Common Mistakes In Data Visualization

Embed Size (px)

Citation preview

Common mistakes in data visualizationAmedee Van GasseBarCampGhent319 December, 2009

$ whoami

Current day job:

MS Office/VBA specialist

Desktop & Mail Security (Antivirus, Spam,...)

Geek stuff:

Ubuntu tweaking

Drupal (http://foutparkeerders.be - with @peterdedecker)

Tik vzw (http://tik.be)

[email protected]
http://amedee.be
http://twitter.com/amedee
http://be.linkedin.com/in/amedee
http://facebook.com/amedee.vangasse

The power of data visualization

Chinese proverb: One Picture is Worth Ten Thousand Words

The Fast-Food Information Age

Pr0n for the stats junkPointy-Haired Boss

Twitter: Infograph of the day (@bnox)

What Makes Good Information Design?

Interestingness

Function

Form

Integrity

Interestingness: relevant, meaningfull, newFunction: easiness, usefulness, useability, fitForm: beatty, structure, appearanceIntegrity: truth, consistency, honesty, accuracy

Interesting + function = experimentFunction + form = sketchForm + integrity = eye-candyIntegrity + interesting = proof of concept

Not interesting = boringNot functional = uselessNo form = uglyNo integrity = rubbish

With Great Power...

... Comes Great ResponsibilityLook for the Hidden Agenda

Trust No One

Lies, Damned Lies, and Statistics

Let's just face it:

Why do we lie?

Charts, infographics,... are only a simplified representation of a complex reality

Compare with maps: they are also (white) lies

Used to convice others

Because it looks cool

Because we don't know any better

The axis are (not) your allies

Y axis starts at 500. Why 500?

Gradients

Grey on grey

Original chart has animation...

Why?

December = 1/5 size of January
= CRISIS!!!

Fear, Uncertainty and Doubt

Convince employees of needed budget cuts

Or... just blind faith in Excel???

What really happened

2008 up to Q3 was an exceptionally good year

Yes, there was a crisis, but...

We went to 1/2, not to 1/5

Excel bug: lowest value on Y-axis is magic...

What we saw in the meeting

versus 2008: - 1,3 %versus OB : + 0,6 %

47.988

47.115

47.385

2008

OB

2009

No scale

Why start at 40.000? Why not 30.000 or 45.000?

3D adds no value

A more accurate version

Scale added

Starts at 0

No 3D

versus 2008: - 1,3 %versus OB : + 0,6 %

You need X-Ray Vision for this one

Confession: I made this one

What if executed > planned? -> invisible!

Better chart: bars (executed) + lines (planned)

Pie chart hell

I made these too!

3D

Gradients

Could have been combined with previous chart

Combining the two previous charts

The donut: evil twin of the pie

Psychedelic

Hard to compare series

Can't get enough pie?

Only one thing is worse than one pie: two pies

Little information

You can tell your story with one chart

Pies are to scale

Pretty colors

3D: same diameter?Perspective?

The Mother of all Pie Charts

http://www.slideshare.net/netlash/old-media-vs-new-media

Sorry, @netlash, but this one really sucks

Why are Greece and Germany the same size?

Red = good or bad?

Lots of comments on InformationIsBeautiful.net

The original chart

Boring

Ugly

Easy to understand

By the way, did I already tell you that I really hate pie charts?

The only good pie charts

Pac Man Pie: @flexyflow

Want to see some more silly charts?

Spiderman loves this one!

Radar chart

Hard to read

Needs a lot of tweaking

Ugly even with perfect data

3D: Manhattan chart

You can always go... downtown

3D: unstacked area chart

This one comes with X-Ray glasses

Can you see S3 (yellow)?

Use a 2D line chart!

Some practical tips

Colors

Axes, legend & title

Logarithmic scale

Colors

No gradients in background

Background contrasts with objects

Use color only when needed

Different color = difference of meaning in data

Soft colors + bright/dark for highlights

Colors

Stick with a single hue

Non-data just visible enough

Avoid red+green (colorblind)

Leave special effects to Hollywood!!!

And the Oscar goes to...

Axes, legend & title

X Axis: Category or Value?

Line charts or XY Scatter charts?

Zero-based or not?

Second Y axis?

Good title and axis labels = no legend needed

If your chart needs a legend, it is a bad chart

Axes gone wrong

X axis is time + location

X and Y axes ar not linear

A (bad) joke from http://graphjam.com

Logarithmic scale

Nonlinear stimulus-response: 10x pay raise makes you only 2x happier

Avoid unless you know what you are doing

Can't be used with 0 in data

Use a trick: add 0.001

General tips

What is your message?

Who is your audience?

Prepare your data

Don't have blind faith in your software

Links

http://peltiertech.com/WordPress/category/chart-bustershttp://chandoo.org/wp/category/visualization/http://graphjam.com/http://www.informationisbeautiful.net/http://www.smashingmagazine.com/2009/09/11/
25-useful-data-visualization-and-infographics-resources/http://www.smashingmagazine.com/2007/08/02/
Data-visualization-modern-approaches/

Muokkaa otsikon tekstimuotoa napsauttamalla

Muokkaa jsennyksen tekstimuotoa napsauttamallaToinen jsennystasoKolmas jsennystasoNeljs jsennystasoViides jsennystasoKuudes jsennystasoSeitsems jsennystasoKahdeksas jsennystasoYhdekss jsennystaso

TON x1000

08/20071093

09/20071037

10/2007970

11/2007909

12/2007953

01/20081078

02/20081159

03/20081287

04/20081276

05/20081207

06/20081316

07/20081269

08/20081166

09/20081200

10/20081028

11/2008758

12/2008645

01/2009678

02/2009673

03/2009659

04/2009587

05/2009575

06/2009606

07/2009748

08/2009868

09/2009919

10/2009926

11/2009933

12/2009897

Planning migratie Notes -> Exchangeweekaantal mailboxen gemigreerdal gemigreerduitgevoerdniet gemigreerdcumul gepland

35/20072273316173

36/2007951353030210

37/20072281172904327

38/20073441072762440

39/20074511562639622

40/20076081822453820

41/200779313723251062

42/200792929820341360

43/2007122822318501574

44/2007145611917421698

45/2007157021715351928

46/2007179312013872062

47/2007191423011592304

48/200721562628892558

49/200724192416562799

50/200726602436703050

51/200729022484843279

52/2007315004843279

1/2008315004843279

2/2008315005063279

3/20083112514933327

4/20083163314323358

5/2008319404323358

6/20083190524193376

7/2008324204193376

8/2008323004263376

9/20083232234013399

10/2008325514033399

11/2008325604033399

12/20083256483053447

13/2008330403053447

14/20083362661813513

15/20083428211603534

16/2008301416883558

17/2008303068203624

18/200830980203624

19/200830980203624

20/200830981193624

21/20083106903644

2008OB2009

YTD 10/2009 (in K)479884711547385