---
title: "Comment by Azhar Abbas on \"Natural-Language Calculations in APEX 26.2\" | Hashnode"
description: "The 30% versus 38.18% example is a useful reminder that a correct row formula can still produce a misleading report total. I’d include unequal-sized sales, retu"
url: "https://hashnode.com/posts/natural-language-calculations-in-apex-26-2/6af5ebb1987f32cbcb5a52b6/comment/6af6af2a5bc1768a1dba7b60"
---

[Discussion](https://hashnode.com/posts/natural-language-calculations-in-apex-26-2/6af5ebb1987f32cbcb5a52b6)

Comment [Natural-Language Calculations in APEX 26.2](https://hashnode.com/posts/natural-language-calculations-in-apex-26-2/6af5ebb1987f32cbcb5a52b6)

[Azhar Abbas](https://hashnode.com/@azharabbas)

[15h ago](https://hashnode.com/posts/natural-language-calculations-in-apex-26-2/6af5ebb1987f32cbcb5a52b6/comment/6af6af2a5bc1768a1dba7b60)

The 30% versus 38.18% example is a useful reminder that a correct row formula can still produce a misleading report total. I’d include unequal-sized sales, returns and zero-revenue rows in the test fixtures before treating the generated margin as a shared business metric.

[Jon Dixon](https://hashnode.com/@orclapex)

[8h ago](https://hashnode.com/posts/natural-language-calculations-in-apex-26-2/6af5ebb1987f32cbcb5a52b6/comment/6af6af2a5bc1768a1dba7b60#reply-6af70dc1cc91786c56692c87)

Thanks Azhar, those are useful test cases. The same applies to manually created Interactive Report computations too. AI makes it easier to add the calculation, but either way we need to be careful about an exploratory calculation becoming a business metric without an agreed definition and testing.