PRINCIPLES OF GOOD DATA VIZ AND STORYTELLING ANA-MARIA MOCANU DATA SCIENTIST, KANTAR INSIGHT
PRINCIPLES OF GOOD DATA VIZ AND STORYTELLING
ANA-MARIA MOCANU
DATA SCIENTIST, KANTAR INSIGHT
1. KNOW YOUR AUDIENCE
Exec team?
Finance team?
Dev team?
2. KNOW WHAT TO AVOID
What's my favourite data task?
Data viz
Machinelearning
Text Analytics
Dataintegration
Communication
Data cleaning
Data viz Machine learning
Text Analytics Data integration
Communication Data cleaning
Data viz
Machine learning
Text Analytics
Data integration
Communication
Data cleaning
0% 10% 20% 30% 40%
3. KNOW THE DATA INSIDE OUT
• Visuals are much easier once the underlying data model is robust and correct
• Know differences between Counts vs. Distinct counts; Left join vs. Inner join
• Understand how you want to slice the data early on
• If you need to use assumptions in the data, validate them and keep them in mind throughout the way
• Always check the results of the visualisation!
4. KEEP IT CLEAN AND SIMPLE (IT’S MORE DIFFICULT THAN IT SOUNDS)
• Don’t overcomplicate dashboards - great stories can be told with simple KPIs and charts
• Limit yourself to answering max. 3 questions in your project. Examples can be:
• Where are your users coming from?
• Which activities are contributing to the highest ROI ? Where are you making the most impact?
• Where are you losing out? Where can you do better?
• Who is your competition and how are they doing?
5. A DATA STORY CAN BE TOLD IN MANY WAYSNumber of U.S. soldiers killed in the Iraq war from 2003 to 2017*
*Source: https://www.statista.com/statistics/263798/american-soldiers-killed-in-iraq/
6. USE CHART FEATURES – COLOUR, SIZE AND SHAPE
Colour: Sentiment (from extremely fun
to plain boring/ not helpful)
X-axis: Amount of knowledge gained
Y-axis: Level of contribution to my
career
We can add different shapes and
sizes to analyse other variables (e.g.
number of hours spent)
7. USE CHART TYPES
*Source: https://plot.ly/r/sankey-diagram/
There are many chart types in the
world. Sankey diagrams are good for
visualising flows, mosaic charts are
good for categories etc.
Don’t forget Principle 4 – Keep it
clean and simple
8. USE YOUR IMAGINATION
• Bring in an
additional dataset
• Create a chart no
one saw before
• Find something funny
in the data that’s
worth telling
*Source:
http://www.tylervigen.co
m/spurious-correlations
For example, how to educate that correlation does not imply causation:
9. REACH OUT FOR HELP
• Ask your end users for their opinion. Show your dashboard to someone creative. Or someone who
has no idea what’s going on – do they understand your story?
• Most of the visualisation tools have forums you can reach out. Have a regular look.
• Also, keep an eye on things like “Viz of the day” and media reports and infographics.
SO, IN SUMMARY:
1. Know your audience
2. Know what to avoid
3. Know the data inside out
4. Keep it clean and simple
5. Choose how you want to tell the story – It can be told in many ways
6. Use chart features
7. Use chart types
8. Use your imagination
9. Reach out for help
“NUMBERS HAVE AN IMPORTANT STORY TO TELL. THEY RELY ON YOU TO GIVE THEM A CLEAR AND CONVINCING
VOICE.”
STEPHEN FEW