Storytelling with Data

Most data presentations fail before the first chart is drawn. The analyst knows the numbers well, but nobody has asked who will use them, what they need to decide or which single finding should change their mind. The result is a tour of everything that was measured, and an audience that leaves unsure what to do. This chapter gives you the habits that come before any chart. You will start by naming your audience and the decision in front of them. You will then compress your analysis into one main message that fits in a single sentence. Finally, you will learn to give every important number the context and comparison it needs, so that people can see at once whether it is good news, bad news or a reason to act.

Start with the decision, not the data

When you finish an analysis, the natural instinct is to show what you found, in the order you found it. That order makes sense to you because it follows your work. It rarely makes sense to the person who has to act on it. Before you open a slide or a spreadsheet tab, ask a different question: what decision will someone make, or make differently, because of this? Hannah, an operations analyst in London, was asked to look at delivery times across three warehouses. Her first draft had eleven charts covering every route and week. When she asked her director what the review was for, the answer was short. The board needed to decide whether to fund extra weekend shifts at one site. Nine of her charts had nothing to say about that choice.

Three questions about your audience

A clear picture of your audience saves hours of work and makes your message sharper. You do not need a formal profile, just honest answers to a few questions. Write them down before you build anything.

• Who exactly will use this, and what do they already know about the topic?
• What decision or action is in their hands, and by when?
• What would make them say yes, and what would make them hesitate or push back?

The third question matters more than it looks. A finance director may care most about cost and risk, while a regional manager may care about workload and customer complaints. The same analysis can need a different emphasis for each, even when the numbers do not change. If you have several audiences, design first for the decision maker, the person who owns the choice. Others can receive a version adapted to their concerns later.

Exploring is not explaining

Analysis has two stages. In exploratory analysis you look at the data from many angles to find what is interesting. In explanatory analysis you show other people the few things that matter. Problems start when the first stage is presented as if it were the second. Your audience does not need to retrace your path. They need the destination and enough of the route to trust it.

One analysis, one main message

Once you know the decision, you can decide what your analysis says about it. Aim for one main message, stated as a full sentence with a point of view. Not a topic such as "Q3 sales performance", but a claim such as "Q3 sales fell short because two large renewals slipped, and recovering them should be the sales team's first priority." The presentation designer Nancy Duarte calls this the big idea in her book Resonate. She says it should express your own perspective, convey what is at stake and be a complete sentence. Cole Nussbaumer Knaflic, author of Storytelling with Data, adopted the same test for data work. If you cannot write the sentence, you are not yet ready to build the charts.

Notice what each message on the right does. It narrows the topic to the part that matters, it names a cause or a pattern, and it points towards an action. Each one is longer than its topic, yet far quicker to understand, because the reader no longer has to work out the point for themselves. A good main message saves your audience effort.

How to find your sentence

Grace, a programme manager in Nairobi, had a long report on a mobile savings pilot. She found her message by talking it through rather than writing it. She explained the findings to a colleague in under a minute, then wrote down what she had just said and cut it to one sentence. Speaking forces you to choose because you cannot say everything at once. Try these prompts when your own sentence will not come:

• If your audience remembered only one thing tomorrow, what should it be?
• What is surprising or different from what they expect?
• What do you want them to do, approve or stop?

Your message may need to be tentative if the data is thin, for example "early signs suggest". That is still a message. What you cannot do is hide behind a neutral title and leave the audience to guess.

A number alone means little

Tell a manager that a team handled, say, 4,200 requests last month and they will nod politely. Is that good? Is it more than usual? Is it enough? A single number has no meaning until it sits next to something it can be compared with. Context turns a figure into information, and comparison turns information into a judgement. Kevin, a marketing analyst in Seattle, once reported a 2.1 per cent email click rate to his leadership team. Nobody reacted. The next month he wrote that the rate was 2.1 per cent against 3.4 per cent the previous quarter, and that the drop began when the newsletter moved to a later send time. The same number started a useful discussion.

Choose the comparison that answers the question

Different questions need different comparisons. Pick the one that speaks to the decision you identified earlier, and say clearly which one you are using.

• Against a target or plan, when the question is whether you are on track.
• Against a previous period, when the question is whether things are improving.
• Against a peer, region or competitor, when the question is how you stand relative to others.
• Against a total, when the question is how big a share something represents.

Watch the scale as well. A 50 per cent rise sounds dramatic, but if complaints went from four to six, the absolute change is small. Give both the relative and the absolute figure when either one alone could mislead. Make sure the denominator is clear too: per customer, per week or per thousand orders. Name your baseline in words, not only on an axis. "Up 12 per cent" invites the question "compared with what?", while "up 12 per cent on the same month last year" answers it. When different people in the room have different baselines in mind, such as last year's result, this year's budget or a rival's figure, they can read the same number as good and bad at once. Stating yours settles that before the discussion begins.

Finish with the so what

After the number and its comparison comes the most important part, which is what it means for your audience. This is the so what. It links the evidence back to the decision. "Weekend delivery times are 30 per cent slower than weekdays at the busiest site, which means about one in five weekend orders now misses its promised date and puts our service agreement at risk." A simple pattern helps: state the number, give the comparison, then say the implication. If you cannot write the implication, ask whether the number belongs in your story at all.

Checkpoint: Kevin's churn briefing

What good looks like

Good data stories start before any chart. Name the person who will use your analysis and the decision they face. Compress what you found into one main message, a complete sentence with a point of view and something at stake. Give each key number a fair comparison that matches the question, show absolute and relative change when either could mislead, and finish with the so what. Everything that does not serve the message can go.