iziRisk Ops · loss distribution approach
Load a loss-event register — or one of the three examples: a Basel II/III bank, a manufacturer or a cyber and IT incident log — and every event type gets a frequency and a severity fitted to its own events. The simulation convolves them year by year, correlates the event types, and returns expected loss, OpVaR, expected shortfall, the capital each event type consumes and a multi-year loss horizon. It runs in your browser; the register never leaves it.
1 · Loss-event register
CSV, TSV, Excel, XML or JSON, or pasted from Excel, in any of the five languages. One row per event.
Any layout: CSV, Excel, XML or JSON. AI works out which columns are the date, the losses and the recoveries, maps your categories to Basel event types, drops rows that are not losses, and runs the model.
Sent to the AI (Anthropic): the column headers, a summary of each column with sample values, the category values and the first 8 rows. The rest of the register stays in your browser. Remove names or ID numbers first if you do not want to share them. What is sent and how it is handled →
Sample bank (Basel II/III) · 583 events · 9 columns · Basel event-type codes detected
- This is the sample bank of the Event register tab: the same invented events, read exactly as “Send to the model” reads a register.
- Only accepted incidents with a loss are here. Reported and rejected events, incidents with no loss and potential events stay in the register.
- Amounts are in colones; the few events recorded in dollars are converted at the rate stored with each. The cells are Basel level-2 event types, with SUGEF business line, area, process and product kept as breakdowns.
- With the Professional plan you can load these events into your own register from that tab, change them and build the supervisors’ files.
- Each event type starts on its best fit: the severity family with the smallest KS distance, and a negative binomial where the monthly counts are overdispersed. “Workbook defaults” in step 4 switches every event type back to Poisson and lognormal by moments.
Column mapping
Guessed from the headers. The date and one loss column are required; net loss is gross − recovery when there is no net column.
Categories
Columns the cells, the scope and the heat map can use: business unit, cause, product, line of business…
Run the simulation to see the results.
Mitigation
Treat the risk the results just measured: give the event types that matter prevention, containment or insurance, price each strategy against the inherent register, and see where the plan leaves the risk matrix.
Strategies by event type
Event types with the largest expected loss first. Open one to add strategies, or let AI propose them; mark one strategy per event type as the plan.
| Cell | E[loss] | Capital allocated | Mitigation |
|---|---|---|---|
| 2.2 - Systems security | 16.5M | — | |
| 7.1 - Transaction capture, execution and maintenance | 15.8M | — | |
| 7.3 - Customer intake and documentation | 12.5M | — | |
| 1.1 - Unauthorised activity | 12.5M | — | |
| 2.1 - Theft and fraud | 11.2M | — | |
| 6.1 - Systems | 10.8M | — | |
| 4.3 - Product flaws | 9.9M | — | |
| 4.4 - Selection, sponsorship and exposure | 8.9M | — | |
| 4.2 - Improper business or market practices | 7.1M | — | |
| 4.5 - Advisory activities | 6.9M | — | |
| 4.1 - Suitability, disclosure and fiduciary | 5.7M | — | |
| 7.4 - Customer and client account management | 5.5M | — | |
| 1.2 - Theft and fraud | 5.4M | — | |
| 3.3 - Diversity and discrimination | 5.2M | — | |
| 5.1 - Disasters and other events | 5.2M | — | |
| 7.5 - Trade counterparties | 4.6M | — | |
| 7.6 - Vendors and suppliers | 4.2M | — | |
| 3.1 - Employee relations | 3.2M | — | |
| 7.2 - Monitoring and reporting | 2.8M | — | |
| 3.2 - Safe environment | 552.9K | — |
Mitigation — inherent vs residual
Every strategy is simulated on the same random numbers as the inherent register, so each difference is the mitigation and not sampling noise. Value per year = expected loss avoided + cost of capital × capital freed − cost per year. Capital is the unexpected loss (OpVaR − EL), as in the rest of the model.
Add strategies to the event types above to price them here.