How Do We Actually 'Pull Stories Out of Data'?
Source: https://counting.substack.com/p/how-do-we-actually-pull-stories-out ↗
Randy Au writes the "Counting Stuff" substack and has more practical wisdom about doing data work inside product organisations than most books on the subject.
This post is about what happens between a SQL query and an insight: the specific craft of turning numbers into stories that other people can act on.
Au is honest about the amount of judgement involved and the amount of self-deception it is possible to inject without noticing.
For product direction it is essential reading because most metrics conversations collapse under their own weight; Au teaches the discipline that keeps them useful.
His broader substack is worth subscribing to as a companion.
Central argument
Au argues that 'pulling stories out of data' is not a mechanical process of reading what the numbers say, but an active, judgement-laden act of construction — the analyst chooses which patterns to foreground, which comparisons to draw, and which causal framings to adopt. The danger he surfaces is not incompetence but well-intentioned self-deception: the story feels earned by the data when it was partly authored by the analyst's prior beliefs. His practical contribution is making that invisible craft legible, so that practitioners can apply it more deliberately and audit it more honestly.
Critique
Au's framing centres the individual analyst as the locus of craft and potential bias, but this understates how much narrative distortion is structural rather than personal — incentive systems, stakeholder pressure, and the way questions get commissioned in the first place shape which stories are even attempted before anyone runs a query. A thoughtful reader might argue that improving individual judgement, without addressing the organisational context that rewards confirming stories over challenging ones, leaves the harder problem untouched.
Why it matters for product
For a CPO, the most consequential moment in any metrics conversation is not the dashboard review but the sentence that starts 'what this tells us is' — and Au gives product leaders a precise vocabulary for interrogating that sentence rather than accepting it. This matters directly in OKR cycles, where teams routinely construct narratives that make flat metrics look like progress, and in discovery, where data is selectively storied to justify decisions already made on instinct.