Jean Villedieu's slides from his talk at Connected Data London. Jean, co-founder of Linkuriuos presented a sarcastic presentation of what developers should consider when creating a graph visualisation.
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
The 8 do’s and don’ts of graph visualisations.
1. The 8 do’s and don’ts
of graph visualizations.
SAS founded in 2013 in Paris | http://linkurio.us | @linkurious
2. Introduction.
● Linkurious is a graph visualization startup.
● We help companies understand graph data.
● Linkurious Enterprise, an enterprise-ready
graph visualization platform.
● Customers like NASA, French Ministry of
Finances, F500s.
● Partnerships with Data to Value, Neo
Technology.
3. Why data visualization?
“The greatest value of
a picture is when it
forces us to notice
what we never
expected to see.” John Tukey (1962)
4. Some data is best represented as a
network of nodes and edges.
What are X's connections? What is
the influence of X in the network?
What's the shortest path between X
and Y?
Fraud, cyber-security, intelligence,
medical research.
Why graph visualization?
PERSON
name: Séb
age: 29
PERSON
name: Jean
age: 31
LOCATION
name: Paris
Lives
in Lives
in
Knows
5. No need to define goals and
expectations.
Your graph visualization will
automagically have positive results.
Administrate, understand, monitor?
Advice #1: don’t set (business) objectives.
6. Why understand your users, their
challenges, their habits.
You know what is right, why ask
other people?
Developers, data scientists,
analysts, public?
Advice #2: don’t consider your users.
7. You’re an artist and your graph
visualizations need to entertain.
3D, colored backgrounds, fancy
interactions.
Colors, sizes, glyphs, icons for
nodes & colors and sizes for edges.
Advice #3: treat it as an art project.
8. You know best, why would your
users need to ask their own
questions?
A static visualization means your
user is passively consuming (vs
answering his own questions).
Zooming, hover & tooltips, expand
on demand, search, filter, select.
Advice #4: don’t add interactivity.
9. Preparing and modelling your
(graph) data is simple and intuitive.
Data preparation is always time-
consuming, there are various ways
to model graph data.
Test and iterate.
Advice #5: don’t think about your data.
Software engineer
preparing a graph
visualization
project.
10. No need to provide guidance to
interpret your graph visualization.
Help your users correctly interpret
the information you provide.
Legend, labels, tooltips.
Advice #6: let the user figure it out.
11. It’s a contest, you need to display
as many nodes and edges as
possible.
Hardware constraints and cognitive
constraints, hairball.
Display what matters (10s, not
100,000s).
Advice #7: always display everything.
12. You can do it all, your prototype will
nicely move into production and be
maintained.
Security, collaboration, stability,
scalability, support, training.
Are you reinventing the wheel?
Advice #8: don’t worry about operational questions.
13. Disclaimer.
Some* of the advice
in these slides should
not be followed.
* actually all of the 8 advices should not be followed if you want your graph visualization project to
be successful ;)