Data Storytelling, and Every Way a Chart Can Mislead

1Why storytelling works

A raw fact activates only the language centres of the brain, and recall is weak.
The same fact inside a story activates language, emotion and sensory centres together.
That is where the roughly 22 times more memorable figure, attributed to Bruner, comes from.
People recall the plot of a favourite film in detail and almost nothing from their own company's annual report.

Fig 6.1The Significant Objects Project

Cheap items bought for around $1.25 were resold with an invented story attached, and fetched up to $8,000. The object never changed. The narrative created the value.

Note the axis is logarithmic, because on a linear axis the purchase prices would be invisible against the sale prices. Even in objective boardrooms, decisions are emotional, and storytelling is what bridges logic and emotion.

What data storytelling is

Fig 6.2Three parts, and what is missing when one is absent

All three are required, and this is a three-part definition worth learning as a definition. A beautiful chart with no narrative is a picture. A narrative with no context lands on the wrong audience. Hover each region for what you are left with when a part is missing.

The canonical example: the 1854 cholera map

Fig 6.3Broad Street, and the pump nobody had suspected

During a London outbreak nobody knew how the disease spread. A physician plotted cases on a street map, with bar thickness showing affected households. The cases clustered tightly around one water pump and thinned with distance from it. One visual, one narrative, one context, and a public health crisis was solved. This is the origin story of the field.

Colour conventions become narratives

The 2000 US election was the first time Democrats were shown in blue and Republicans in red. That association is now permanent, and appears even on plain bar charts with no map involved. Once a narrative embeds in perception it is very hard to change, which is both a tool and a warning.

Presentation frameworks

Presentation Zen · Garr Reynolds

Restraint: eliminate every unnecessary element. Simplicity: use white space deliberately, one idea per slide. Naturalness: authentic delivery, eye contact, genuine connection.

SUCCESS · Heath Brothers

Simple, Unexpected, Concrete, Credible, Emotional, Storied. The result is a presentation that is engaging, focused and memorable.

Pecha Kucha

20 slides × 20 seconds = 6 min 40 sec. The constraint does the teaching: with 20 seconds you cannot read out the slide, so the visual has to carry itself.

Fig 6.4Twenty seconds is how long you have per slide

20.0 s

Read the bullet-heavy slide out loud while the bar runs down. You will not finish. That is the whole point of the format: it forces you to identify what actually matters and cut the rest. Talk less, show more.

Used in businesses, business schools and conferences worldwide. Simplicity is the ultimate sophistication, and a clean visual delivers more than a cluttered one.

Death by presentation: NASA Columbia, 2003

Fig 6.5The foam-strike slide, and how it should have been built

This is Chapter 3's visual hierarchy lesson with the consequences attached. Where the most important line sits on the slide is not a style question. The critical warning, that the flight condition was far outside the tested database, was in the smallest font at the bottom of the slide.

Flaw	What happened
Clutter	Dense bullets, multiple font sizes, long sentences on one slide
Vague language	"Significant" repeated, never meaning statistical significance, so the word carried no fixed meaning
Poor hierarchy	The critical warning was in the smallest font at the bottom
Misleading title	"Review of Test Data Indicates Conservatism" made the slide sound reassuring

Storytelling frameworks

Fig 6.6Three shapes, and the data story each one carries

Which one you pick depends on the audience, not on the data. A story works through ordering: incident, rising action, climax, then the push to act, and you build the same arc into a sequence of visuals rather than dumping charts in the order you made them.

Types of narrative

Fig 6.7The author-driven to reader-driven spectrum, and the martini glass

The martini glass is the one to remember by name, because it names the sequence: control first, freedom second. It begins as a tight author-driven narrative, the stem, then opens out into free exploration, the glass. The other hybrid techniques are drill-down and the interactive slideshow.

	Author-driven	Reader-driven	Hybrid
Character	Linear, static, prescribed message	Interactive, exploratory	Guided structure with interactive parts
Format	Magazine infographics, decks, video	Tableau or Power BI dashboards	Interactive slideshow, dashboard with a set entry point
Use when	You need the audience to reach one conclusion	Users must find answers relevant to them	You want both

Three dimensions separate two people telling the same story with the same data: genre (annotated chart, comic strip, slideshow, video, flowchart), visual narrative tactics (highlighting, transitions, animations, visual structuring) and narrative structure (ordering, interactivity, messaging).

The seven data story types

Fig 6.8Each type, with the chart it usually needs

Drilling down and zooming out are opposites, and both are legitimate stories: one narrows to find the cause, the other widens to find the significance. Intersection is about the crossover moment, not the levels, so the story is when the order swapped, not what the values are.

Misleading with charts

The same tools that make a story clear can make it false. Most of these are not deliberate deception, which is why you need to recognise them in your own work.

Axis problems

Fig 6.9Drag the baseline and watch a modest change become a cliff

Truncating the y-axis to a narrow band turns a change of a few percentage points into a visual cliff. Show the same data against a zero baseline and the change is modest. Bar length stops being proportional to value the moment the baseline is not zero.

Fig 6.10The inverted axis, and why it is the sharpest case

A chart of firearm deaths was drawn with the y-axis running downward, so a genuine increase read as a decrease to every casual viewer. The designer's stated reason was that deaths are negative, so a downward line should mean improvement. The result was a chart that told the opposite of the truth. The rule: never fight the natural ordering. Use colour to carry the emotional signal and leave the axis conventional.

On dual axis charts, label both scales explicitly. Without that, a flat-looking series may be steep and a steep one trivial. As a reader, always check the axis first. As an analyst, never manipulate it.

Bar length that does not match the quantity

Fig 6.11Shrinking bars over rising prices

A chart showed petrol prices with shrinking bars and a large "28% decrease" callout. Prices had actually risen. What had fallen was the rate of increase. Because detection reads bar length as quantity, the chart says something its own numbers contradict.

Simpson's paradox

Fig 6.12B wins both days. A wins overall

	Saturday	Sunday	Overall
Person A	7/8 = 87.5%	1/2 = 50%	8/10 = 80%
Person B	2/2 = 100%	5/8 = 62.5%	7/10 = 70%
Winner	B	B	A

Nothing is wrong with any of the four numbers. The reversal comes from how many tasks each person did on each day: A did most of their work on the easy day and B did most of theirs on the hard one. The design response: present the aggregate, but let the user drill down. This is exactly why dashboards need both author-driven and reader-driven elements.

Drill-down order changes the narrative

Fig 6.13Same voters, same total, two different stories

The final number is identical. The story is not. Same endpoint, two different emphases, and each nudges the audience toward a different conclusion. Offering both drill paths is the honest option, so the audience reaches its own reading.

Cherry picking

Fig 6.14Drag the start year and choose your own conclusion

A chart of UK national debt starting at 1995 shows debt at an all-time high. Extend the same series back through the full history and earlier periods exceeded 200%, far above the recent figure. The recent rise is real; presenting only the recent window inflates it out of proportion. Prioritise accuracy over aesthetics, and if you must truncate a range, add a caveat giving the wider context.

Colour saturation working against the data

Fig 6.15A legend that points attention at the wrong states

On a case-count map the legend gave the darkest shade to one band and lighter shades to others in a way that did not track the actual counts. States with far fewer cases drew the eye more strongly than the states with the most. Saturation must increase with the quantity it represents, and readers should check the legend before drawing conclusions.

Key points

Why stories are retained better, and the roughly 22× figure attributed to Bruner.
The Significant Objects Project, and what it demonstrates about narrative and value.
Data storytelling as the combination of visualization, narrative and context.
The 1854 cholera map, and why it is the founding example.
Presentation Zen: restraint, simplicity, naturalness.
The SUCCESS model, all six letters.
Pecha Kucha: 20 slides, 20 seconds, and why the constraint improves the talk.
The NASA Columbia slide: four flaws and the three fixes.
Three storytelling frameworks: monomyth, story mountain, nested loops.
Three narrative types, and what the martini glass technique does.
Three dimensions of narrative: genre, visual tactics, structure.
All seven data story types with the objective of each.
Why drilling down and zooming out are opposite but equally valid.
Four axis manipulations and the effect of each.
Why bar length must stay proportional, using the petrol price case.
Correlation against causation, and what a confounding variable does.
Simpson's paradox: be able to construct the reversal and explain it.
Why drill-down order changes the narrative even when the answer is identical.
Cherry picking, and prioritising accuracy over aesthetics.
How a badly built colour legend can point attention at the wrong values.