compute the interval, but it cannot judge whether the sample is representative, whether weighting is needed, whether the question was biased or whether the sampling design changes the standard error. Good interpretation still depends on research design, fieldwork quality and the decision being made.