We developed heuristic plausibility checks for a forecasting system, which are regularly executed and inform the responsible employees about conspicuous forecasts via automatic e-mail generation.

The basis used to define heuristic plausibility checks was prior project experience on the part of VIA Software employees: which centered around technical support for forecasting system operations. Heuristics applied in this project were formalised and suitably parameterised, which—compared to manual checks—enabled an automated heuristic evaluation for a large quantity of forecasts. Anomalies were also automatically assigned a “severity” based on potential monetary impact estimates for possible forecast errors, which were in turn used to prioritise downstream manual analyses.

Furthermore, we extended automated analyses with additional heuristic checks and statistics defined by the customer, establishing regular anomaly reporting via e-mails to respective market experts.