Scorecard Gini, KS and PSI calculator
Drop in a CSV with a score and whether each loan went bad. The file is read in your browser and never uploaded.
Your files
These are generated demo loans. Drop in a file to see your own.
This generated recent sample has 3,000 loans, with scores that have drifted a little from the first.
Your files are read in this browser and nothing is uploaded.
File comparison
| Development sample | Recent sample | Change | |
|---|---|---|---|
| Gini | 0.710 | 0.634 | −0.076 |
| AUC | 0.855 | 0.817 | −0.038 |
| KS | 0.559 | 0.478 | −0.082 |
| Bad rate | 14.6% | 18.9% | +4.2 pp |
| Loans | 5,000 | 3,000 |
A common rule of thumb reads below 0.1 as stable, 0.1 to 0.25 as worth a look, and above 0.25 as shifted.
ROC curve
Score bands
| Score range | Development sample | Recent sample | Cut-off at band start | PSI by band | ||||
|---|---|---|---|---|---|---|---|---|
| Loans | Bad loans | Bad rate | Loans | Bad rate | Approved | Approved bad rate | ||
| 300 – 487 | 497 | 294 | 59.2% | 448 | 55.6% | 100.0% | 14.6% | 0.0203 |
| 488 – 526 | 490 | 160 | 32.7% | 373 | 31.6% | 90.1% | 9.7% | 0.0063 |
| 527 – 554 | 496 | 110 | 22.2% | 360 | 18.6% | 80.3% | 6.9% | 0.0040 |
| 555 – 578 | 500 | 60 | 12.0% | 346 | 16.5% | 70.3% | 4.8% | 0.0022 |
| 579 – 601 | 513 | 42 | 8.2% | 316 | 10.1% | 60.3% | 3.6% | 0.0001 |
| 602 – 621 | 490 | 23 | 4.7% | 256 | 7.0% | 50.1% | 2.6% | 0.0018 |
| 622 – 645 | 505 | 18 | 3.6% | 257 | 4.3% | 40.3% | 2.1% | 0.0025 |
| 646 – 671 | 502 | 21 | 4.2% | 236 | 3.4% | 30.2% | 1.7% | 0.0053 |
| 672 – 709 | 507 | 3 | 0.6% | 221 | 1.8% | 20.1% | 0.4% | 0.0089 |
| 710 – 850 | 500 | 1 | 0.2% | 187 | 1.1% | 10.0% | 0.2% | 0.0178 |
Get the result by email
After measuring a scorecard, test its cut-off on past applications that were already decided before running it as the live credit decision. Start free
How figures are calculated
Gini = 2 × AUC − 1 KS = max ( Share of bad loans declined − Share of good loans declined ) Bad rate = Bad loans ÷ Loans PSI = Σ ( Second file share − First file share ) × ln ( Second file share ÷ First file share )
Bands are the ten tenths of the first file by score, and the second file is counted in the same bands. Ties in the score are shared evenly in the AUC.
Calculate Gini, KS and PSI
Use a scored file of loans with their outcomes to calculate Gini, AUC and KS. The tool also shows the ROC curve, bad rate by score band and the approval result at each cut-off.
Compare a recent sample with the development sample to see whether the score still ranks loans well and whether the population has shifted. Use the results when you validate a scorecard and review its cut-off before running it as a live credit decision.
Scorecard questions
How do you calculate Gini from AUC?
Gini is calculated from AUC by doubling AUC and subtracting one. This tool shows both figures from the same scored file.
What is the KS statistic in credit scoring?
KS is the largest gap between the share of bad loans and the share of good loans at or below a score. The tool calculates KS from the score and outcome columns in your file.
How do you calculate and read PSI?
PSI compares the score distribution in a recent file with the development sample across the same score bands. The tool shows PSI overall and by band, with a common rule of thumb that labels the result as stable, worth a look or shifted.
How do you validate a credit scorecard on a new sample?
Compare the new sample with the development sample using Gini, AUC, KS and bad rate. Review the ROC curves, score bands and cut-off results, then check PSI for population movement.
Why does Gini drop on recent loans?
Gini drops when the score ranks recent loans less well than the development sample. Comparing the two files shows the change in Gini, AUC and KS alongside the score bands and bad rates.