The Attribution Model Confidence Interval Calculator estimates a plausible range around an observed attribution-model proportion, such as the share of conversions assigned correctly in a validation sample. It is useful when a single observed percentage feels too precise and you need to communicate sampling uncertainty alongside the point estimate. Enter the number of validated outcomes, the number counted as successful under your evaluation rule, and the confidence level. The calculator reports the observed rate, an approximate normal confidence interval, margin of error, and standard error. Use the interval to compare model evaluations more cautiously, especially when sample sizes differ.
1. Count evaluated outcomes
Enter the total number of attribution outcomes reviewed under one consistent validation rule.
2. Count successful cases
Enter how many of those outcomes met your definition of a successful or acceptable attribution.
3. Choose confidence
Select 90%, 95%, or 99%. Higher confidence produces a wider interval.
4. Review the interval
Use the point estimate together with the lower and upper bounds rather than relying on the observed rate alone.
Where:
• p = observed successful-attribution proportion
• SE = standard error of the observed proportion
• z = normal critical value for the selected confidence level
• total = number of validated attribution outcomes
Assumptions: This page uses the normal approximation for a binomial proportion and clips interval bounds to 0% and 100%. For very small samples or proportions near 0% or 100%, an exact or Wilson interval may be more appropriate.
A narrower interval indicates greater precision around the observed attribution success rate, while a wider interval indicates more sampling uncertainty.
This interval describes uncertainty from the sampled outcomes under the stated model; it does not capture systematic bias in the attribution method or validation labels.
Given: 1,200 validated conversions, 888 successful attributions, 95% confidence.
Calculation: p = 888 / 1,200 = 0.7400. SE = √(0.74 × 0.26 / 1,200) ≈ 0.01267. Margin = 1.959964 × 0.01267 ≈ 0.02484.
Result: 74.00% with an approximate 95% confidence interval of 71.52% to 76.48%.
Interpretation: The observed validation rate is 74%, but sampling uncertainty supports reporting a range of roughly 71.5%–76.5%.
Smaller samples produce a larger standard error, so the same observed rate is estimated less precisely. Increasing the validated sample usually narrows the interval.
Use the level that matches your reporting standard and tolerance for uncertainty. Higher confidence creates a wider interval because it covers a broader range of plausible values.
This calculator expects a binary count: each evaluated outcome is classified as success or not success. A continuous revenue-share metric needs a confidence interval based on its own variance rather than this proportion formula.
The normal approximation can be less reliable at extreme proportions, particularly with small samples. Consider a Wilson or exact binomial interval when boundary behavior matters.
This page quantifies uncertainty around a success proportion. An error-rate estimator focuses on the observed share of incorrect or disputed outcomes and may be used as a monitoring metric.