ChatGPT for Qlik: explain why a KPI ignores your selection

Give ChatGPT the Qlik expression and the current selections when a KPI shows an unexpected total. In set analysis, $ starts from the current selection in the default state; 1 starts from the full app record set regardless of those selections. A Year modifier does not automatically restore a Region filter discarded by the leading identifier.

For a KPI that must respect Region while overriding Year, first compare the record sets selected by $ and 1. Changing the aggregation cannot fix a selection policy that asks the wrong question.

Name the selected records

This original fictional app has one flat table and a scalar KPI with no chart dimension or alternate state:

Record Year Region Sales
R1 2025 East 10
R2 2025 West 20
R3 2026 East 30
R4 2026 West 40

The current selections are Region=East and Year=2026. The requested KPI is 2025 sales for the selected region, so it should use R1 and return 10.

These two candidate expressions request different sets:

Sum({$<Year={2025}>} Sales)
Sum({1<Year={2025}>} Sales)

For this fixture, the first starts with the selected state and replaces the Year selection with 2025, retaining East. The second starts without the current selections, then restricts Year to 2025. It includes R1 and R2 and returns 30. Those are authored record traces, not captured Qlik output.

Qlik's set-identifier reference, checked October 24, 2026, defines the starting sets. The comparison concerns selections within the app's available records.

Test the intended selection policy

Clear the Region selection in a working copy of the app. Both expressions should now return 30 for the stated fixture. Select West and retain Year=2026: the current-selection expression with its Year override should return 20, while the 1-based version remains 30.

That filter check tests whether the KPI follows the requested region. It does not establish how a real app behaves with additional associations, alternate states, dimensions or conflicting selections. Include those details before transferring the simple expression into a more complex model.

Copy the expression and the selection state into the request

Review this Qlik scalar KPI without editing the app.
Qlik product/version and relevant data model associations: [ENTER]
Exact expression and current selections: [PASTE]
State/alternate state and chart dimensions: [ENTER]
Required filters to preserve, replace or ignore: [ENTER]
Permitted record fixture and observed result: [PASTE]

Explain the starting record set for $ versus 1 before applying any
modifier. Trace record IDs contributing to the total. In the simple
fixture, Region=East and Year=2026 are selected; the KPI must override
Year to 2025 while preserving Region. Show expected results with East,
West and Region cleared. Propose the smallest expression correction.
Do not alter source amounts, introduce TOTAL or alternate states to
hide the discrepancy, infer permissions, or claim an actual app test.

The DAX guide handles a different expression language and filter model; do not paste DAX into a Qlik measure. Check the three selection cases in Qlik itself and keep the accepted filter policy with the KPI definition, so the next editor knows which selections it intentionally preserves.

Keep reading