Science Of Data Visualization - Tableau Conference 2018

Transcription

Welcome

#TC18The Science of Data VisualizationLarry SilversteinStrategic Sales ConsultantTableau

Start Your (Visualization) Engines

AgendaThe science of data visualizationBest practices for building dashboards and visualizations

The Science ofData Visualization

Share of Worldwide Urban Population Growth2010–2050

Share of Worldwide Urban Population Growth2010–2050

89

55.7386.89ValueMean of x in each case9 (exact)Variance of x in each case11 (exact)Mean of y in each case7.50 (to 2 decimal places)Variance of y in each case4.122 or 4.127 (to 3 decimalplaces)Correlation between x andyin each case0.816 (to 3 decimal places)Linear regression line ineach casey 3.00 0.500x (to 2 and 3decimal places, respectively)

Thinking is Slow. Seeing is FAST.CerebralCortexThinkingVisualCortexSeeing

Preattentive Attributes

Memory LimitsHave we gained or lost customers over the last 4 years?Which city has grown the fastest?Total # of Registered Customers2009201020112012New 390AustinTotal3614825855951550176717871779

Visual Representation vs. TableHow do each of our locations compare?

Memory Limits16 Numbers4 Patterns

Overcoming Memory LimitsUse familiar chart typesDon’t make people remember viewsAvoid large legendsUse intuitive colors and shapes

Familiar Chart Type? Intuitive Encoding?

Familiar Chart Type? Intuitive Encoding?

Familiar Chart Type. Intuitive Encoding.

Visual Interruptions Make People Slow

Visual Interruptions Make People Slow

Color, Data Types, ChartTypes and Layouts

Best Practices for Color

Color Vision Deficiency (CVD)

Bars and Colors—Caution!Be thoughtful when applying color to bars

Bars and Colors—Caution!Be thoughtful when applying color to barsThis is OK

Color PerceptionColor perception is relative, not absolute

Color PerceptionProvide a consistent background

Too Many ColorsHumans can only distinguish 8 colorsThis is not helpful.

Right Number of ColorsHumans can only distinguish 8 colorsThis is helpful.

Types of DataQualitative (nominal)Arizona, New York, TexasSarah, John, MariaCoors, Bud Light, Stella ArtoisQualitative (ordinal)Gold, silver, bronzeExcellent health, good health, poor healthLove it, like it, hate itQuantitativeWeight (10 lbs, 20 lbs, 5000 lbs)Cost ( 50, 100, 0.05)Discount (5%, 10%, 12.8%)

How Do Humans Like Their Data?PositionMore importantColorSizeShapeLess important

1. Position

1. Position

What Goes Where?EmphasizedNeither emphasizednor de-emphasizedEmphasizedNeither emphasizednor de-emphasizedDe-emphasized

What Goes Where? Research re-72395

Best Practices forBuilding Dashboardsand Visualizations

Tables

Graphs Tables for Spotting Trends

Chart Type GeneralizationsTime: on an x-axisLocation: on a mapComparing values: bar chartExploring relationships: scatter plotRelative proportions: treemap

ShowMe: Visual Best Practices - Automatically

Pies vs. Bars (An Unfair Fight, IMO)

A Tip for Using Stacked Bars

Bullet Chart GaugesBackground fill colors that encode qualitative ranges like bad,satisfactory and goodQuantitative ScaleText LabelBar that encodes theperformance measureSymbol marker that encodes thecomparative measure

Bars vs. LinesUse line graphs to encode trends over time

Avoid Food Graphs

Tastier Solutions—Part 1

Tastier Solutions—Part 1

Tastier Solutions—Part 2

Tastier Solutions—Part 2

Tastier Solutions—Part 3

Tastier Solutions—Part 3

Making a Point—About Lines

Making a Point—About Lines

Making a Point—About Lines

Be Careful Using Dual s

3D?!? Make Mine Small Multiples!

Mapping to InsightUse maps when location is relevant

“Choropleth” Maps

“Choropleth” Maps

Symbol Maps

A Strike Zone is a Map?!?

Tell the Truth!

So This Is What Really Happened

Sometimes, the Default isn’t Best

Experiment—You Won’t “Break” your Data

Let’s Talk About Dashboards

Are All Dashboards the Same?

20-look-at-our-participation.html

Iraq: Deaths on the lofiej-20-look-at-our-participation.html

Which One to Use?

The Same GoalsMake better, datadriven decisionsMake changes forthe better

The 5 Second Test

Dashboarding for the 5 Second TestMost important view goes ontop or top-leftLegends go near their viewsAvoid using multiple colorschemes on a singledashboardUse 5 views or fewer indashboardsProvide interactivity

Dashboarding for the 5 Second TestUse your words!TitlesAxes & UnitsKey facts and figuresRemove extra digitsin numbersGreat tooltips

Dashboarding for the 5 Second Test

Dashboarding for the 5 Second Test

Is Beautiful Design Important?

Recommended irtual?course 2346

#TC18Thank you!Larry Silversteinlsilverstein@tableau.com

Please complete thesession survey from the MyEvaluations menuin your TC18 app

The Science of Data Visualization #TC18 . Tableau. Start Your (Visualization) Engines. Agenda The science of data visualization Best practices for building dashboards and visualizations. The Science of Data Visualization. Share of Worldwide Urban Population Growth