Reproducible scores grounded in your actuals
Climingo scores providers against your historical observations with established meteorological verification metrics. There is no proprietary black box or vendor-supplied reference dataset.
Data capture and storage
Forecasts archived at issuance
Climingo records provider forecasts when issued, before outcomes are observed. This avoids post-hoc revisions and preserves what the provider predicted at the decision-relevant time.
Collection matches your use case: hourly values, site coordinates, or grid-cell averages. Coverage includes major providers serving European energy markets.
Your actuals define the truth
We verify against your observed values, including readings from your assets, purchased station data, or SCADA output from your generation fleet. Climingo standardises units and manages gaps. Your data stays private and is benchmarked only for you.
Upload CSV or JSON files, or connect directly by API to most major SCADA and market data platforms.
Score model
Deterministic scores: RMSE and MAE
Root Mean Square Error (RMSE) gives greater weight to large errors than Mean Absolute Error (MAE). In many energy trading use cases, RMSE for temperature and wind speed during peak-demand hours ranks providers first. MAE adds a linear-scale error measure for planning comparisons.
We calculate each metric at your chosen horizon, from 6h to 7-day. This shows which provider loses skill fastest with longer lead times and helps calibrate your hedge.
Probabilistic score: CRPS
Continuous Ranked Probability Score (CRPS) assesses the entire predictive distribution rather than its mean alone. It suits ensemble or quantile forecasts, rewarding accuracy and calibration while penalising excessive confidence in uncertainty estimates.
For deterministic-only output, CRPS becomes MAE, allowing direct comparison with probabilistic providers.
Validation and repeatability
Audit log
Each benchmark run creates a downloadable audit package containing matched forecast-observation pairs, errors at every time step, and summary statistics. You can independently reproduce the calculation with any standard verification library. The results are not confined to Climingo's platform.
We retain audit packages for 12 months after each run. They are also available through API for your reporting infrastructure.
Rolling window, point-in-time
Dashboard scores use a 90-day rolling window to reduce provider-level noise and show sustained performance. Point-in-time exports support event analysis, such as a wind ramp week, cold snap, or peak demand period. Both views rely on the same verification data.
Seasonal breakdowns support customers with materially different exposure in summer and winter, or across shoulder and peak periods.
First benchmark in 15 minutes
Upload actuals, choose providers, and receive a reproducible skill score table.