Pitching data insights to decision-makers: recommendation first, three pieces of evidence, honest answers

Start from the decision, in one sentence

Before building any slide, write the single sentence you want the room to repeat after you leave: what to do, why, and what it should change.

An analytics pitch is not a tour of the work. The audience came to decide something, so the first thing it needs is the decision you propose. The course gives a template to fill in before touching the slides, and the recommendation it produces goes on the first content slide.

The test is simple and slightly brutal: after the pitch, can a listener say your recommendation in one sentence? In the session, every listener writes it down on a feedback form. If half the room writes something different from what you meant, the message did not land, however good the analysis behind it.

Example: the delivery brief

One brief of the course asks which delivery stage of Olist (approval, seller handling, carrier transit) should receive next quarter's improvement budget. An illustrative sentence, not a result: we recommend putting the budget into the stage where late orders lose most of their days, because the breakdown of delays by stage shows it, which should reduce the share of orders that arrive after the promised date. Replace each bracket with your own figures and the sentence is ready to open the pitch.

Back to contents.

Ten minutes in four blocks

Recommendation, evidence, impact, then limits and next steps: the audience knows where you are going before it sees a single chart.

The order is the one consultants and analysts use when time is short: conclusion first, support afterwards. The deck gives a suggested split of the ten minutes, explicitly not a rule; what is not negotiable is that the recommendation is heard in the opening minute.

Each slot in the session also contains four minutes of questions and one minute to change over, which is why the timing is held strictly: at zero the speaker finishes the sentence and the questions begin.

The impact block is where many pitches become vague. A useful estimate says what moves, for whom and in what order of magnitude, and states the assumptions behind it so that the audience can disagree with an assumption rather than with you. The limits block is not an apology: it shows that you know where the evidence stops, and the next checks you name are often what a decision-maker funds.

Back to contents.

What convinces a decision-maker

A short chain of evidence that leads straight to the recommendation persuades more than a long list of interesting charts.

The course treats storyline and evidence as one thing: every slide should support the recommendation, and the evidence should lead there without detours. In practice that means choosing three pieces of evidence at most and cutting the rest, even good analyses that answer another question. A slide that cannot be connected to the decision in one sentence belongs in the appendix or in your notes for the questions.

Habits that make evidence persuasive

1. A finding as the title. Each chart title says what to see, as taught in the visualisation session, so the slide can be read in five seconds.
2. Rates with their denominator. A comparison between groups of different sizes uses rates, and keeps the size of each group visible.
3. Numbers you have checked. Any figure that reaches the recommendation slide has been recomputed or reconciled with a known total, including figures drafted with an AI assistant.
4. A model read through its errors. When a prediction supports the case, say what each kind of error costs the business, not only how accurate the model is.
5. A dashboard that opens. The link to the group's dashboard is tested in a private window, so it works for someone outside the group.

Delivery matters, but it is not the heart of the matter. Clear, confident and on time is enough. Who presents which block is decided in advance, the first minute is rehearsed aloud, and the whole pitch is run once with a timer before the day.

Back to contents.

The questions are where you prove the analysis is yours

Answer first, then give the evidence, and when you do not know, say so and explain how you would find out.

In the course, the four minutes of questions are taken as seriously as the evidence itself, and without any AI help. That reflects real meetings: a strong deck followed by evasive answers loses the room. Every member of the group answers at least one question, so the team decides beforehand who covers the data and the cleaning, the analysis, the recommendation and the limits.

• How did you clean the data, and what did the cleaning remove?
• What result would make you change your recommendation?
• Would the conclusion still hold with today's data rather than the historical file?
• How did you check what the AI assistant produced for you?

Those four questions come straight from the course and deserve a short prepared answer each. The reply itself follows one pattern: the answer in the first sentence, then the evidence. Uncertainty is stated honestly; I do not know, followed by how you would find out, is a better answer than a confident guess, and decision-makers usually recognise the difference.

Back to contents.

Mistakes that sink a good analysis

Most failed pitches bury the recommendation, show too much, or run out of time before the questions.

Step by step

1. Opening with the method. Starting with how the data were cleaned is a reliable way to lose the room before the message arrives.
2. Saving the recommendation for the end. If time runs short, the one thing the audience needed is the thing it never hears.
3. Too many slides. Every slide that does not support the recommendation competes with it; cut it or move it to the appendix.
4. Running over. A pitch stopped at ten minutes loses its last slide, and the minutes of questions matter more than that slide.
5. One speaker for everything. When only one person can answer, the audience doubts that the group owns the work.
6. Rehearsing instead of listening. Groups waiting their turn learn more from hearing which messages land in other pitches than from a last run-through.

Listening is part of the exercise. For each pitch other than its own, every student fills a short form with three fields that mirror the structure of a pitch: the recommendation as understood, the evidence that convinced most, and the question they would have asked. The presenting group receives the fronts unsigned, so the feedback can be frank. The written brief is due a few days later, on purpose, so that the questions and these forms can still change it.

Back to contents.

After the course: six questions, resources and roles

Tools will change over a career; the questions you learned to ask of a number, a chart or a model will stay useful.

For those who want to keep practising for free, the deck points to three routes: Elements of AI (University of Helsinki), which explains what AI can and cannot do without code and makes a good base for governance work; Orange Data Mining, desktop software that builds the kind of models seen in Session 4 by drag and drop; and the short browser courses of Kaggle Learn (SQL, machine learning, data visualisation), which follow on from Sessions 3 and 4 with some Python. Two certificates are mentioned at two depths: the Google Data Analytics Professional Certificate on Coursera, covering the whole analyst workflow, and the Microsoft Power BI Data Analyst Associate (exam PL-300), narrower and centred on Power BI. Choose by the tools named in job offers where you want to work, and check names and prices before enrolling, as they change often.

None of the roles that use these skills requires being the best programmer in the room: business analyst, product manager, analytics translator, data product owner, AI governance and compliance, consultant. Since job titles differ from one employer to the next, judge a post by what its description asks for. In Europe the AI Act also creates work: training staff in AI literacy, and meeting the high-risk obligations, for instance when AI helps with recruitment, from 2 December 2027. The course ends on its first rule: start from the decision, not from the data.

Back to contents.

Key takeaways

1. Write the one-sentence recommendation before any slide, and say it in the first minute.
2. Use three pieces of evidence at most, each one tied directly to the decision.
3. State the expected impact with its assumptions, and the limits with the checks you would run next.
4. Answer questions with the answer first, then the evidence; admit what you do not know and say how you would find out.
5. Keep the method for the questions, and make sure every member of the team can take one.
6. Rehearse with a timer: a pitch that runs over sacrifices the minutes that test it most.
7. Start from the decision, not from the data.

Glossary

Pitch

A short, structured presentation that defends one recommendation before the people who will decide.

Recommendation

The action you propose, stated in one sentence with its main evidence and expected impact.

Storyline

The order in which evidence is shown so that it leads to the recommendation without detours.

Finding title

A slide or chart title that states what the data show rather than the topic.

Expected impact

What should change if the recommendation is followed, for whom, and roughly by how much.

Limits

What the data and the analysis cannot establish, stated openly.

Next steps

The checks or actions you would take after the decision, often what gets funded.

Peer feedback

Short written feedback from listeners: the message as understood, the best evidence, an open question.

Analytics translator

A role that links business teams and data scientists around a decision.

Data product owner

The person responsible for a dataset, dashboard or model on behalf of its users.

The other notes of the course

One public summary per deck, in the order of the course.

Previous: Generative AI, GDPR and the AI Act

1. Session 1From data to decisions
2. Session 2Data visualisation and storytelling
3. Session 3SQL essentials for managers
4. Session 4Predictive analytics for decision-makers
5. Session 5Generative AI, GDPR and the AI Act
6. Session 6Pitching data insights

All the notes, as a reading list

About these notes

Written by the lecturer from the slides of the deck, as a public summary: the key ideas, the formulas and one worked example. The exercises, their solutions, the tests and the grading stay in the deck, for the students of the course.

Revision: 16 October 2026.