RecallWeave · A small data lab
When groups and totals disagree
An option can have the higher success rate in each group and the lower rate overall. Follow the denominators, then give both options the same group mix.
Fictional examplesWorks offlineNo account or upload01Inspect the recorded counts
Enter whole numbers from 0 to 1,000,000. Successes cannot exceed total attempts. A total of zero means that group's rate is unavailable.
Compare like groups
| Group | Option A | Option B | Comparison |
|---|---|---|---|
| Newcomers | |||
| Experienced |
02Combine each recorded sample
Add successes, then add attempts. Each option keeps its own group proportions.
Option B · pooled
03Ask a same-mix question
Keep the observed group rates. Apply one chosen group mix to both options. These reference rates describe a different comparison; they do not change the recorded counts.
Option A · reference
Option B · reference
Inspect exact fractions
Fractions are proportions, before multiplying by 100 for percentages. Negative differences favor B. Comparison labels use these exact values.
Which question are you answering?
These recorded attempts? The pooled fraction describes each option's actual sample. A chosen common group mix? Use the stated reference weights. Neither question makes the other arithmetic wrong.
What would changing the option cause? The table alone does not answer that. Assignment, common causes, and when groups were measured matter. Grouping by a consequence of the option can change the effect being studied.
No confidence intervals, significance tests or causal effects are estimated. Rates are displayed to two decimal places; exact comparisons can remain different even when the displayed percentages match.
The comparison file keeps your current counts, reference mix and exact fractions. The separate course contains original questions: . It uses the RecallWeave deck format; open it through a compatible lesson importer. This explorer does not load or replace a learner session.