Section 1
Numbers do not carry their own meaning
A number on its own is inert. Churn is 4 percent tells a room nothing until they know the period, the denominator, and what it was last quarter. Add those and you have a fact. Add what it costs at current growth and you have an argument. Add what you would do differently if it stays at 4 percent for another two quarters and you have a decision, which is the only thing a meeting can actually act on. Most reporting stops at the fact stage and then wonders why it produces discussion instead of action. The fix is to write the claim first, in one sentence, before touching a chart. If the claim cannot be stated in a sentence, the analysis is not finished. Every chart then exists to support or falsify that one claim, and anything that does neither comes out. [How to Structure a Business Story for Maximum Impact](/blog/how-to-structure-a-business-story-for-maximum-impact) sets out the sequencing that keeps a data argument in order.
Section 2
Why the story outlives the statistic
People retain structure, not values. A month after a review, nobody recalls that support tickets rose 14 percent. They recall that the new onboarding flow generated a wave of confused customers and that the team turned it off. The second version encodes cause, actor and outcome, which is why it survives being retold by someone who was not in the room. Memory research supports the same point. In a classroom exercise Chip Heath ran at Stanford, after a round of short presentations 63% of listeners remembered the stories while only 5% remembered any individual statistic (Heath and Heath, Made to Stick). The lesson is not to abandon the number. It is to attach the number to the mechanism it came from, so the two travel together.
Section 3
A structure for a data narrative
A repeatable shape stops each analyst inventing their own. The model below sets the claim, the comparison, the mechanism and the decision. On making this stick across a team rather than a slide, see [Using Story Circles to Improve Team Communication](/blog/using-story-circles-to-improve-team-communication).
Section 4
Rebuilding one recurring report
Take the report your team produces most often and that changes the least behaviour. Strip it to the single claim it is really making this month. Everything else becomes an appendix, available if asked, not presented. Then add the two things that are almost always missing: the comparison that makes the headline number meaningful, and the decision that the number should trigger. Run it once in that form and watch what happens in the room. If people argue about the claim, the format is working, because arguing about a claim is how a decision gets made. If they ask for the other charts back, ask which decision those charts would change. Sometimes the honest answer is none, and then the report gets shorter permanently.
Section 5
Ways a data story misleads
Simplification is the risk that comes with the method. Four patterns cause most of the damage. A percentage without a base, where a 200 percent rise turns out to be two customers becoming six. A comparison chosen because it flatters, usually the weakest month in the previous year. A cause asserted from a correlation because the story needed one. And survivorship, where the analysis only includes accounts that stayed. State the denominator, name the comparison you chose and why, and flag where you are inferring cause rather than measuring it. A story that admits its own uncertainty is more useful than a clean one that quietly does not hold. On the data groundwork underneath any of this, see [The Role of Data in Effective AI Automation](/blog/the-role-of-data-in-effective-ai-automation).