Performance Metrics

The performance metrics page lets you analyse your model’s forecast accuracy over time. It displays calculated metrics for each forecast category, filtered by a date range and model-specific options.

Accessing performance metrics

From the forecast dashboard, click Performance Metrics in the top navigation bar. This link is only enabled for models that have performance metrics configured. If the link is greyed out, the current model doesn’t have metrics set up.

Page layout

The performance metrics page is organised into three areas:

Sub-navigation bar — the same category selector as the forecast dashboard, letting you filter which categories are displayed.
Main panel — metrics tables showing accuracy data for each selected category.
Options sidebar — date range pickers and model-specific filter options.

Metrics tables

Each selected category gets its own table. Tables have a green header bar showing the category name and sample size (the number of data points used in the calculation).

The table body displays your configured metric values. Numeric values are formatted to two decimal places using Australian locale formatting. The specific metrics shown depend on how your model’s performance queries are configured.

Note: Metrics data is calculated from a nightly extraction process that saves model run output into Parquet files. There may be a short delay before the most recent runs appear in the metrics.

The sidebar contains controls that filter and configure the metrics display.

Run date range

The run date range filter is always present. It provides two date pickers:

Min Date — the start of the date range. Defaults to 3 months ago.
Max Date — the end of the date range. Defaults to today.

Adjust these dates to narrow or widen the window of historical runs included in the metrics calculation. The tables update automatically as you change the dates.

Model-specific options

Below the date range, you may see additional options specific to the current model. These are defined by the model author and can include:

Select dropdowns — choose from a list of predefined values.
Boolean toggles — enable or disable a filter.
Number inputs — enter a numeric value.
Date pickers — select a specific date.

These options are injected as variables into the underlying metrics query, giving you fine-grained control over the calculation.

Exploring metrics data

To dig deeper into the raw data behind your metrics, click the Explore link. This opens the SQL Viewer with the metrics DuckDB data pre-loaded, so you can write custom queries against the historical forecast output stored in Parquet files.

Next steps

To learn more about exploring raw data with SQL, see Exploring Data with SQL. For details on how metrics queries are defined, see the Configuration File documentation.