Turn experiments into uplift metrics you can trust. Validate changes with statistical testing. Download clear reports for stakeholders in minutes.
| Group | Visitors | Conversions | Conversion rate | Revenue/Conversion |
|---|---|---|---|---|
| Control | 10,000 | 520 | 5.20% | 35.00 |
| Variant | 10,000 | 575 | 5.75% | 35.00 |
| Result | — | +55 | +0.55 pp | +1,925.00 (approx.) |
Use visitors as the exposure count and conversions as the success count. Exclude bots, duplicate sessions, and post‑purchase refreshes. Keep tracking rules consistent across groups. When segmentation is required, rerun the calculator per segment rather than mixing rates. This preserves interpretability and avoids Simpson’s paradox. For example, 10,000 visitors with 520 conversions implies a 5.20% baseline.
Absolute uplift reports the percentage‑point change, which maps directly to incremental conversions at a fixed traffic level. Relative uplift scales the change by the baseline rate, which helps compare experiments across funnels. For low baseline conversion, even small absolute gains can produce large relative lifts. A 0.55 pp gain from 5.20% to 5.75% equals about 10.6% relative.
The z‑test summarizes how surprising the observed difference is under a no‑effect assumption. The confidence interval gives a plausible range for the true uplift and is often more decision‑useful than a single p‑value. If the interval crosses zero, the effect may be positive or negative at the chosen confidence. Narrow intervals require higher traffic and stable measurement.
Incremental conversions estimate the additional outcomes the variant produced versus the control rate applied to variant traffic. Adding revenue per conversion converts uplift into expected value. When revenue differs by group, use separate values and interpret the result as an approximation of realized value, not guaranteed profit. Always sanity‑check with metrics like refunds and churn.
A statistically significant result can still be too small to matter. Compare incremental conversions against operational costs and capacity. Longer tests reduce variance but increase opportunity cost. If results are noisy, consider extending duration or consolidating variants to concentrate traffic and improve precision. As a rule, avoid stopping early based on one day.
Export CSV for dashboards and PDF for stakeholder team updates. Record hypothesis direction, confidence level, and any exclusions used. Pair uplift with context: traffic, seasonality, and launch risk. Use the calculator’s interval and incremental metrics to support a decision: ship, iterate, or stop. Keep a changelog of wins and losses to improve priors.
Uplift is the size of the change in conversion rate. Significance is evidence the change is unlikely under no effect. Use uplift to judge impact and confidence intervals to judge uncertainty.
Use two-sided when any change matters or when direction is not pre-registered. Use one-sided only when you committed in advance to shipping only if the variant improves conversions.
When baseline conversion is low, small absolute changes inflate relative lift. Low traffic also widens confidence intervals. Focus on absolute uplift and the interval width before celebrating a large percent lift.
It estimates extra conversions produced by the variant compared with applying the control conversion rate to the variant’s visitor count. It translates a rate change into an expected count difference.
No. It summarizes outcomes after a test. Power analysis helps plan sample size before running. Use both: plan with power, then evaluate results with uplift, p-values, and confidence intervals.
Avoid relying on it with very small samples or extreme rates near 0% or 100%. In those cases, consider exact tests or Bayesian models, and confirm tracking and randomization quality.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.