An analysis only has value once someone understands it well enough to act on it.
Communication is therefore part of the analysis, not a final decoration.
Most decision makers have little time, so the message must come first and be short.
A useful structure is answer, reason, evidence: what we found, why it matters and how we know.
The recommendation should be specific, such as offering existing customers a 10% discount, not looking into pricing.
Numbers need context, such as a comparison with last year or with a target.
Precise-looking figures, such as 23.47%, often suggest more certainty than the data supports, so rounding helps.
Uncertainty can be shown plainly, such as somewhere between 2% and 5%, instead of a lone number.
Charts should be simple, labelled and titled with their message.
Technical details, such as the tests used, belong in an appendix that anyone can check.
Knowing the audience changes the version: engineers want the method, while executives want the decision.
Saying what the data cannot answer builds trust and stops results being stretched too far.
Stating the main message in 1 sentence is a good test of whether it is clear.
In practice, a short written summary often travels further than a slide deck, since it gets forwarded.
The usual mistake is presenting everything that was done, in the order it was done, so the finding gets lost.