Storytelling with Data

The answer is not “So we pulled six months of data and ran a cohort analysis…”

The answer is the recommendation. Every time.

The analyst’s last mile

You can build a perfect model, find a real pattern, and produce a chart that is technically accurate — and still change nothing. If the person who needs to act on your insight does not understand it, or runs out of time before you get to the point, the work was wasted.

The last mile of analysis is communication. The goal of a data story — any presentation, slide deck, or email that delivers findings — is not admiration of your charts. It is a decision: a concrete action someone takes because of what you showed them.

That reframe changes everything about how you structure the work.

The pyramid principle

In the 1970s, McKinsey consultant Barbara Minto observed that most analysts present like detectives: they lay out all the clues, walk through every step of the reasoning, and save the conclusion for the end. Busy executives — who may have dozens of such presentations in a week — lose the thread, interrupt, or simply check out before the punchline.

Minto’s fix is called the pyramid principle: flip the order entirely.

1. The recommendation (the single answer to the question) — one sentence, right at the top.
2. Supporting arguments — the two or three reasons the recommendation is correct.
3. The data — the charts, numbers, and analysis that prove each argument.

You present top-down: answer first, evidence second. The audience hears the conclusion, decides whether they agree or want more, and then receives the supporting detail. They are never guessing where you are going.

Two terms worth knowing:

• The lede (from journalism) is the main point — the most important information in a story.
• Burying the lede means hiding the main point at the end, after all the setup, so a skimming reader never reaches it. It is the single most common mistake in business presentations.

Before vs after: one concrete example

Here is the same finding delivered two ways.

Detective order (buried lede): “We pulled six months of funnel data and ran a cohort analysis across the three signup variants. Completion rates in the control group were 34%, in variant A they were 31%, and in variant B — the new screen we launched last quarter — they were 28%. Projecting that delta onto monthly traffic, we estimate…”

By the time you finish that sentence, the CFO is checking her phone.

Pyramid order (answer first): “We should revert the new signup screen — it is costing us about 500 paying customers a month. Here is why…”

The CFO puts her phone down. Now she wants the supporting arguments. You have the room.

Choosing the right chart

A chart is a sentence. Before you build one, finish this sentence: “This chart shows ___.” If you cannot fill in the blank in eight words or fewer, the chart is not ready.

Four rules of thumb:

Pick the chart that matches the message. A pie chart used to show change over time confuses; a line chart used to compare five unrelated categories clutters. Match first, then build.

One message per chart. If a chart needs a paragraph in the footnotes to explain what it is showing, it has failed. Either simplify the chart or split it into two.

Chartjunk and data-ink

Data visualization pioneer Edward Tufte gave us two complementary ideas:

• Chartjunk is any decoration that carries no information — three-dimensional bars that add visual depth but encode nothing, heavy gridlines that compete with the data, gradient backgrounds that exist only to look “designed.”
• Data-ink is the pixels that actually encode the data — the bars themselves, the line on a line chart, the numbers in a table.

Tufte’s principle: maximize data-ink; delete chartjunk. Every element on a chart that does not help the reader understand the message is working against you. Remove it.

A clean chart with four data points and no clutter communicates faster than a “beautiful” chart with thirty visual elements.

The pyramid principle, visualized

Putting it together: a checklist

Before you send any analysis or walk into any presentation, run through these four checks:

1. Does my first sentence state the recommendation? If not, rewrite the opening.
2. Does each chart have one clear message? If you cannot say the message in eight words, simplify the chart.
3. Have I deleted every element that is not data-ink? Remove the gradient, the heavy grid, the 3-D effect.
4. Would a busy executive understand the point in five seconds? If no, simplify until yes.

The goal is not to impress with complexity. The goal is a decision.

Next

You have finished the Business Analytics track. Every chapter has built toward this point: finding real patterns in data and communicating them so decisions actually happen. Circle back to any chapter to deepen a concept, or jump into the SQL and Visualization track to build these analyses yourself from raw data.