Troubleshooting guide

Outcome

By the end of this guide, you can identify why a node formula returns unexpected results (empty values, wrong granularity, or wrong numbers) and fix the issue with a structured checklist.

Prerequisites

• You have access to the Model Editor

• You can edit the node’s formula in the Function Editor

Steps

1. Fix syntax and validation issues first

• If the editor shows a validation or syntax error, resolve that before troubleshooting logic.

• Re-check the Data preview after the error is resolved.

2. Define what “correct” looks like (before editing)

• Have two things ready:

  • One spot-check value: a specific intersection you expect to be correct.

  • Expected output shape: which dimensions/levels should be present.

3. Confirm the symptom in Data preview

Classify the issue as one of the following:

• Symptom: Empty output (values missing where you expect data) → Go to: Step 8, then Step 9

• Symptom: Wrong shape (too aggregated or too granular) → Go to: Step 5, then Step 6

• Symptom: Wrong numbers (shape looks right, values don’t) → Go to: Step 4, then Step 7

4. Isolate where the issue starts

• Temporarily simplify the formula:

  • start with a single node reference (e.g., 'Sales')

  • then add parts back one by one

• If the formula is large, move parts into helper nodes to inspect intermediate results or wrap and comment them out via #

5. Fix “wrong shape” first (dimensions and levels)

• First check whether the connected data sources have the dimensionality you expect.

• If the result is too aggregated:

  • make aggregation explicit (for example with ROLLUP) instead of relying on implicit behavior.

• If the result is too granular:

  • check for reshaping functions (EXPAND, EXPANDSINGLE)

  • check whether * or / introduced expansion when inputs have different dimensionality.

6. Check operator alignment when combining nodes

• + and - typically align to shared dimensions (rolling up).

• * and / typically apply across the combined dimension set (expanding).

• If the output shape changes right after adding an operator, the operator is the likely cause.

7. Check constants (0-dimensional numbers)

• If you add/subtract constants like + 1 or - 100000, verify whether the result becomes unexpectedly aggregated.

• If you intended an element-wise adjustment, use ADDEACH(...) instead of relying on raw constants.

8. If the output is empty, check filters and level values

• If you used FILTER(...) with explicit level values like ["EMEA", "APAC"], confirm:

  • labels match exactly (spelling/case)

  • the referenced dimension/level is correct

• Test with a single level value first, then expand the list.

9. Check missing vs zero

• If values disappear after comparisons, filtering, or join-like logic:

  • check whether one input is missing where the other has values at the same intersections.

• If results differ from what you expect due to assumptions/scenarios:

  • verify whether assumptions are applied upstream and whether you need BASELINE/NONSIM behavior for validation.

10. Re-validate with two spot checks

• Pick two representative intersections (e.g., one Product and one Month).

• Confirm:

  • the shape is correct

  • the numbers match expectation

11. Harden the final formula

• Split complex logic into helper nodes if it improves readability.

• Add short # comments that capture intent (why this is done, not what the syntax is).

Still in trouble?

Contact Valsight Customer Support
support@valsight.com

Related sections

• Formula basics syntax rules, constants, and how formulas evaluate

• Operators arithmetic, comparisons, and alignment behavior

• Function catalog A–Z index and categories for detailed behavior

• Browse by category choose the right function family by intent