Last reviewed: Next review due:
The central challenge of data storytelling
Data journalism produces rigorous findings. Data storytelling communicates them in a way that a reader will understand, care about, and remember. The two are not the same skill, and many technically excellent analyses are buried in unreadable articles. The best data journalism combines both: the finding is rock-solid, and the writing is clear, human, and purposeful.
The core discipline: ask yourself, after every paragraph, “So what?” If you cannot answer that with a sentence about impact or significance, cut the paragraph.
Structure for a data story
1
The finding
Lead with the most significant finding, expressed in plain English with human impact. Not "the data shows a 23.7% increase" — "nearly a quarter more children went to food banks last year, new figures show."
2
The human angle
Place a real person in the data as soon as possible. The number is real — find the person who embodies it. This is where the data story connects with the reader.
3
The context
How does this compare to last year? To other areas? To the national average? Context transforms a number into meaning. Is 23% high or low? Good or bad? Compared to what?
4
The explanation
Why might this be happening? What do experts say? What does the institution responsible say? Data tells you what happened; sources tell you why.
5
The methodology note
Explain briefly how you calculated the finding and link to the underlying data. This builds trust and allows readers to verify your work.
What NOT to include
- ×Every number in your analysis — publish only the numbers that support the specific editorial point. Everything else belongs in a data table for those who want it.
- ×Technical methodology details in the body copy — move these to a footnote or separate methodology page.
- ×Too many caveats in the main text — note uncertainty clearly but briefly; do not let caveats overwhelm the story.
- ×Charts that duplicate prose — if you describe a trend in words, you do not usually need a chart of it too.
- ×Jargon — "coefficient," "p-value," "regression" are fine in methodology notes, not in news copy.
When data storytelling is hardest
- 1When the finding is genuinely uncertain — you must communicate honest uncertainty without losing reader trust.
- 2When the number is counterintuitive and readers may not believe it.
- 3When the story requires understanding context that most readers lack.
- 4When the finding contradicts a widely-held belief — you must show your working carefully.
Red flags
- The headline makes a stronger claim than the data supports.
- The human case study is the only person in your dataset with that experience — it may be an outlier, not the story.
- You are hiding uncertainty in the methodology note that belongs in the main text.
- The story has a lot of numbers but no explanation of what they mean for real people.
- You are asking readers to believe a counterintuitive finding without showing your working.
Data storytelling checklist
- My headline accurately reflects what the data shows — not a stronger claim.
- I have found a human case study who embodies the data finding.
- I have provided context: comparison to previous period, national average, or comparable area.
- I have explained uncertainty clearly but briefly in the main text.
- I have removed any number that does not directly support the editorial point.
- I have a methodology note (inline or linked) that explains my key calculation.
- I have offered the institution or individuals involved a right of reply before publication.
Ready to publish?
Before publishing, check our data journalism publishing workflow — including fact-checking, source citation, and the corrections process for data errors.
Common mistakes
- Leading with methodology ("we analysed...") rather than the finding.
- Using jargon that the average reader will not understand.
- Publishing a chart without a headline that says what the chart shows.
- Picking a human case study before checking whether their experience is typical of the data.
- Burying the most significant finding deep in the article.
Related guides
Primary sources
Frequently asked questions
What is the difference between data-driven and data-led journalism?
Data-led journalism starts with a dataset and asks what story is inside it — the data suggests the angle. Data-driven journalism starts with a question or hypothesis and uses data to answer it. Both are valid approaches. Data-led can surface stories a journalist would never have thought to look for; data-driven ensures the evidence genuinely answers the editorial question the headline implies.
When should I use a chart instead of prose?
Use a chart when: the comparison between multiple values is the point (bar chart); the trend over time is the point (line chart); the geographic distribution is the point (map). Use prose when: there is only one or two numbers to communicate; the relationship between numbers is complex and needs explanation; or the chart would distract from a more powerful human narrative. Do not use a chart just because you can.
How do I explain uncertainty to a general audience?
Avoid statistical jargon. Instead of 'margin of error ±3pp at 95% confidence,' write 'the poll suggests around 42% support, but it could be anywhere between 39% and 45%.' Instead of 'correlation of r=0.7,' write 'areas with more fast food outlets tend to have higher rates of obesity, though other factors are also involved.' The goal is honesty about uncertainty without losing your reader.
Related guides
Primary sources
- Datawrapper — chart headline and description guide— Datawrapper
- Guardian Visuals — data journalism methodology— The Guardian
- Full Fact — reporting numbers accurately— Full Fact
- ONS — official statistics for story angles— ONS
- Bureau Local — collaborative data storytelling— The Bureau of Investigative Journalism
- Flourish — storytelling templates— Flourish