Control
50%
5.00%
Projected evidence over time (1 − p)
Traffic sensitivity
Projected 1 − p at different traffic volumes, conditional on the assumed effect. Not a confidence level or a stopping rule.
Visitors needed by power
Based on your expected uplift assumption - larger real effect = fewer visitors needed
Power over time
Assuming your expected uplift is correct - how likely the test is to detect it after each time unit
Deep dive
Paste daily numbers to see how your test developed over time.
One row per group and day - e.g. an Adobe Workspace table with Experience → Day, or a sheet with Group · Date · Visitors · Conversions. Totals rows are fine, they're used for the cards.
Traffic per day
Visitors per day and group - spikes and gaps show up here first
Observed evidence over time (1 − p)
Your real data, added up day by day. The line will cross the threshold by chance now and then - don't stop the test at the first crossing.
Uplift over time
Line = uplift so far, shaded = likely range. Dots = uplift within each week - a downward trend means the effect is wearing off.
Breakdowns
Patterns, not winners: slice data often enough and some slices look like big wins or losses by pure chance. Use this to spot patterns - e.g. weekends behave differently - not to crown a winner per slice.