Storytelling with Data: Why Order Matters

Presenting data is as important as exploring it. Stories give facts a memorable shape, and the order of the pieces changes what people understand and remember.

Based on Kosara & Mackinlay, Storytelling: The Next Step for Visualization, and Hullman et al., A Deeper Understanding of Sequence in Narrative Visualization (2013)
The case

From exploring data to presenting it

Visualization research has focused on helping analysts find answers. But analysts usually are not the people who decide, so findings have to be communicated.

1

Explore

Find patterns with flexible, interactive tools.

2

Analyze

Understand causes and test ideas.

3

Present

Persuade and be remembered. The next research frontier.

Not new. Nightingale's charts, Snow's cholera map and Minard's Napoleon map were made to show and explain, not to discover.
The idea

What is a data story?

An ordered sequence of steps with a clear path. Each step can be a chart, text, an image or video.

Overview→Detail→Cause→Outcome
Order carries meaning. It signals time and causality, and audiences understand and recall scrambled stories worse than well-ordered ones.
Settings

Three ways stories are told

Self-running

News graphics. A teaser draws readers in; no author to answer questions.

Live talk

Like Hans Rosling's Gapminder. The presenter can pause, annotate and respond.

Small group

Most interactive. Off-script questions are expected, and can feed the next revision.

Structure

Patterns professionals use

An analysis of 42 professional narrative visualizations found a few recurring moves.

Martini glass

Broad introduction, narrow to a point, then open up to exploration.

Parallelism

Repeat the same pattern of slides for each part, as in the NYT Copenhagen piece on drought, flooding and crops.

Dialogue

Pose a question, then show the chart that answers it.

Slide-to-slide transitions usually change just one thing. How common each family was:

Temporal
88%
Granularity
71%
Comparison
64%
Causal
24%
Dialogue
17%
Key principle

Change one thing at a time

Designers hold most of a chart constant and vary a single dimension, building complexity gradually. Gapminder does this step by step: distribution, area chart, bubbles, then a second axis.

Transformation cost

Count how many of four attributes change between two charts: independent variable, dependent variable, time, level of detail.

Example

Male SAT scores 2010 to female SAT scores 2009 costs 2. Switching to TOEFL scores costs 3.

Balance. Too little change is boring; too much overloads the viewer.
Evidence

What audiences prefer

In a crowdsourced study (143 people), participants picked which chart should come next.

Low coststrongly preferred over higher-cost transitions
Time firstchronological moves beat other types

Preference order when cost is equal:

Temporal>Dimension = Measure>Granularity
Memory

Repetition helps recall

73 students saw six charts in one of four orders, then tried to reconstruct the order.

Perfect parallelism

Best memory for the original order (p = 0.002 overall).

Reversed variants

Lower total cost, yet remembered worse and slightly harder to explain.

Lesson. Minimizing local changes is not enough. Good sequences also need repeated, recognizable structure.
Craft

Interaction, annotation, tools

Use interaction sparingly

Best after the story ends, when pausing for questions, or as simple highlighting that keeps readers on the thread.

Guide with annotation

Text, arrows and narration help, but too much text drowns the data.

Let tools suggest orders

Treat each chart as a node in a graph, score transitions by cost and type, and favor parallel patterns.

Takeaways

A short checklist

  • Build the story as an ordered path with clear time or cause.
  • Change one dimension between views; build complexity gradually.
  • Prefer time-based moves; use detail shifts with care.
  • Repeat structural patterns across parts to aid memory.
  • Keep interaction light so the thread is not lost.
  • Judge success by engagement, memory and decision quality, not just speed and accuracy.
Open questions

What research still needs

Which animation helps, and which only distracts?

Can narration or annotation offset a complex transition?

Which chart types work best as storytelling devices?

How should stories be evaluated, and what can film and journalism teach?