18
Selecting a Visual Analytics Application AUTHORS: Professor Pat Hanrahan Stanford University CTO, Tableau Software Dr. Chris Stolte VP, Engineering Tableau Software Dr. Jock Mackinlay Director, Visual Analytics Tableau Software

Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Sele

ctin

g a

Vis

ual A

naly

tics A

pp

licatio

n

AU

TH

OR

S:

Pro

fes

so

r Pa

t Ha

nra

ha

n

Sta

nfo

rd U

niv

ers

ity

CT

O, T

ab

lea

u S

oft

wa

re

Dr. C

hris

Sto

lte

VP, E

ng

ine

erin

g

Ta

ble

au

So

ftw

are

Dr. J

oc

k M

ac

kin

lay

Dire

cto

r, Vis

ua

l An

aly

tics

Ta

ble

au

So

ftw

are

Page 2: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Na

me

Infla

tion

: Vis

ua

l An

aly

tics

Vis

ua

l an

aly

tics is

be

co

min

g th

e fa

ste

st w

ay fo

r pe

op

le to

ex

plo

re

an

d u

nd

ers

tan

d d

ata

of a

ny s

ize. M

an

y c

om

pa

nie

s to

ok n

otic

e w

he

n

Ga

rtne

r cite

d in

tera

ctiv

e d

ata

vis

ua

lizatio

n a

s o

ne o

f the

top

fiv

e

tren

ds tra

nsfo

rmin

g b

usin

ess in

tellig

en

ce

. Ne

w c

on

fere

nce

s h

av

e

em

erg

ed

to p

rom

ote

rese

arc

h a

nd

be

st p

ractic

es in

the a

rea

,

inclu

din

g V

AS

T (V

isu

al A

na

lytic

s S

cie

nce &

Te

ch

no

log

y), o

rga

nize

d

by th

e 1

00,0

00 m

em

be

r IEE

E. T

ech

no

log

ies b

ase

d o

n v

isu

al

an

aly

tics h

av

e m

ov

ed

from

rese

arc

h in

to w

ide

sp

rea

d u

se in

the la

st

fiv

e y

ea

rs, d

rive

n b

y th

e in

cre

ase

d p

ow

er o

f an

aly

tica

l da

tab

ase

s

an

d c

om

pu

ter h

ard

wa

re. T

he IT

de

pa

rtme

nts

of le

ad

ing

co

mp

an

ies

are

incre

asin

gly

reco

gn

izing

the n

ee

d fo

r a v

isu

al a

na

lytic

s s

tan

da

rd.

No

t su

rpris

ing

ly, e

ve

ryw

he

re y

ou

loo

k, s

oftw

are

co

mp

an

ies a

re

ad

op

ting

the te

rms “

vis

ua

l an

aly

tics”

an

d “

inte

ractiv

e d

ata

vis

ua

lizatio

n.”

To

ols

tha

t do

little m

ore

tha

n p

rod

uce c

ha

rts a

nd

da

sh

bo

ard

s a

re n

ow

lay

ing

cla

im to

the la

be

l.

Ho

w c

an

yo

u te

ll the c

lev

erly

na

me

d fro

m th

e g

en

uin

e? W

ha

t sh

ou

ld

yo

u lo

ok fo

r? It’s

imp

orta

nt to

kn

ow

the

de

fin

ing

ch

ara

cte

ristic

s o

f

vis

ua

l an

aly

tics b

efo

re y

ou

sh

op

. Th

is p

ap

er in

trod

uce

s y

ou

to th

e

se

ve

n e

sse

ntia

l ele

me

nts

of tru

e v

isu

al a

na

lytic

s a

pp

lica

tion

s.

Fig

ure

1: T

he

re a

re s

eve

n e

sse

ntia

l ele

me

nts

of a

vis

ual a

naly

tics a

pp

licatio

n.

Do

es a

true v

isu

al a

na

lytic

s a

pp

lica

tion

als

o in

clu

de s

tan

da

rd

an

aly

tica

l fea

ture

s lik

e p

ivo

t-tab

les, d

ash

bo

ard

s a

nd

sta

tistic

s? O

f

co

urs

e—

all g

oo

d a

na

lytic

s a

pp

lica

tion

s d

o. B

ut n

on

e o

f tho

se

fea

ture

s c

ap

ture

s th

e e

sse

nce

of w

ha

t vis

ua

l an

aly

tics is

brin

gin

g to

the w

orld

’s le

ad

ing

co

mp

an

ies. V

isu

al a

na

lytic

s is

a n

ew

are

a o

f

tech

no

log

y, a

nd

it ad

ds s

om

eth

ing

sp

ecia

l to y

ou

r bu

sin

ess

inte

llige

nce to

olk

it.

© 2

00

9 T

ab

lea

u S

oftw

are

1

Page 3: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

De

finin

g V

isu

al A

na

lytic

s: F

irst, W

ha

t It’s

No

t

Le

t’s s

tart w

ith w

ha

t vis

ua

l an

aly

sis

is n

ot: A

gra

ph

ica

l de

pic

tion

of

da

ta. V

irtua

lly a

ny s

oftw

are

ap

plic

atio

n c

an

pro

du

ce a

ch

art, g

au

ge

or d

ash

bo

ard

. Vis

ua

l an

aly

tics o

ffers

so

me

thin

g m

uch

mo

re

pro

fou

nd

. Vis

ua

l an

aly

tics is

the p

roce

ss o

f an

aly

tica

l rea

so

nin

g

facilita

ted

by in

tera

ctiv

e v

isu

al in

terfa

ce

s.

Pe

op

le w

ho

un

de

rsta

nd

vis

ua

l an

aly

tics k

no

w th

ere

’s s

om

eth

ing

mis

sin

g in

the s

ha

llow

ch

artin

g w

izard

s a

nd

da

sh

bo

ard

s fo

un

d in

mo

st b

usin

ess in

tellig

en

ce p

acka

ge

s a

nd

sp

rea

dsh

ee

ts. W

hile

ch

arts

an

d d

ash

bo

ard

s a

re in

de

ed

“v

isu

aliza

tion

s,”

the

y le

av

e o

ut

thre

e c

ritica

l ste

ps: e

xp

lora

tion

, an

aly

sis

an

d c

olla

bo

ratio

n. A

ch

art,

for in

sta

nce

, sh

ow

s c

on

clu

sio

ns, b

ut n

ot th

e th

ou

gh

ts b

eh

ind

it. No

r

ca

n u

se

rs u

se a

ch

art to

ask q

ue

stio

ns a

nd

thin

k fu

rthe

r. In a

ch

art,

the th

inkin

g h

as ta

ke

n p

lace a

lrea

dy

an

d th

e re

su

lting

vis

ua

lizatio

ns

are

little m

ore

tha

n a

sh

ow

.

Vis

ua

l an

aly

tics is

a m

ea

ns o

f ex

plo

ring

an

d u

nd

ers

tan

din

g d

ata

. It

su

pp

orts

an

d a

cce

lera

tes th

e a

na

lysis

pro

ce

ss its

elf. Y

ou

ca

n a

sk a

qu

estio

n, g

et th

e a

nsw

er, a

nd

ask fo

llow

-up

qu

estio

ns—

all w

ithin

vis

ua

l inte

rface

s. A

sto

ry u

nfo

lds fro

m o

ne

vis

ua

l su

mm

ary

to

an

oth

er. Y

ou

ma

inta

in y

ou

r train

of th

ou

gh

t with

ou

t takin

g y

ou

r

ey

es o

ff the d

ata

. La

ter, y

ou

ca

n re

trace th

e s

tory

to re

thin

k, e

xp

lore

furth

er a

nd

sh

are

. In s

ho

rt, vis

ua

l an

aly

tics a

llow

s y

ou

to g

o in

an

y

dire

ctio

n w

ith y

ou

r tho

ug

hts

wh

ile le

ve

rag

ing

yo

ur v

isu

al p

erc

ep

tua

l

sy

ste

m to

gu

ide y

ou

do

wn

the m

ost u

se

ful p

ath

s.

Wh

o is

Ad

op

ting

Vis

ua

l An

aly

tics?

Vis

ua

l an

aly

tics is

be

ing

ad

op

ted

by th

e w

orld

’s le

ad

ing

co

mp

an

ies,

un

ive

rsitie

s a

nd

go

ve

rnm

en

t ag

en

cie

s. F

rom

the w

orld

’s la

rge

st a

nd

mo

st in

no

va

tive o

rga

niza

tion

s –

Pro

cto

r & G

am

ble

, Ap

ple

, Pfi

zer,

Mic

roso

ft, Co

ca C

ola

, Go

og

le, C

orn

ell U

niv

ers

ity, P

rog

ressiv

e

Insu

ran

ce

, Am

azo

n, G

eo

rge

tow

n U

niv

ers

ity, th

e V

A (V

ete

ran

’s

Ad

min

istra

tion

), Blu

e C

ross B

lue S

hie

ld –

to o

ne

-pe

rso

n c

on

su

lting

sh

op

s, v

isu

al a

na

lysis

too

ls a

re n

ow

ma

instre

am

.

Wh

ere

vis

ua

l an

aly

tics w

as o

nce th

ou

gh

t to b

e in

the d

om

ain

of

scie

ntis

ts a

nd

en

gin

ee

rs, p

eo

ple

no

w re

co

gn

ize th

at v

isu

al a

na

lysis

acce

lera

tes b

usin

ess a

na

lytic

s. E

ve

n p

eo

ple

wh

o h

av

e m

ad

e d

o w

ith

Exce

l an

d w

izard

-driv

en

ch

artin

g c

an

be

ne

fit fro

m to

da

y’s

vis

ua

l

an

aly

tics a

pp

lica

tion

s –

the a

pp

lica

tion

s h

av

e b

eco

me

ea

sy e

no

ug

h

tha

t an

yo

ne c

an

use th

em

.

20

09 T

ab

lea

u S

oftw

are

Page 4: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

As

se

ss

ing

Fe

atu

res o

f Vis

ua

l An

aly

tics

Ap

plic

atio

ns fo

r Bu

sin

es

s D

ec

isio

n M

ak

ing

If vis

ua

l an

aly

tics is

mo

re th

an

just p

rese

ntin

g d

ata

vis

ua

lly, th

en

asse

ssin

g v

isu

al a

na

lysis

ap

plic

atio

ns is

mo

re th

an

just lo

okin

g fo

r

so

ftwa

re th

at c

an

pro

du

ce c

oo

l vis

ua

lizatio

ns. T

o d

istin

gu

ish

go

od

vis

ua

l an

aly

tics to

ols

from

ba

d, it’s

critic

al to

focu

s o

n th

e s

ev

en

esse

ntia

l ch

ara

cte

ristic

s th

at d

istin

gu

ish

a v

isu

al a

na

lytic

s

ap

plic

atio

n.

Vis

ua

l Ex

plo

ratio

n

Th

e fi

rst c

ha

racte

ristic

of a

vis

ua

l an

aly

tics a

pp

lica

tion

is th

e m

ost

imp

orta

nt: T

he a

pp

lica

tion

un

ifie

s th

e s

tep

s o

f qu

ery

ing

, ex

plo

ring

an

d v

isu

aliza

tion

da

ta in

to a

sin

gle

pro

ce

ss. D

o th

e d

ata

an

d th

e

vis

ua

lizatio

n w

ork

in ta

nd

em

? W

he

n th

e u

se

r pu

lls o

n th

e

vis

ua

lizatio

n th

e d

ata

sh

ou

ld c

om

e a

lon

g. W

ha

t do

es th

is m

ea

n? It

me

an

s p

eo

ple

ca

n g

o in

an

y d

irectio

n w

ith th

eir th

ou

gh

ts, w

itho

ut

ev

er lo

sin

g th

eir tra

in-o

f-tho

ug

ht. T

he

y m

ay n

ot e

ve

n h

av

e a

sp

ecifi

c q

ue

stio

n, b

ut a

s th

ey m

ov

e th

rou

gh

the d

ata

vis

ua

lly, th

ey

no

tice s

om

eth

ing

an

d th

at p

rom

pts

a q

ue

stio

n a

nd

the

n a

follo

w u

p

an

d s

o o

n, e

ve

ntu

ally

lea

din

g to

insig

ht. It m

ea

ns th

e v

isu

aliza

tion

s

in a

vis

ua

l an

aly

tics a

pp

lica

tion

allo

w p

eo

ple

to s

top

an

d ta

ke a

clo

se

r loo

k. F

ilterin

g, g

rou

pin

g, s

ortin

g a

nd

drillin

g a

ll take p

lace

with

in th

e v

isu

aliza

tion

itse

lf, with

a c

lick. A

use

r ma

y s

tart w

ith

ba

sic

qu

estio

ns a

nd

the

n, b

ase

d o

n v

isu

al c

ue

s a

nd

insig

ht, d

ee

pe

n

the in

qu

iry. T

he q

ue

stio

ns m

ay m

ea

n e

limin

atin

g s

om

e d

ata

,

va

lida

ting

it, or re

ach

ing

for a

ne

w s

et o

f da

ta a

ltog

eth

er. A

vis

ua

l

an

aly

tics a

pp

lica

tion

he

lps p

eo

ple

do

all o

f this

vis

ua

lly, a

nd

on

-the

-

fly. It’s

like th

e e

arly

da

ys o

f the w

eb

an

d M

osa

ic, w

he

n p

eo

ple

firs

t

sta

rting

ge

tting

the id

ea o

f stre

am

of c

on

scio

usn

ess e

xp

lora

tion

via

hy

pe

rlinks. It w

as a

ga

me c

ha

ng

er.

20

09 T

ab

lea

u S

oftw

are

Page 5: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Fig

ure

2: V

isu

al a

naly

sis

ap

plic

atio

ns s

up

po

rt the c

ycle

of v

isu

al a

naly

sis

. Wh

en

pe

op

le a

re e

xp

lorin

g d

ata

, they e

nte

r into

a c

ycle

wh

ere

at a

ny p

oin

t they m

ay

ne

ed to

skip

ste

ps, b

ack-u

p, s

ee

k a

dd

ition

al d

ata

, or e

ve

n s

tart o

ve

r. Vis

ual

an

aly

sis

ap

plic

atio

ns s

up

po

rt this

pro

ce

ss o

f vis

ual d

ata

exp

lora

tion

.

Ho

w d

o y

ou

kn

ow

if an

an

aly

tics a

pp

lica

tion

is a

pp

rop

riate

ly

de

sig

ne

d to

su

pp

ort v

isu

al e

xp

lora

tion

? H

ere

are

a fe

w te

sts

:

da

ta? U

sin

g th

e s

oftw

are

sh

ou

ld b

e s

o e

asy th

at p

eo

ple

do

n’t

ev

en

thin

k a

bo

ut th

e m

ech

an

ics o

f cre

atin

g a

vis

ua

lizatio

n.

from

the v

isu

aliza

tion

? F

or in

sta

nce

, pe

op

le s

ho

uld

be

ab

le to

lasso

item

s a

nd

exclu

de th

em

with

a c

lick. T

he

y s

ho

uld

be a

ble

to p

erfo

rm d

rag

-an

d-d

rop

cu

lling

. Filte

rs s

ho

uld

als

o b

e

ap

plic

ab

le, w

he

n n

ece

ssa

ry, to

a c

olle

ctio

n o

f inte

ractiv

e

vis

ua

lizatio

ns b

ein

g v

iew

ed

sim

ulta

ne

ou

sly

.

with

in th

e v

isu

aliza

tion

? T

his

inclu

de

s d

rag

-an

d-d

rop

gro

up

ing

an

d o

n-th

e-fl

y b

inn

ing

.

20

09 T

ab

lea

u S

oftw

are

Page 6: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

vis

ua

lizatio

n—

for in

sta

nce

, by c

ha

ng

ing

ag

gre

ga

tion

s o

r cre

atin

g

ne

w c

alc

ula

tion

s (e

.g., Y

ea

r ov

er Y

ea

r ch

an

ge) in

sta

ntly

?

reco

rds w

itho

ut a

ny s

pe

cia

l co

nfi

gu

ratio

n? V

isu

al a

na

lytic

s

mo

ve

s a

t the s

pe

ed

of th

ou

gh

t. It req

uire

s to

ols

to s

tay o

ut o

f

the w

ay a

nd

rea

ct in

sta

ntly

.

Vis

ua

l an

aly

tics p

aid

off d

ram

atic

ally

at

Cle

ve

lan

d C

linic

. Th

ere

, the

acco

un

ts

rece

iva

ble

sta

ff sim

ply

co

uld

n’t a

na

lyze

reje

cte

d c

laim

s fa

st e

no

ug

h. E

ach

ye

ar, th

e o

rga

niza

tion

lost

su

bsta

ntia

l am

ou

nts

from

un

fulfi

lled

insu

ran

ce re

imb

urs

em

en

ts

for s

om

e o

f the 5

0,0

00

-plu

s p

atie

nts

the h

osp

ital s

erv

es. C

laim

s

an

aly

sts

co

uld

no

t ex

am

ine

cla

ims re

co

rds ite

rativ

ely

to u

nco

ve

r

wh

y th

e c

laim

wa

s re

jecte

d. T

he

re w

as n

o w

ay

the

y c

ou

ld lo

ok a

t

a s

et o

f cla

ims a

nd

follo

w m

ultip

le lin

es o

f rea

so

nin

g to

ide

ntify

wh

at c

ou

ld b

e c

au

sin

g th

e p

rob

lem

.

Co

nve

ntio

na

l rep

ortin

g s

olu

tion

s d

idn

’t allo

w th

em

to g

rou

p c

laim

s

into

diffe

ren

t se

gm

en

ts, d

rill into

de

tails

or fi

lter in

tera

ctiv

ely

. A

vis

ua

l an

aly

tics a

pp

lica

tion

from

Ta

ble

au

ga

ve th

em

the u

se

r-

dire

cte

d e

nviro

nm

en

t the

y re

ally

ne

ed

ed

. Ru

nn

ing

on

top

of a

n

ex

istin

g d

ata

wa

reh

ou

se, T

ab

lea

u e

na

ble

d C

linic

sta

ff to a

na

lyze

un

pa

id a

nd

reje

cte

d c

laim

s a

s th

ey ro

lled

in, e

ve

n th

ou

gh

insp

ectio

n re

qu

irem

en

ts v

arie

d fro

m c

ase to

ca

se

. Re

paym

en

ts ro

se

dra

ma

tica

lly. In

on

e y

ea

r, the h

osp

ital a

ttribu

ted

ove

r $2

0 m

illion

in

reco

ve

red

paym

en

ts to

the

ir use o

f vis

ua

l an

aly

tics.

Au

gm

en

tatio

n o

f Hu

ma

n P

erc

ep

tion

Ge

nu

ine v

isu

al a

na

lytic

s a

pp

lica

tion

s e

nco

ura

ge v

isu

al th

inkin

g b

y

lev

era

gin

g th

e p

ow

ers

of h

um

an

pe

rce

ptio

n. T

he h

um

an

bra

in

po

sse

sse

s a

n a

ma

zing

ca

pa

city

to p

roce

ss g

rap

hic

s fa

ste

r tha

n it

ca

n p

roce

ss ta

ble

s o

f nu

mb

ers

. Un

fortu

na

tely

, mo

st b

usin

ess

inte

llige

nce p

acka

ge

s a

nd

sp

rea

dsh

ee

ts d

o n

ot ta

ke fu

ll ad

va

nta

ge

of th

e b

rain

’s p

erc

ep

tive c

ap

ab

ilities. F

or in

sta

nce

, the

y u

se c

olo

r

an

d v

isu

al e

ffects

irresp

on

sib

ly, a

nd

the

y ig

no

re p

rov

en

rese

arc

h

on

dis

pla

yin

g d

ata

with

ou

t bia

s.

In a

vis

ua

l an

aly

tics a

pp

lica

tion

, pro

pe

rly v

isu

alize

d d

ata

“p

op

s.”

Fo

r ex

am

ple

, rela

tion

sh

ips, tre

nd

s a

nd

ou

tliers

sh

ow

up

brig

ht a

nd

cle

ar—

aid

ing

bo

th th

e u

se

r an

d th

e a

ud

ien

ce

alik

e.

20

09 T

ab

lea

u S

oftw

are

Page 7: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Fig

ure

3: V

isu

al a

naly

tics le

ve

rag

es a

n im

po

rtant a

nd p

ow

erfu

l truth

ab

ou

t the

hu

man b

rain

: a p

ictu

re re

ally

is w

orth

mo

re th

an a

ll the w

ord

s a

nd n

um

be

rs

de

scrib

ing it. In

the to

p illu

stra

tion

, so

me d

ata

are

sh

ow

n in

a ty

pic

al re

po

rt or

sp

read

sh

eet. E

ve

n w

ith th

e u

se o

f co

lor (a

vis

ual c

ue), it’s

hard

to u

nd

ers

tan

d

mu

ch a

bo

ut s

ale

s a

nd p

rofi

t. Bu

t the im

ag

e b

elo

w, b

ase

d o

n th

e e

xact s

am

e d

ata

,

make

s th

e p

atte

rns a

nd m

ag

nitu

de

s c

lear a

t a g

lan

ce

.

Ho

w d

o y

ou

kn

ow

if yo

u a

re lo

okin

g a

t a tru

e v

isu

al a

na

lytic

s

ap

plic

atio

n? H

ere

are

so

me te

sts

:

vis

ua

l pro

pe

rties?

Pre

cis

e u

se o

f size

, co

lor, s

ha

pe a

nd

tex

t, for

ex

am

ple

, ma

ke a

diffe

ren

ce

. Wh

en

ha

nd

led

we

ll, the

y a

id

inte

rpre

tatio

n; w

he

n d

on

e p

oo

rly, th

ey d

istra

ct a

nd

mis

lea

d. T

he

ap

plic

atio

n s

ho

uld

co

nta

in e

ffectiv

e v

isu

al d

efa

ults

tha

t wo

uld

ma

ke v

isu

aliza

tion

ex

pe

rts lik

e E

dw

ard

Tu

fte a

nd

Ste

ph

en

Fe

w

pro

ud

.

vis

ua

l pe

rce

ptio

n; fo

r insta

nce th

ey e

mp

loy

cu

te, w

acky o

r

irresp

on

sib

le m

eta

ph

ors

like fra

cta

ls, 3

D o

bje

cts

, dia

ls a

nd

sp

ee

do

me

ters

. Vis

ua

l an

aly

tics s

oftw

are

he

lps a

mp

lify a

pe

rso

n’s

thin

kin

g in

the m

ost e

leg

an

t, co

mm

un

ica

tive a

nd

pro

ve

n

wa

ys.

20

09 T

ab

lea

u S

oftw

are

Page 8: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

for in

form

atio

n v

isu

aliza

tion

? A

co

mm

on

mis

take is

to e

mp

loy

free In

tern

et m

ap

pin

g s

erv

ice

s d

esig

ne

d fo

r ge

ne

ratin

g o

nlin

e

driv

ing

dire

ctio

ns. S

uch

ma

ps a

re in

effe

ctiv

e fo

r mo

st d

ata

vis

ua

lizatio

n n

ee

ds b

eca

use th

ey

vis

ua

lly o

ve

rwh

elm

wh

at y

ou

mo

st c

are

ab

ou

t (yo

ur d

ata

), ov

ere

mp

ha

size

roa

ds a

nd

lan

dm

ark

s, a

nd

are

ge

ne

rally

infl

ex

ible

.

Co

rne

ll Un

ive

rsity

fou

nd

ou

t wh

at a

diffe

ren

ce p

erc

ep

tion

ma

ke

s. T

he

y trie

d m

an

ag

ing

an

d m

on

itorin

g K

PIs

with

a

trad

ition

al b

usin

ess in

tellig

en

ce to

ol—

bu

t in n

ine m

on

ths

bu

sin

ess o

ffice

rs c

ou

ldn

’t ma

ke

se

nse o

f the d

ash

bo

ard

s o

r

rep

orts

tha

t the s

yste

m p

rod

uce

d. T

he

n th

ey d

ep

loye

d a

vis

ua

l

an

aly

tics s

uite

from

Ta

ble

au

. Th

ey e

na

ble

d C

orn

ell to

cre

ate

ea

sy-to

-un

de

rsta

nd

da

sh

bo

ard

s.

Vis

ua

l Ex

pre

ss

ive

ne

ss

No

asp

iring

pa

inte

r wo

uld

pu

t up

with

a p

ain

t-by

-nu

mb

ers

ca

nva

s.

Bu

t tha

t’s w

ha

t ma

ny p

rog

ram

s fo

rce

on

pe

op

le w

he

n th

ey u

se

ch

artin

g w

izard

s a

nd

da

sh

bo

ard

s. G

oo

d v

isu

al a

na

lytic

s to

ols

acco

mm

od

ate

pe

op

le’s

ne

ed

for d

ep

th, fl

ex

ibility

an

d

ex

pre

ssiv

en

ess in

the v

isu

al d

isp

lay

s.

Th

is is

esp

ecia

lly im

po

rtan

t wh

en

pe

op

le n

ee

d to

loo

k a

t mo

re th

an

two

or th

ree d

ime

nsio

ns o

f a p

rob

lem

sim

ulta

ne

ou

sly

. Ima

gin

e

pu

tting

fiv

e d

ime

nsio

ns o

f a p

rob

lem

(e.g

., Ye

ar, M

on

th, R

eg

ion

,

Pro

du

ct F

am

ily a

nd

Un

its S

old

) into

a c

ha

rting

wiza

rd: th

e re

su

lt just

do

esn

’t co

me o

ut w

ell. V

isu

al a

na

lytic

s a

pp

lica

tion

s le

t pe

op

le

vis

ua

lize m

ultip

le d

ime

nsio

ns o

f a p

rob

lem

effo

rtlessly

, in fo

rma

ts

tha

t are

ea

sy to

un

de

rsta

nd

. Wh

ere

cro

ss-ta

bs a

nd

piv

ot-ta

ble

s

ofte

n c

on

fuse a

nd

ov

erw

he

lm, m

ulti-d

ime

nsio

na

l vis

ua

lizatio

ns

cla

rify. V

isu

al a

na

lytic

s a

pp

lica

tion

s d

isp

lay c

om

ple

x p

rob

lem

s w

ith

ele

ga

nt s

imp

licity

.

Mu

ltidim

en

sio

na

l ex

pre

ssiv

en

ess is

pa

rticu

larly

imp

orta

nt w

he

n

time a

nd

ge

og

rap

hy a

re in

vo

lve

d. L

et’s

face it, tim

e a

nd

sp

ace

are

sp

ecia

l. Da

ta u

nd

er g

eo

gra

ph

ic a

na

lysis

ma

y n

ot h

on

or g

eo

gra

ph

ic

bo

un

da

ries, a

nd

the in

terfa

ce s

ho

uld

let th

e u

se

r follo

w a

ny lin

e o

f

ge

og

rap

hic

inq

uiry

. Do

es th

e a

pp

lica

tion

co

nta

in b

uilt-in

ge

og

rap

hic

inte

llige

nce th

at e

ffectiv

ely

dis

pla

ys m

ultip

le le

ve

ls o

f ge

og

rap

hic

de

tail?

Eq

ua

lly im

po

rtan

t is th

e tre

atm

en

t of tim

e d

ime

nsio

ns. H

an

dlin

g

time a

pp

rop

riate

ly is

n’t a

s s

imp

le a

s a

dd

ing

tren

d lin

es. P

eo

ple

sh

ou

ld b

e a

ble

to v

isu

ally

dis

pla

y d

ate

s a

nd

da

te/tim

es a

t mu

ltiple

lev

els

of d

eta

il sim

ulta

ne

ou

sly

. Fo

r ex

am

ple

, ex

am

inin

g s

ale

s tre

nd

s

by w

ee

kd

ay (e

.g. M

on

da

ys) a

nd

by

mo

nth

(e.g

., Ja

nu

ary

, Fe

bru

ary

,

Ma

rch

) sim

ulta

ne

ou

sly

ma

y b

e c

ritica

l to d

isco

ve

ring

an

up

tick in

ca

ffein

ate

d b

ev

era

ge s

ale

s o

n w

inte

r Mo

nd

ay m

orn

ing

s. M

an

y

sy

ste

ms d

o n

ot h

av

e th

e c

ap

ab

ility to

loo

k a

t time in

fle

xib

le, u

se

ful

wa

ys.

20

09 T

ab

lea

u S

oftw

are

Page 9: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Fig

ure

4 G

oo

d v

isu

al e

xp

ressiv

en

ess c

ap

ab

ilities a

cco

mm

od

ate

pe

op

le’s

ne

ed

s fo

r

de

pth

an

d fl

exib

ility in

vis

ual d

isp

lays.

He

re a

re s

om

e te

sts

of a

n a

pp

lica

tion

’s m

ultid

ime

nsio

na

l

ex

pre

ssiv

en

ess:

mu

ltiple

s,”

as e

sp

ou

se

d b

y E

dw

ard

Tu

fte? O

r do

es th

e s

oftw

are

prim

arily

pro

du

ce th

e s

am

e b

asic

gra

ph

s fo

un

d in

Exce

l’s

ch

artin

g w

izard

? M

ost b

usin

ess in

tellig

en

ce a

pp

lica

tion

s

de

libe

rate

ly c

op

ied

the c

ho

ice

s fo

un

d in

Exce

l, no

t un

de

rsta

nd

ing

the lim

itatio

ns o

f tem

pla

te-d

rive

n g

rap

hic

s.

dim

en

sio

ns o

f a p

rob

lem

sim

ulta

ne

ou

sly

? U

se

rs o

ften

ne

ed

to

vis

ua

lize m

an

y d

ime

nsio

ns o

f a p

rob

lem

at o

nce

.

as te

mp

late

s th

at c

an

be a

pp

lied

to o

the

r da

ta? C

an

the

y e

asily

sh

are

the

ir vis

ua

l cu

sto

miza

tion

s?

ge

og

rap

hic

lay

ers?

Wo

rldw

ide m

ap

pin

g a

t mu

ltiple

lev

els

of

ge

og

rap

hy s

ho

uld

be p

rov

ide

d b

y d

efa

ult a

nd

req

uire

no

sp

ecia

l

da

ta p

rep

ara

tion

or g

eo

co

din

g.

sim

ulta

ne

ou

sly

?

ov

er tim

e a

t sp

ecifi

c lo

ca

tion

s?

© 2

00

9 T

ab

lea

u S

oftw

are

Page 10: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Qu

est D

iag

no

stic

s, w

hic

h p

rovid

es m

ed

ica

l testin

g to

larg

e h

ea

lthca

re fa

cilitie

s, fa

ce

d th

e lo

ss o

f a m

ulti-

millio

n d

olla

r clie

nt. T

est re

su

lts w

ere

co

min

g in

too

slo

wly

, the c

lien

t sa

id. T

ho

ug

h th

e Q

ue

st d

irecto

r sa

w

the s

tatis

tics d

iffere

ntly

, rep

orts

an

d c

ha

rts p

rod

uce

d

co

nve

ntio

na

lly w

ere

no

t co

nvin

cin

g. A

nd

wh

en

the

clie

nt h

ad

ne

w q

ue

stio

ns, Q

ue

st c

ou

ldn

’t resp

on

d fa

st e

no

ug

h w

ith

cle

ar, m

ea

nin

gfu

l an

aly

tics.

Ta

ble

au

, a v

isu

al a

na

lytic

s a

pp

lica

tion

, ch

an

ge

d a

ll tha

t. Th

e d

irecto

r

fle

w th

rou

gh

the v

isu

alize

d d

ata

inste

ad

of tru

dg

ing

thro

ug

h ro

ws a

nd

co

lum

ns—

an

d c

rea

ted

ne

w v

iew

s o

f clie

nt d

ata

, askin

g d

iffere

nt

qu

estio

ns a

nd

ge

tting

ne

w p

ers

pe

ctiv

es o

n Q

ue

st’s

pe

rform

an

ce

.

Eve

n a

s h

e p

rese

nte

d h

is fi

nd

ing

s to

his

clie

nt, h

is c

lien

t be

ga

n a

skin

g

ad

ditio

na

l qu

estio

ns a

nd

askin

g fo

r diffe

ren

t vis

ua

lizatio

ns. T

og

eth

er,

he a

nd

his

clie

nt v

isu

alize

d th

e te

stin

g tim

elin

e b

y te

st ty

pe, p

riority

an

d fa

cility

typ

e. T

he

y d

isco

ve

red

tha

t mu

ch

of th

e p

rob

lem

wa

s

ca

use

d b

y th

e c

lien

ts’ s

taff n

ot c

on

sis

ten

tly fo

llow

ing

pro

ce

du

res

acro

ss d

iffere

nt te

sts

, diffe

ren

t prio

rity s

tatu

se

s a

nd

facility

typ

e.

Ultim

ate

ly, th

e c

lien

t no

t on

ly s

taye

d w

ith Q

ue

st b

ut a

ye

ar la

ter

bu

cke

d th

e c

orp

ora

te m

an

da

te to

sw

itch

pro

vid

ers

.

Au

tom

atic

Vis

ua

liza

tion

Ima

gin

e a

n a

pp

lica

tion

tha

t tells

yo

u h

ow

yo

u s

ho

uld

loo

k a

t the

sp

ecifi

c p

rob

lem

yo

u h

av

e. F

or to

o lo

ng

, an

aly

sts

ha

ve b

ee

n ta

ug

ht

to th

ink in

nu

mb

ers

alo

ne

. A v

isu

al a

na

lytic

s a

pp

lica

tion

jum

psta

rts

the a

na

lysis

pro

ce

ss its

elf. T

his

inclu

de

s a

uto

ma

tica

lly s

ug

ge

stin

g

effe

ctiv

e v

isu

aliza

tion

s.

Fig

ure

5: A

uto

matic

vis

ualiza

tion n

ot o

nly

he

lps p

eo

ple

fin

d p

atte

rns e

asily

bu

t

als

o tra

ins th

em

to th

ink v

isu

ally

an

d m

ore

rap

idly.

20

09 T

ab

lea

u S

oftw

are

Page 11: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

10

© 2

00

9 T

ab

lea

u S

oftw

are

A k

ey b

en

efi

t of a

uto

ma

tic v

isu

aliza

tion

is n

ot ju

st th

at it re

du

ce

s

wo

rk tim

e. It a

lso

he

lps p

eo

ple

lea

rn to

thin

k v

isu

ally

. If the

y c

an

thin

k in

pic

ture

s, th

ey c

an

wo

rk fa

ste

r an

d re

ca

ll tren

ds a

nd

pa

ttern

s

mo

re e

asily

. He

re is

the e

sse

ntia

l thin

g to

loo

k fo

r:

mo

st a

pp

rop

riate

gra

ph

ica

l de

pic

tion

ba

se

d o

n th

e s

ele

cte

d d

ata

ele

me

nts?

In o

the

r wo

rds, th

e a

uto

ma

tic v

isu

aliza

tion

sh

ou

ld n

ot

just b

e a

ran

do

m s

ele

ctio

n b

ut s

ho

uld

be

ba

se

d o

n h

eu

ristic

s.

Sa

nD

ieg

o.c

om

is a

n o

nlin

e g

uid

e to

Sa

n D

ieg

o’s

ne

igh

bo

rho

od

s, e

ve

nts

an

d a

ttractio

ns. T

he v

olu

me

s o

f

traffi

c d

ata

we

nt u

nu

se

d b

eca

use th

e C

EO

did

n’t h

ave tim

e

or e

xp

ertis

e to

an

aly

ze it. W

ith T

ab

lea

u’s

au

tom

atic

vis

ua

lizatio

n c

ap

ab

ilities—

ca

lled

“S

ho

w M

e”—

the C

EO

wa

s

ab

le to

ma

ke im

me

dia

te d

isco

ve

ries a

bo

ut th

e s

ite’s

traffi

c.

Vis

ua

l Pe

rsp

ec

tive

-Sh

ifting

Th

ere

is n

ev

er a

sin

gle

vis

ua

lizatio

n th

at o

ffers

the b

est s

um

ma

ry o

f

ev

ery

fin

din

g in

yo

ur d

ata

. Ty

pic

ally

pe

op

le n

ee

d to

loo

k a

t a v

arie

ty

of v

isu

aliza

tion

s, d

ep

en

din

g o

n th

e ta

sks y

ou

wa

nt to

ach

iev

e.

Effe

ctiv

e v

isu

al a

na

lytic

s a

pp

lica

tion

s s

ho

uld

su

gg

est a

se

ries o

f

alte

rna

tive v

isu

aliza

tion

s w

hic

h c

an

be e

ffortle

ssly

flip

pe

d th

rou

gh

.

Fo

r ex

am

ple

, if yo

u’re

tryin

g to

fin

d o

utlie

rs, lo

ok a

t a s

ca

tterp

lot.

Try

ing

to u

nd

ers

tan

d tim

e-b

ase

d tre

nd

s in

the

da

ta? T

he

n a

line

or

Ga

ntt c

ha

rt mig

ht b

e id

ea

l. Try

ing

to u

nd

ers

tan

d m

ulti-d

ime

nsio

na

l

ge

og

rap

hic

va

riatio

n? T

ry a

sm

all m

ultip

le o

f ma

ps. N

o o

ne v

iew

ca

n

an

sw

er a

ll qu

estio

ns.

His

tory

ha

s s

ho

wn

tha

t loo

kin

g a

t da

ta fro

m th

e rig

ht p

ers

pe

ctiv

e is

as im

po

rtan

t as lo

okin

g a

t the rig

ht d

ata

in th

e fi

rst p

lace

. Ph

ysic

ist

Ric

ha

rd F

ey

nm

an

fam

ou

sly

sh

ow

ed

tha

t en

gin

ee

rs w

ork

ing

on

the

Sp

ace S

hu

ttle C

ha

llen

ge

r ha

d in

fron

t of th

em

all o

f the d

ata

ne

ed

ed

to c

on

clu

de th

at la

un

ch

ing

the

Sh

uttle

at lo

w te

mp

era

ture

s p

ose

d

un

pre

ce

de

nte

d ris

ks. O

ne o

f the

ir failin

gs w

as th

at th

ey n

ev

er

plo

tted

the d

ata

in th

e m

ost re

ve

alin

g w

ay.

Th

is w

as w

ell d

ocu

me

nte

d b

y E

dw

ard

Tu

fte in

his

bo

ok V

isu

al

Ex

pla

na

tion

s: Im

ag

es a

nd

Qu

an

tities, E

vid

en

ce a

nd

Na

rrativ

e.

Sh

ifting

vis

ua

l pe

rsp

ectiv

es o

n a

pro

ble

m is

als

o a

gre

at w

ay to

ge

ne

rate

ne

w q

ue

stio

ns. It jo

stle

s th

e b

rain

a little

an

d m

ake

s y

ou

mo

re c

urio

us a

bo

ut w

ha

t is a

ctu

ally

go

ing

on

in th

e d

ata

.

Page 12: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Fig

ure

6: D

iffere

nt v

isu

aliza

tion

s a

nsw

er d

iffere

nt q

ue

stio

ns a

nd s

olv

e d

iffere

nt

issu

es. A

bove w

e s

ee th

e s

am

e d

ata

set v

isu

alize

d in

mu

ltiple

ways. E

ach

pe

rsp

ectiv

e a

nsw

ers

diffe

rent q

ue

stio

ns a

nd h

igh

lights

diffe

rent p

atte

rns. V

isu

al

an

aly

tics le

ts y

ou m

ove fro

m o

ne p

ers

pe

ctiv

e to

an

oth

er w

ith a

clic

k.

Fo

r the

se re

aso

ns, v

isu

al a

na

lytic

s a

pp

lica

tion

s c

on

tain

stro

ng

su

pp

ort fo

r pe

rsp

ectiv

e s

hiftin

g. H

ere

are

so

me

qu

alitie

s to

loo

k fo

r:

insta

ntly

? In

just a

few

clic

ks, a

pe

rso

n s

ho

uld

be

ab

le to

mo

ve

from

vie

w to

vie

w u

ntil s

he

fin

ds e

xa

ctly

the rig

ht v

iew

for th

e

qu

estio

n a

t ha

nd

.

da

ta w

ith a

clic

k?

sim

ulta

ne

ou

sly

?

Pe

op

le o

n th

e lo

oko

ut fo

r use

ful in

form

atio

n v

isu

aliza

tion

s lik

e to

fora

ge fre

ely

in d

ata

. So

the la

st th

ing

pe

op

le n

ee

d is

a to

ol th

at

co

nfi

ne

s th

em

to a

sin

gle

, line

ar p

ath

. An

infl

ex

ible

too

l cre

ate

s a

da

tase

t an

d a

ch

art a

nd

tries to

stic

k w

ith it. A

vis

ua

l an

aly

tics

ap

plic

atio

n in

ste

ad

offe

rs d

irect a

cce

ss to

a m

yria

d o

f vis

ua

lizatio

ns,

with

no

bo

un

da

ries.

An

aly

sts

at T

he M

artin

Ag

en

cy

’s In

ge

nu

ity M

ed

ia G

rou

p

sifte

d th

rou

gh

mo

un

tain

s o

f da

ta fro

m in

tera

ctiv

e m

ed

ia

ca

mp

aig

ns to

pro

du

ce p

erio

dic

rep

orts

an

d d

ash

bo

ard

s

for c

lien

ts. P

rod

uctio

n to

ok h

ou

rs. A

na

lysts

on

ly h

ad

time

to c

rea

te w

ere

sta

tic, s

tan

da

rd c

ha

rts, w

ith n

o tim

e to

fin

d n

ew

insig

hts

. Bu

t with

Ta

ble

au

’s p

ers

pe

ctiv

e s

hiftin

g fe

atu

res, a

na

lysts

co

uld

loo

k a

t mu

ltiple

pe

rsp

ectiv

es o

f the

ir clie

nts

’ da

ta a

nd

rap

idly

fin

d th

e m

ost m

ea

nin

gfu

l on

e. It o

pe

ne

d a

ne

w w

orld

of e

xp

lora

tion

an

d d

isco

ve

ry.

11

© 2

00

9 T

ab

lea

u S

oftw

are

Page 13: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Vis

ua

l Pe

rsp

ec

tive

Lin

kin

g

A lo

gic

al b

ut p

ow

erfu

l ad

ditio

n to

pe

rsp

ectiv

e s

hiftin

g is

pe

rsp

ectiv

e

linkin

g. A

ltho

ug

h th

e tw

o to

pic

s a

re re

late

d, lin

kin

g e

nta

ils a

diffe

ren

t se

t of c

ap

ab

ilities th

an

pe

rsp

ectiv

e s

hiftin

g. In

sh

ort, it is

n’t

en

ou

gh

to lo

ok a

t mu

ltiple

pe

rsp

ectiv

es o

n a

pro

ble

m in

rap

id

su

cce

ssio

n –

or e

ve

n s

imu

ltan

eo

usly

. So

me

time

s th

e p

ers

pe

ctiv

es

ne

ed

to b

e in

tima

tely

linke

d. O

ne v

isu

aliza

tion

ma

y d

isp

lay a

se

t of

ou

tliers

, for in

sta

nce

. Ca

n a

pe

rso

n s

ele

ct a

n o

utlie

r an

d in

sta

ntly

se

e a

no

the

r vis

ua

lizatio

n th

at d

isp

lay

s g

rea

ter d

eta

il? A

s a

n

ex

am

ple

, a p

ers

on

ma

y n

otic

e th

at s

ale

s fo

r a p

artic

ula

r sta

te s

ee

ms

to b

e d

om

ina

ting

. By c

lickin

g o

n th

e m

ark

rep

rese

ntin

g th

at s

tate

’s

sa

les, h

e c

an

insta

ntly

up

da

te a

vis

ua

lizatio

n re

ga

rdin

g s

ale

s

am

ou

nt b

y c

om

pa

ny a

nd

se

e a

no

the

r vis

ua

lizatio

n re

dra

w a

line

ch

art s

ho

win

g s

ale

s b

y d

ate

usin

g d

ata

for ju

st th

at s

tate

. Th

at

inte

ractio

n m

ay s

he

d lig

ht o

n w

ha

t is d

rivin

g s

ale

s in

tha

t sta

te to

be

do

min

an

t.

Fig

ure

7 In

a v

isu

al a

naly

tics a

pp

licatio

n, y

ou c

an u

se p

ers

pe

ctiv

e lin

kin

g to

dis

cove

r rela

tion

sh

ips in

data

or to

un

cove

r hid

de

n s

torie

s. In

this

exam

ple

, a u

se

r

se

lects

a s

tate

on th

e m

ap (C

alifo

rnia

). Tab

leau th

en in

sta

ntly

reve

als

wh

ich d

ata

in th

e o

the

r vis

ualiza

tion

s a

re re

late

d.

12

© 2

00

9 T

ab

lea

u S

oftw

are

Page 14: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Ge

nu

ine v

isu

al a

na

lytic

s a

pp

lica

tion

s s

up

po

rt pe

rsp

ectiv

e lin

kin

g in

se

ve

ral w

ay

s. H

ere

are

so

me fe

atu

res to

eva

lua

te:

vis

ua

lizatio

ns?

Fo

r insta

nce

, ca

n u

se

rs s

ele

ct a

tren

d lin

e in

a

firs

t vie

w a

nd

se

e th

e g

eo

gra

ph

ic e

ntitie

s re

late

d to

tha

t line in

a

se

co

nd

vie

w?

ca

n p

eo

ple

use a

firs

t vis

ua

lizatio

n to

se

arc

h fo

r da

ta o

f inte

rest,

an

d th

en

au

tom

atic

ally

ap

ply

the s

ele

cte

d d

ata

as a

filte

r on

a

se

co

nd

vis

ua

lizatio

n?

ho

ve

ring

ov

er d

ata

?

pro

gra

mm

ing

? In

a v

isu

al a

na

lytic

s a

pp

lica

tion

, rich

da

ta

asso

cia

tion

s s

ho

uld

be d

isco

ve

red

au

tom

atic

ally

.

Co

llab

ora

tive

Vis

ua

liza

tion

An

oth

er d

efi

nin

g c

ap

ab

ility o

f effe

ctiv

e v

isu

al a

na

lytic

s a

pp

lica

tion

s

is th

e a

bility

to ite

rativ

ely

cre

ate

use

ful in

form

atio

n v

isu

aliza

tion

s in

a te

am

se

tting

. Th

is p

roce

ss u

su

ally

sta

rts w

ith a

“h

ey, lo

ok a

t this

mo

me

nt. B

ut th

e re

al q

ue

stio

n is

: Do

es th

e s

oftw

are

su

pp

ort th

e

invo

lve

d c

olla

bo

rativ

e p

roce

ss th

at s

ho

uld

follo

w th

ese m

om

en

ts?

Sh

are

d fi

nd

ing

s le

ad

to s

olu

tion

s, a

ctio

n, a

nd

resu

lts. In

fact, in

mo

st o

rga

niza

tion

s u

nsh

are

d d

isco

ve

ries a

re u

se

less. S

om

e

so

ftwa

re p

acka

ge

s, w

hile

me

etin

g o

the

r crite

ria, fa

il he

re. E

ffectiv

e

vis

ua

l an

aly

tics s

oftw

are

en

co

ura

ge

s c

olla

bo

ratio

n b

y le

tting

resu

lts

be s

ha

red

in w

ha

tev

er fo

rm th

e u

se

r pre

fers

. Th

e a

pp

lica

tion

’s

arc

hite

ctu

re s

ho

uld

be b

uilt e

xp

licitly

for c

olla

bo

ratio

n.

He

re a

re s

om

e o

f so

me s

pe

cifi

c fe

atu

res to

ev

alu

ate

:

se

co

nd

s? V

isu

aliza

tion

s s

ho

uld

be a

va

ilab

le v

ia a

bro

wse

r. Th

ey

sh

ou

ld b

e liv

e a

nd

ch

an

ge a

s d

ata

ch

an

ge

s.

role

or g

rou

p o

f the p

artic

ula

r info

rma

tion

co

nsu

me

rs? In

ad

ditio

n, w

ill pe

op

le w

ho

are

vie

win

g o

r inte

ractin

g w

ith th

e

vis

ua

lizatio

ns h

av

e th

e a

bility

to c

usto

mize

the

ir pa

rticu

lar

ch

an

ge

s to

it an

d s

av

e th

at v

iew

?

with

it all?

Th

e a

pp

lica

tion

sh

ou

ld s

up

po

rt Go

og

le-s

tyle

se

arc

h

tha

t en

ab

le g

rou

ps to

loca

te v

isu

aliza

tion

s q

uic

kly

.

cu

sto

mizin

g a

nd

sh

arin

g v

isu

aliza

tion

s a

nd

da

sh

bo

ard

s?

13

© 2

00

9 T

ab

lea

u S

oftw

are

Page 15: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

an

d d

ata

mo

de

ls? W

ho

le d

ata

mo

de

ls (o

r ind

ivid

ua

l ca

lcu

latio

ns,

de

fin

ed

gro

up

s, e

tc.) s

ho

uld

be

ea

sily

sh

are

d a

nd

ve

rsio

n

co

ntro

lled

.

into

oth

er a

pp

lica

tion

s a

nd

po

rtals?

Fig

ure

8: In

this

illustra

tion

, pe

op

le s

hare

key fi

nd

ing

s fro

m th

e v

isu

al a

naly

tics

syste

m w

ith c

olle

ag

ue

s u

sin

g S

hare

po

int. E

ve

n th

ou

gh th

e re

su

lts a

re e

mb

ed

de

d,

pe

op

le c

an s

lice a

nd d

ice th

e d

ata

usin

g c

he

ck-b

ox fi

lters

, so

rt, vie

w u

nd

erly

ing

data

, zoo

m in

to s

pe

cifi

c p

oin

ts a

nd e

ve

n v

iew

the u

nd

erly

ing fi

nd

ing

s.

14

© 2

00

9 T

ab

lea

u S

oftw

are

Page 16: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

De

plo

ym

en

t Co

ns

ide

ratio

ns

Th

is p

ap

er h

as d

escrib

ed

the s

ev

en

de

fin

ing

ele

me

nts

of v

isu

al

an

aly

tics a

pp

lica

tion

s. Y

et th

ere

are

oth

er fa

cto

rs to

co

nsid

er b

efo

re

ma

kin

g a

ch

oic

e o

n a

vis

ua

l an

aly

tics s

tan

da

rd. F

or in

sta

nce

, if the

vis

ua

l an

aly

tics a

pp

lica

tion

is g

oin

g to

be u

se

d in

a c

orp

ora

te

en

viro

nm

en

t, the

re a

re te

ch

nic

al a

nd

infra

stru

ctu

re c

on

sid

era

tion

s

as w

ell.

Un

ive

rsa

l Da

ta A

cce

ss. D

ata

co

me

s fro

m a

ll dire

ctio

ns. D

oe

s th

e

so

ftwa

re c

on

ne

ct to

virtu

ally

an

y s

ou

rce

, from

da

ta w

are

ho

use

s to

Exce

l or te

xt fi

les?

Do

es it c

on

ne

ct to

all m

ajo

r da

ta fo

rma

ts,

inclu

din

g re

latio

na

l da

tab

ase

s, O

LA

P d

ata

cu

be

s a

nd

fla

t file

s?

Fu

rthe

rmo

re, is

co

nn

ectin

g to

ne

w d

ata

so

urc

es e

asily

do

ne?

Sca

lab

ility. D

oe

s th

e s

oftw

are

su

pp

ort re

al-tim

e in

tera

ctiv

e

vis

ua

lizatio

n o

f da

ta o

f ne

arly

an

y s

ize—

ev

en

millio

ns o

r billio

ns o

f

reco

rds?

Th

e a

pp

lica

tion

sh

ou

ld b

e a

ble

to h

an

dle

larg

e a

mo

un

ts o

f

da

ta a

nd

pro

vid

e s

olid

pe

rform

an

ce

.

Ge

ne

rate

s E

fficie

nt S

QL fo

r DB

MS

. Re

po

rts s

ho

uld

ge

ne

rate

qu

ickly

.

Pe

op

le s

ho

uld

ha

ve th

e o

ptio

n to

take

, mo

dify

an

d ru

n S

QL fro

m th

e

rep

orts

an

d re

pro

du

ce th

e re

su

lts. T

he

y s

ho

uld

als

o b

e a

ble

co

nn

ect

to S

QL s

tate

me

nts

rath

er th

an

actu

al d

ata

tab

les a

nd

vie

ws.

Ab

ility to

Jo

in T

ab

les. P

eo

ple

sh

ou

ld b

e a

ble

to jo

in ta

ble

s e

asily

in

wa

ys th

at g

uid

e th

em

to a

pp

rop

riate

an

d w

ell-d

esig

ne

d jo

ins. T

ab

les

ca

n b

e jo

ine

d a

uto

ma

tica

lly o

r ba

se

d o

n u

se

r’s d

irectio

n.

Se

cu

rity. D

oe

s th

e a

pp

lica

tion

ha

ve a

full s

ecu

rity m

od

ule

to s

up

po

rt

co

llab

ora

tion

with

in e

xis

ting

pe

rmis

sio

n s

yste

ms?

Th

is s

ho

uld

inclu

de o

ptio

na

l inte

gra

tion

with

oth

er s

ecu

rity m

eth

od

s lik

e A

ctiv

e

Dire

cto

ry.

Min

ima

l IT S

up

po

rt. Go

od

so

ftwa

re fre

es IT

from

the s

ma

ll stu

ff.

Go

od

so

ftwa

re in

sta

lls e

asily

, ge

ts th

e a

ve

rag

e u

se

r up

in m

inu

tes

with

ou

t he

lp, a

nd

pro

vid

es fre

e tra

inin

g o

n d

em

an

d.

Mic

roso

ft Offi

ce C

om

pa

tibility

. MS

Offi

ce is

ub

iqu

itou

s o

n th

e

bu

sin

ess p

ers

on

’s d

eskto

p. D

oe

s th

e s

oftw

are

pro

vid

e g

ate

wa

ys in

an

d o

ut o

f MS

Offi

ce? C

an

use

rs n

ativ

ely

acce

ss d

ata

from

Offi

ce

ap

plic

atio

ns?

Ca

n th

ey o

utp

ut im

ag

es, ta

ble

s, d

ata

lists

, cro

ss-ta

bs

ea

sily

an

d d

irectly

to O

ffice?

Da

ta M

od

elin

g a

nd

Ma

na

ge

me

nt. C

an

use

rs s

ha

re e

ntire

da

ta

mo

de

ls u

se

d to

co

nstru

ct v

isu

aliza

tion

s—

alo

ng

with

ind

ivid

ua

l

15

© 2

00

9 T

ab

lea

u S

oftw

are

Page 17: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

ca

lcu

latio

ns, g

rou

ps, e

tc? U

se

rs s

ho

uld

be a

ble

to re

ly o

n

dim

en

sio

ns a

nd

me

asu

res th

at a

re c

on

sis

ten

tly d

efi

ne

d a

nd

ha

ve

da

ta in

teg

rity. C

an

the s

oftw

are

allo

w th

e u

se

r to m

od

el d

ata

–- s

uch

as to

mo

dify

va

riab

le ty

pe

s, re

plic

ate

va

riab

les, c

ha

ng

e fi

eld

na

me

s,

sta

nd

ard

ize d

ime

nsio

ns -- w

itho

ut d

ev

elo

pe

r su

pp

ort?

Th

is is

critic

al

for g

ivin

g u

se

rs th

e a

na

lytic

al p

ow

er w

hile

als

o re

du

cin

g th

e IT

bu

rde

n.

Fig

ure

9 T

ab

leau D

eskto

p a

nd T

ab

leau S

erv

er c

an b

oth

be d

ep

loye

d in

min

ute

s.

Tab

leau’s

softw

are

arc

hite

ctu

re is

bu

ilt to m

ake c

olla

bo

rativ

e a

naly

tics e

asy to

de

plo

y a

nd s

up

po

rt. Au

tho

rs p

ub

lish liv

e, in

tera

ctiv

e d

ata

vis

ualiza

tion

s, re

po

rts

an

d d

ash

bo

ard

s to

Tab

leau S

erv

er w

he

re p

eo

ple

use w

eb b

row

se

rs a

nd w

eb

ap

plic

atio

n.

Tra

inin

g. H

ow

mu

ch

is n

ee

de

d? Id

ea

lly p

eo

ple

ca

n g

et s

tarte

d

with

ou

t an

y tra

inin

g –

the u

se

r inte

rface s

ho

uld

be o

bv

iou

s a

nd

ea

sy

to u

se

. Is th

ere

free o

nlin

e tra

inin

g a

va

ilab

le a

ny tim

e fo

r mu

ltiple

lev

els?

As p

eo

ple

ad

va

nce in

the

ir vis

ua

l an

aly

tic s

kills

, mo

re

so

ph

istic

ate

d tra

inin

g s

ho

uld

be a

va

ilab

le o

n th

eir tim

e s

ch

ed

ule

ba

se

d o

n th

eir n

ee

ds.

Lic

en

sin

g. L

ice

nsin

g m

od

els

sh

ou

ld b

e fl

ex

ible

ba

se

d o

n u

se

r ne

ed

s

with

ou

t min

imu

m c

on

fig

ura

tion

s: b

uy

1 lic

en

se

, bu

y 1

0, b

uy 1

00,

bu

y 1

00

0.

16

© 2

00

9 T

ab

lea

u S

oftw

are

Page 18: Selecting a - KB Earlekbearle.com/Data/Sites/1/SharedFiles/selecting... · existing data warehouse, Tableau enabled Clinic staff to analyze unpaid and rejected claims as they rolled

Fu

rth

er R

ea

din

g o

n V

isu

al A

na

lytic

s1. J

.J. T

ho

ma

s a

nd

K.A

. Co

ok, Illu

min

atin

g th

e P

ath

, IEE

E C

S P

ress,

20

05

.

2. S

tua

rt Ca

rd, J

ock M

ackin

lay, B

en

Sch

ne

ide

rma

n, R

ea

din

gs in

Info

rma

tion

Vis

ua

lizatio

n: U

sin

g V

isio

n to

Th

ink, M

org

an

Ka

ufm

an

n, 1

99

9.

3. T

ufte

, E. R

., Th

e V

isu

al D

isp

lay o

f Qu

an

titativ

e In

form

atio

n,

Gra

ph

ics P

ress, 2

001.

4. S

tep

he

n F

ew

, Pe

rce

ptu

al E

dg

e, Im

pro

ve

Yo

ur V

isio

n a

nd

Ex

pa

nd

Yo

ur M

ind

with

Vis

ua

l An

aly

tics, w

hite

pa

pe

r.

5. S

tep

he

n F

ew

, Sh

ow

Me th

e N

um

be

rs: D

esig

nin

g T

ab

les a

nd

Gra

ph

s to

En

ligh

ten

, An

aly

tics P

ress, 2

00

4.

6. J

acq

ue

s B

ertin

, Th

e S

em

iolo

gy o

f Gra

ph

ics, O

ut o

f prin

t, 19

67.

7. D

on

ald

A. N

orm

an

, Th

e P

sy

ch

olo

gy o

f Ev

ery

da

y T

hin

gs, B

asic

Mo

rga

n K

au

fma

nn

Pu

blis

he

rs, 2

00

0.

An

aly

sis

, an

d V

isu

aliza

tion

of M

ultid

ime

nsio

na

l Da

tab

ase

s, C

AC

M,

11. C

ha

bo

t, C., D

em

ystify

ing

Vis

ua

l An

aly

tics, IE

EE

/VA

ST

ba

se

d o

n

“P

ractic

al A

pp

lica

tion

s o

f Vis

ua

l An

aly

tics: O

n th

e C

usp

of

Vis

ua

l An

aly

tics S

cie

nce a

nd

Te

ch

no

log

y.

12

. Jo

ck D

. Ma

ckin

lay, P

at H

an

rah

an

, an

d C

hris

Sto

lte, S

ho

w M

e:

Au

tom

atic

Pre

se

nta

tion

for V

isu

al A

na

lysis

, wh

itep

ap

er.

13

. Pa

t Ha

nra

ha

n, C

hris

Sto

lte a

nd

Jo

ck M

ackin

lay, V

isu

al A

na

lysis

for E

ve

ryo

ne - U

nd

ers

tan

din

g D

ata

Ex

plo

ratio

n a

nd

Vis

ua

lizatio

n, w

hite

pa

pe

r.

17

“T

ab

lea

u” a

nd

“T

ab

lea

u S

oftw

are

,” a

re re

gis

tere

d tra

de

ma

rks o

f Ta

ble

au

So

ftwa

re.

Re

fere

nce

s to

oth

er c

om

pa

nie

s a

nd

the

ir pro

du

cts

use tra

de

ma

rks o

wn

ed

by th

e re

sp

ectiv

e

co

mp

an

ies a

nd

are

for re

fere

nce p

urp

ose

s o

nly

.C

op

yrig

ht ©

20

09 T

ab

lea

u S

oftw

are

, Inco

rpo

rate

d a

nd

its lic

en

so

rs. A

ll righ

ts re

se

rve

d.