Forecast fraud losses in a minute
Drop in confirmed fraud cases with the date the fraud happened, the date it was confirmed and the amount. The tool completes recent months for fraud still to be confirmed, then forecasts the next months.
Fraud cases file
1,635 fraud cases from 2024-04 were read, confirmed up to 2026-09, and 0 rows without readable dates or with a confirmation before the fraud were left out.
The file is read in this browser and never uploaded. Only monthly totals are sent by email if you request the result.
Later confirmations count at this lag, so set it to the longest delay you see.
These are generated fraud cases over 30 months. Fraud grows about 2% a month, and about a third is confirmed in the month it happens.
Fraud loss forecast for {h} months
The next 6 months are forecast at 573,348, and 162,319 of fraud has already happened and is still to be confirmed. Only 33.4% of 2026-09 is confirmed so far, which is why recent months look low.
Fraud and forecast
Confirmed Still unconfirmed Forecast
Fraud by month
| Month | Confirmed | Share | Estimate | Pending |
|---|---|---|---|---|
| 2026-09 | 25,670 | 33.4% | 76,776 | 51,106 |
| 2026-08 | 52,550 | 58.1% | 90,511 | 37,961 |
| 2026-07 | 76,230 | 71.7% | 106,282 | 30,052 |
| 2026-06 | 64,130 | 81.2% | 78,977 | 14,847 |
| 2026-05 | 71,020 | 87.8% | 80,864 | 9,844 |
| 2026-04 | 69,060 | 90.6% | 76,187 | 7,127 |
| 2026-03 | 70,780 | 93.4% | 75,797 | 5,017 |
| 2026-02 | 69,650 | 95.4% | 72,987 | 3,337 |
| 2026-01 | 73,320 | 97.5% | 75,208 | 1,888 |
| 2025-12 | 77,580 | 98.7% | 78,625 | 1,045 |
| 2025-11 | 66,480 | 99.9% | 66,575 | 95 |
| 2025-10 | 70,280 | 100.0% | 70,280 | 0 |
| 2025-09 | 68,900 | 100.0% | 68,900 | 0 |
| 2025-08 | 58,670 | 100.0% | 58,670 | 0 |
| 2025-07 | 56,350 | 100.0% | 56,350 | 0 |
| 2025-06 | 59,270 | 100.0% | 59,270 | 0 |
| 2025-05 | 47,850 | 100.0% | 47,850 | 0 |
| 2025-04 | 52,240 | 100.0% | 52,240 | 0 |
| 2025-03 | 49,500 | 100.0% | 49,500 | 0 |
| 2025-02 | 41,510 | 100.0% | 41,510 | 0 |
| 2025-01 | 58,930 | 100.0% | 58,930 | 0 |
| 2024-12 | 49,490 | 100.0% | 49,490 | 0 |
| 2024-11 | 55,390 | 100.0% | 55,390 | 0 |
| 2024-10 | 55,390 | 100.0% | 55,390 | 0 |
Factors by lag
| Lag | Factor | Share confirmed |
|---|---|---|
| 0 | 1.736 | 33.4% |
| 1 | 1.235 | 58.1% |
| 2 | 1.132 | 71.7% |
| 3 | 1.082 | 81.2% |
| 4 | 1.032 | 87.8% |
| 5 | 1.030 | 90.6% |
| 6 | 1.022 | 93.4% |
| 7 | 1.022 | 95.4% |
| 8 | 1.012 | 97.5% |
| 9 | 1.012 | 98.7% |
| 10 | 1.001 | 99.9% |
| 11 | 1.000 | 100.0% |
Email this result
The same fraud counts can feed fraud rules and limits in each decision, with losses tracked as they mature. Start free
Calculation method
Factor (l) = Σ Reported by lag l+1 ÷ Σ Reported by lag l Estimate = Confirmed × Factor (l) × … × Factor (Longest lag in months − 1) Pending = Estimate − Confirmed Forecast = a + b × Month, ± 1.96 × σ
Each month is completed from how older months kept growing after the same lag, using the chain ladder insurers use for claims reported late. The trend is a straight line through the completed months, so a change in the pattern or a season is not included.
How to forecast fraud losses
A fraud loss forecast estimates the total loss for each month after later confirmations are included. It uses confirmed fraud cases, the month the fraud happened, the confirmation month and the amount lost. If there is no amount column, it counts cases.
Recent months can look low because some fraud has happened but is still to be confirmed. Completing those months from older months gives a basis for planning, and the straight-line trend forecasts the coming months.
Fraud forecast questions
How do I forecast fraud losses?
Complete recent months for fraud still to be confirmed using how earlier months filled up. Then fit a straight-line trend to completed months to forecast the next months.
Why do recent fraud months look low?
Fraud is confirmed after it happens, so recent months have had less time for confirmations to arrive. The tool estimates the part still to be confirmed from older months.
What does incurred but not reported fraud mean?
It is fraud that has already happened but has not yet been confirmed. The tool estimates this amount by completing each month's reported total using development factors from older months.
How does chain ladder estimate late fraud?
The chain ladder compares amounts reported at one lag with amounts reported at the next lag across months old enough to show both. Those factors estimate how much of each recent month's fraud remains to be confirmed.
How can I plan next quarter's fraud losses?
Use completed monthly estimates as the basis for a straight-line trend over the selected months. The forecast includes a band around the line, but it does not account for seasonality or a change in the pattern.