Calculated Results
Results update after calculation.
Interactive Performance Chart
Targeting Interpretation
Selected Cutoff
Lift Table
Model Comparison
Formula Used
Cumulative Lift = Cumulative Positive Rate / Overall Positive Rate Bin Lift = Bin Positive Rate / Overall Positive Rate Cumulative Gain = Cumulative Positives / Total Positives Profit = Positive Value − Contact Cost − False-Positive Cost − Fixed Cost
How to Use
- Paste data or upload a delimited file.
- Map outcomes, scores, weights, and segments.
- Choose bins, targeting percentage, and campaign values.
- Calculate, interpret, compare, and export results.
Example Data
| ID | Actual | Model A | Model B | Weight | Segment |
|---|---|---|---|---|---|
| C001 | 1 | 0.92 | 0.85 | 1 | Retail |
| C002 | 0 | 0.84 | 0.76 | 1 | Retail |
| C003 | 1 | 0.79 | 0.81 | 1 | Business |
Frequently Asked Questions
What does lift measure?
Lift compares model targeting against random selection. Higher lift shows stronger concentration of positives. Early lift matters most for limited campaigns.
What is cumulative gain?
Cumulative gain shows captured positive cases. It grows as more records are targeted. The perfect curve captures positives earliest.
What is bin lift?
Bin lift measures one group only. It can fluctuate between adjacent bins. Cumulative lift is usually smoother.
Can scores exceed one?
Yes, unrestricted ranking scores are supported. Disable probability validation before calculation. Sorting still determines the ranking.
How are tied scores handled?
Ties can preserve input order. They can also favor either class. Seeded random ordering is available.
Should sample weights be used?
Use weights for sampled populations. Weights affect rates and cumulative totals. Validate weights before interpreting results.
Can several models be compared?
Select several score columns together. Curves and summary metrics are overlaid. The comparison table ranks each model.
Why can lift change between datasets?
Lift depends on positive prevalence. Sampling and segments also matter. Compare models on identical evaluation data.
How is the best cutoff selected?
The calculator checks cumulative profit. Capacity and budget limits are respected. The best feasible cutoff is highlighted.