Chapter 13. Turning numbers into meaning — figures, reports, presentations, practice

Key points

• An analysis becomes valuable only when it lands and moves someone
• Who you are speaking to — a researcher, an executive, the floor — changes the translation
• AI generates figures, reports, slides and drafts at speed; deciding the claim stays human

An analysis is finished only when it lands

However fine the analysis, asleep in a drawer it changes nothing. The journey of data completes only when the result lands, and moves someone's judgement or behaviour. Recall Chapter 1: what matters is not gathering numbers but turning numbers into meaning. Communicating is the work of translating that meaning into a form that reaches the other person.

The same result, translated differently

Here the role setting from the earlier parts pays off. The same result changes emphasis and vocabulary according to who is hearing it.

• To researchers and conferences — careful about the validity of methods, the limitations, and the relationship to prior work. Objectivity and reproducibility first.
• To executives and funders — the conclusion, and how it bears on the business (return on investment, implications for decisions), concisely.
• To floor staff — what changes from tomorrow, at the level of concrete action. No jargon.
• To users and the public — reassurance and understanding. Less number, more "what this means for you".

Producing several translated versions from one analysis is where the skill of communication shows.

AI accelerates all of it

And this translation is exactly what generative AI is best at. Everything from the earlier parts comes together here.

• Figures — "turn this result into a bar chart an executive can read at a glance"
• Reports — "write this up as a research report: a three-line summary, then detail, then limitations"
• Slides — "structure this as ten slides for a five-minute conference talk" (Gemini Notebook, Gamma and the like)
• Papers and manuscripts — "draft the results section in an objective tone"
• Business proposals — "turn this validation into a proposal for improving the service"
• Publishing — "draft this finding as an article" (the knowledge base you grew earlier is the source)

Having reached the point of building and publishing your own site in the previous part pays off here too. Publish your analysis to the world on your own site. From collection through to communication, the whole line is now in your hands.

But never yield the claim

AI produces figures, prose and slides astonishingly fast. But deciding what you are actually claiming, and what you most want to convey, stays a human's work to the end.

Does the beautiful chart the AI made really support the claim? Does the report's conclusion overreach the data's limits? Has correlation been stated as causation, against everything Chapter 12 said? Watching for that is your job. Speed of communication to AI, responsibility for communication to the human. That last line separates researchers and practitioners who are trusted from those who are not.

A prompt to try (multiple translations)

You are a communications specialist who understands both research and business. Below are the key points of my analysis.

[Key points] (e.g. participants in the exercise class had higher satisfaction, but selection bias is possible and causation cannot be asserted)

Translate this same result for three audiences.

1. A conference presentation (objective, limitations stated)
2. A proposal to executives (concise, with implications for decisions)
3. Floor staff (plain, with what to do from tomorrow)

In all three, take care not to overreach the limits of the data.

Chapter 13 summary

• Analysis becomes valuable only when it lands and moves someone
• The same result needs different translations for different audiences; produce several
• AI generates figures, reports, slides and drafts at speed; the claim and the responsibility stay human