Support metric-based (HAVING) filters as dashboard native filters

What is the problem or goal you're trying to solve or accomplish?

The customer wants to filter dashboard charts by an aggregate threshold — for example, "only show customers with total revenue above X" — and control that threshold from a dashboard-level filter so it applies across charts and can be changed by viewers without editing each chart.

Today, aggregate filtering is only available inside Explore, where selecting a saved metric or aggregate expression as a filter subject produces a HAVING clause. The native filter config modal offers only dataset columns as filter targets, and the values a native filter emits are applied as WHERE predicates. There is no metric option and no WHERE/HAVING equivalent to the toggle available in the Explore filter popover.

The practical consequence is that any threshold on an aggregate must be hardcoded per chart. A dashboard with several charts sharing the same measure requires the threshold to be edited in each one, and dashboard viewers cannot adjust it at all.

How are you solving it currently?

Not yet confirmed with the customer — the question was asked as a capability check, not as a report of an existing workaround. The available options are:

• Pre-aggregating in a virtual dataset so the measure becomes a physical column, then filtering it with a Numerical range native filter. This works, but the threshold is fixed to the grain of the pre-aggregation; if a chart re-aggregates to a coarser grain, the filter applies to the fine-grained rows before roll-up, which is not equivalent to a HAVING on the chart's own GROUP BY.

• A Jinja filter_values() block emitting a HAVING clause in virtual dataset SQL, driven by a native filter on a carrier column. This produces a genuine dynamic HAVING, but is limited to discrete threshold values and pushes maintenance burden into hand-written SQL.

• Leaving the threshold in the chart, which is correct per-chart but not dashboard-controllable.

What is your recommended solution?

The customer did not propose an implementation. Inferred from the request:

Allow a native filter to target a dataset metric in addition to a column. When the target is a metric, the emitted filter should be applied as a HAVING clause against the chart's own GROUP BY, matching Explore's existing behaviour. A Numerical range filter on a metric is the most direct fit for threshold use cases.