Turn a Power BI report full of numbers into three insights an executive will actually act on

You are a senior data analyst preparing the narrative that goes on top of a Power BI report before it reaches an executive audience. The report itself has the numbers; your job is to say what matters and why, in a way that survives someone skimming it in 90 seconds.

REPORT CONTENTS
Monthly revenue trend (up 4% MoM), churn rate by segment (SMB churn spiked to 9% this month vs 5% average), NPS trend (flat), regional sales map (unchanged distribution).

AUDIENCE
VP of Customer Success and the exec staff meeting, who care about retention and are not deep in the underlying data themselves.

BUSINESS CONTEXT
We raised SMB pricing 15% at the start of this month; support ticket volume has not notably changed.

WHAT DECISION THIS MIGHT INFORM
Whether to roll back or phase in the SMB pricing change before it hits the enterprise segment next quarter.

Do the following:
1. From everything in the report contents, select only the moves that are both statistically meaningful (not just noise in a volatile metric) and business-relevant to the decision context — most numbers in a report are neither, and shouldn't make the cut.
2. Rank your selected insights by how much they should change what the audience does next, not by how dramatic the percentage change looks — a 3% move in the metric that actually drives the decision outranks a 40% move in a metric nobody's about to act on.
3. For each insight, state it as: what happened, the most likely explanation given the business context (flagged clearly as a hypothesis if you don't have enough context to be sure), and the specific action or question it should prompt.
4. Explicitly call out anything in the report that looks alarming at a glance but is probably not worth executive attention, and say why — this protects the audience from chasing noise.

WHAT NOT TO DO
Do not summarize every visual in the report — an insight narrative that mentions everything is equivalent to mentioning nothing. Do not state a causal explanation as fact when the business context doesn't actually support it; flag it as your best hypothesis and say what would confirm it.

OUTPUT FORMAT
1. Top 3 insights, ranked, each as: what happened / likely explanation (hypothesis-flagged if uncertain) / recommended action or question
2. "Looks alarming but probably isn't" section, one to two items, with the reason
3. One sentence: the single thing this audience should remember if they read nothing else