Skip to content
PerslacePerslace Help Centre

Running an AI Impact Simulator backtest

The AI Impact Simulator (AIS) checks a commission scheme against selected historical months to show how that scheme would have affected results. Use it to understand the potential impact of a scheme before making a change. A simulation is not an approval, calculation record, or instruction to change a scheme.

Who can use this guide

Section titled “Who can use this guide”

This guide is for administrators and authorised Finance users with verified simulator access.

Prerequisites

Section titled “Prerequisites”

Before you begin, confirm that you have:

  • One or more approved historical months
  • Current scheme and assignment context
  • A valid baseline visible in the workspace
  • The organisation’s default currency
  • Authority to review the selected population
  • An agreed question and a human decision owner

Verification required: The reviewed workspace did not provide a closed baseline for a complete test. Model assumptions and baseline requirements remain unconfirmed.

Set up and review a backtest

Section titled “Set up and review a backtest”
  1. Open AI Impact Simulator (AIS).
  2. Select the scheme you want to review.
  3. Choose the objective shown in the interface: Cost Reduction, Improve Fairness, or Boost Motivation.
  4. Choose an available scope: Organisation, Branch, Group, or Individual.
  5. Choose the historical months to test using 1M, 3M, 6M, 1Y, or Custom.
  6. Review the selected population, dates, currency, and any overlap warning.
  7. Confirm that the baseline is present, understandable, and appropriate for the historical period. If it is missing or unclear, stop.
  8. Run the simulation only with approved test or historical data.
  9. Review the available comparison, risk, fairness, cost, and optimisation information without assuming how the values were produced.
  10. Compare material outputs with the source scheme, assignments, transaction data, and standard reports.
  11. Record the objective, scope, period, baseline, currency, warnings, output, limitations, and human reviewer.
  12. Test proposed scheme changes separately in a controlled draft and obtain normal approval before publication.

Expected result

Section titled “Expected result”

The simulation is recorded as exploratory evidence with its baseline, scope, period, currency, warnings, source comparison, and human conclusion.

Important notes

Section titled “Important notes”
  • Never publish or alter a commission scheme from simulator output alone.
  • Do not assume the simulator uses the same method as Dashboard forecasts or scheme Historical Replay.
  • The underlying algorithm, confidence, fairness definition, and optimisation method were not verified.
  • Use synthetic or approved historical data and protect all resulting financial and personnel information.

Troubleshooting

Section titled “Troubleshooting”

No baseline is available

Section titled “No baseline is available”

Stop. Confirm that an eligible historical period exists. Baseline eligibility and closed-cycle requirements need product confirmation.

An overlap warning appears

Section titled “An overlap warning appears”

Review the selected scheme, assignment dates, population, and period before continuing. Do not dismiss a warning you cannot explain.

The result is surprising

Section titled “The result is surprising”

Recheck the objective, scope, period, baseline, currency, and source data. Compare with authoritative reports and do not act on the output.

Section titled “Related articles”

Last updated: August 2026