Big Data Analytics: Turning Big Data into Big Money

By Frank Ohlhor st
Copyright 2013 by John Wiley & Sons, Inc.

CHAPTER

Building the Big

Data Team

n e of th e m ost im portan t elem en ts of a Big Data project is a

rath er obviou s bu t often overlooked item : people. With ou t
h u m an in volvem en t or in terpretation , Big Data an alytics
becom es u seless, h avin g n o pu rpose an d n o valu e. It takes a team to
m ake Big Data work, an d even if th at team con sists of on ly two
in dividu als, it is still a n ecessary elem en t.
Brin gin g people togeth er to bu ild a team can be an ardu ou s process
th at in volves m u ltiple m eetin gs, perh aps recru itm en t, an d, of cou rse,
person n el m an agem en t. Several specialized skills in Big Data are
requ ired, an d th at is wh at de n es th e team . Determ in in g th ose skills is
on e of th e rst steps in pu ttin g a team togeth er.

THE DA TA SCIEN TIST

On e of th e rst con cepts to becom e acqu ain ted with is th e data sci-
en tist; a relatively n ew title, it is n ot readily recogn ized or accepted by
m an y organ ization s, bu t it is h ere to stay.
A data scien tist is n orm ally associated with an em ployee or a
bu sin ess in telligen ce (BI) con su ltan t wh o excels at an alyzin g data,
particu larly large am ou n ts of data, to h elp a bu sin ess gain a com peti-
tive edge. Th e data scien tist is u su ally th e de facto team leader du rin g a
Big Data an alytics project.

29

c04 22 October 2012; 17:56:34

30 BI G DATA ANAL YTI CS

Th e title data scientist is som etim es disparaged becau se it lacks

speci city an d can be perceived as an aggran dized syn on ym for data
analyst. Neverth eless, th e position is gain in g acceptan ce with large
en terprises th at are in terested in derivin g m ean in g from Big Data, th e
volu m in ou s am ou n t of stru ctu red, u n stru ctu red, an d sem istru ctu red
data th at a large en terprise produ ces or h as access to.
A data scien tist m u st possess a com bin ation of an alytic, m ach in e
learn in g, data m in in g, an d statistical skills as well as experien ce with
algorith m s an d codin g. However, th e m ost critical skill a data scien tist
sh ou ld possess is th e ability to tran slate th e sign i can ce of data in a
way th at can be easily u n derstood by oth ers.

THE TEA M CHA LLEN GE

Fin din g an d h irin g talen ted workers with an alytics skills is th e rst step
in creatin g an effective data an alytics team . Organ izin g th at team is th e
n ext step; th e relation sh ip between IT an d BI grou ps m u st be in cor-
porated in to th e team design , leadin g to a determ in ation of h ow m u ch
au ton om y to give to Big Data an alytics profession als.
En terprises with h igh ly organ ized an d cen tralized corporate
stru ctu res will lean toward placin g an an alytics team u n der an IT
departm en t or a bu sin ess in telligen ce com peten cy cen ter. However,
m an y experts h ave fou n d th at su ccessfu l Big Data an alytics projects
seem to work better u sin g a less cen tralized approach , givin g team
m em bers th e freedom to in terpret resu lts an d de n e n ew ways of
lookin g at data.
For m axim u m effectiven ess, Big Data an alytics team s can be
organ ized by bu sin ess fu n ction or placed directly with in a speci c
bu sin ess u n it. An exam ple of th is wou ld be placin g an an alytics team
th at focu ses on cu stom er ch u rn (th e tu rn over of cu stom er accou n ts)
an d oth er m arketin g-related an alysis in a m arketin g departm en t,
wh ile a risk-focu sed data an alytics project team wou ld be better su ited
to a n an ce departm en t.
Ideally, placin g th e Big Data an alytics team in to a departm en t
wh ere th e resu ltin g data h ave im m ediate valu e is th e best way to
accelerate n din gs, determ in e valu e, an d deliver resu lts in an action -
able fash ion . Th at way th e an alyst an d th e departm en tal decision

c04 22 October 2012; 17:56:34

 BUI L DI NG THE BI G DATA TEAM 31

m akers are speakin g th e sam e lan gu age an d workin g in a collaborative

fash ion to eke ou t th e best resu lts.
It all depen ds on scale. A sm all bu sin ess m ay h ave differen t an a-
lytical n eeds th an a large bu sin ess does, an d th at obviou sly affects th e
relation sh ip with th e data an alysis profession als an d th e departm en ts
th ey work with .

DIFFEREN T TEA MS, DIFFEREN T GO A LS

A case in poin t wou ld be an en gin eerin g rm th at is exam in in g large

volu m es of u n stru ctu red data for a tech n ical an alysis. Th e rm itself
m ay be qu ite sm all, bu t th e data set m ay be qu ite large. For exam ple, if
an en gin eerin g rm was design in g a bridge, th e com pon en ts of Big
Data an alytics cou ld in volve everyth in g from cen su s data to traf c
pattern s to weath er factors, wh ich cou ld be u sed to u n cover load an d
traf c tren ds th at wou ld affect th e design of th e bridge. If oth er ele-
m en ts are added, su ch as m arket data (m aterials costs an d an ticipated
n an cial growth for th e area), th e de n ition of a data scien tist m ay
ch an ge. Th at in dividu al m ay n eed an en gin eerin g backgrou n d an d a
keen u n derstan din g of econ om ics an d m ay work on ly with th e pri-
m ary en gin eers on th e project an d n ot with an y oth er com pan y
departm en ts.
Th is can m ean that the rm s m arketin g and sales departments
are left out in the cold. The question then is h ow important is th at style
of analytics to those departments arguably, it is n ot important at all.
In a situation like th at, m arket analysis, com petition , governm ent
fun ding, infrastru cture age and u sage, and population density m ay n ot be
as applicable to the in-place data scien tist but m ay require a differen t
individu al skill set to successfully interpret the results.
As an alytics n eeds an d organ ization al size in crease, roles m ay
ch an ge, as well as th e processes an d th e relation sh ips in volved. Larger
organ ization s ten d to h ave th e resou rces an d bu dgets to better leverage
th eir data. In th ose cases, it becom es im portan t to recogn ize th e pri-
m ary skills n eeded by a Big Data an alytics team an d to bu ild th e team
arou n d core com peten cies. Fortu n ately, it is relatively easy to iden tify
th ose core com peten cies, becau se th e tasks of th e team can be broken
down in to th ree capabilities.

c04 22 October 2012; 17:56:34

32 BI G DATA ANAL YTI CS

DO N T FO RGET THE DA TA

There are three primary capabilities n eeded in a data analytics team: (1)
locating the data, (2) n ormalizing th e data, and (3) analyzing th e data.
For th e rst capability, locatin g th e data, an in dividu al h as to be
able to n d relevan t data from in tern al an d extern al sou rces an d work
with th e IT departm en t s data govern an ce team to secu re access to th e
data. Th at in dividu al m ay also n eed to work with extern al bu sin esses,
govern m en t agen cies, an d research rm s to gain access to large data
sets, as well as u n derstan d th e differen ce between stru ctu red an d
u n stru ctu red data.
For th e secon d capability, n orm alizin g th e data, an in dividu al h as
to prepare th e raw data before th ey are an alyzed to rem ove an y
spu riou s data. Th is process requ ires tech n ical skills as well as an alytics
skills. Th e in dividu al m ay also n eed to kn ow h ow to com bin e th e data
sets, load th ose data sets on th e storage platform , an d bu ild a m atrix of
elds to n orm alize th e con ten ts.
Th e th ird capability, an alyzin g th e data, is perh aps th e team s m ost
im portan t ch ore. For m ost organ ization s, th e an alytic process is con -
du cted by th e data scien tist, wh o accesses th e data, design s algorith m s,
gath ers th e resu lts, an d th en presen ts th e in form ation .
Th ese th ree prim ary ch ores de n e a data an alytics team s fu n c-
tion s. However, th ere are several su bsets of tasks th at fall u n der each
category, an d th ese tasks can vary based on scope an d oth er elem en ts
speci c to th e requ ired data an alytics process.
Mu ch like th e data th em selves, th e team sh ou ld n ot be static in
n atu re an d sh ou ld be able to evolve an d adapt to th e n eeds of th e
bu sin ess.

CHA LLEN GES REMA IN

Locatin g the right talen t to analyze data is the biggest h urdle in buildin g
a team. Such talen t is in h igh dem and, and th e n eed for data analysts
and data scien tists contin ues to grow at an alm ost expon en tial rate.
Fin din g th is talen t m ean s th at organ ization s will h ave to focu s on
data scien ce an d h ire statistical m odelers an d text data m in in g pro-
fession als as well as people wh o specialize in sen tim en t an alysis.

c04 22 October 2012; 17:56:34

 BUI L DI NG THE BI G DATA TEAM 33

Su ccess with Big Data an alytics requ ires solid data m odels, statistical
predictive m odels, an d test an alytic m odels, sin ce th ese will be th e core
application s n eeded to do Big Data.
Locatin g the appropriate talen t takes m ore th an ju st a typical IT job
placement; th e skills required for a good retu rn on investm en t are n ot
sim ple and are n ot solely tech nology oriented. Som e organizations m ay
turn to consulting rms to m eet the n eed for talen t; h owever, m an y
con sultin g rms also h ave trouble n ding th e experts th at can m ake Big
Data pay off.
Neverth eless, th ere is a silver lin in g to th e Big Data storm clou d.
Big Data is abou t bu sin ess as m u ch as it is abou t tech n ology, wh ich
m ean s th at it requ ires a h ybrid talen t. Th is allows th e pool of poten tial
experts to be m u ch deeper th an ju st th e IT profession al workforce. In
fact, a Big Data expert cou ld be developed from oth er departm en ts th at
are n ot IT cen tered bu t th at do h ave a sign i can t n eed for research ,
an alysis, an d in terpretation of facts.
Th e poten tial talen t pool m ay grow to in clu de an y staffers wh o
h ave an in h eren t in terest in th e Big Data tech n ology platform s in play,
wh o h ave a tools backgrou n d from web site developm en t work earlier
in th eir careers, or wh o are ju st n atu rally cu riou s, talen ted, an d self-
tau gh t in a qu est to be better at th eir jobs. Th ese are typically in divi-
du als wh o can u n derstan d th e valu e of data an d th e ideology of h ow to
in terpret th e data.
Bu t organ ization s sh ou ld n ot h ire ju st an yon e wh o sh ows a spark
of in terest in or a basic u n derstan din g of data an alytics. It is im portan t
to develop a litm u s test of sorts to determ in e if an in dividu al h as th e
appropriate skills to su cceed in wh at m ay be a n ew career. Th e can -
didates sh ou ld possess a fou n dation of ve critical skills to im m ediately
brin g valu e to a Big Data team :

1. Data m in in g
2. Data visu alization
3. Data an alysis
4. Data m an ipu lation
5. Data discovery

Th ese de n e wh at a data scien tist sh ou ld be able to accom plish .

c04 22 October 2012; 17:56:34

34 BI G DATA ANAL YTI CS

TEA MS VERSUS CULTURE

Argu ably, n din g an d h irin g talen ted workers with an alytics skills is
th e rst step in establish in g an advan ced data an alytics team . If th at
is in deed th e case, th en th e secon d step wou ld be determ in in g h ow to
stru ctu re th e team in relation to existin g IT an d BI grou ps, as well as
determ in in g h ow m u ch au ton om y to give th e an alytics profession als.
Th at process m ay requ ire bu ildin g a n ew cu ltu re of tech n ology
profession als wh o also h ave sign i can t bu sin ess skills. Developin g th at
cu ltu re depen ds on m an y factors, su ch as m akin g su re th at th e team s
are edu cated in th e ways of th e bu sin ess cu ltu re in place an d
em ph asizin g m easu rem en ts an d resu lts.
Startin g at th e top proves to be on e of th e best ways to tran sform
an IT-cen tered cu ltu re in to an in tern al bu sin ess cu ltu re th at th rives on
advan ced data an alytics tech n ology an d fact-based decision m akin g.
Bu sin esses th at h ave experien ced a ch an ge in sen ior m an agem en t
often clear th e path for th e developm en t of a data an alytics bu sin ess
cu ltu re an d a data wareh ou sin g, BI, an d advan ced an alytics program .
In stitu tin g a ch an ge in cu ltu ral ideology is on e of th e m ost
im portan t ch ores associated with leveragin g an alytics. Man y com pa-
n ies h ave becom e accu stom ed to ru n n in g operation s based on gu t
feelin gs an d wh at h as worked in th e past, both of wh ich lead to a
form u laic way of con du ctin g bu sin ess.
Nowh ere h as th is been m ore eviden t th an in m ajor retail ch ain s,
wh ich u su ally pride th em selves on con sisten cy across location s. Th at
cu ltu ral perspective can prove to be th e an tith esis of a dyn am ic,
com petitive bu sin ess. In stitu tin g a cu ltu re th at u ses th e ideology of
an alytics can tran sform bu sin ess operation s. For exam ple, th e bu sin ess
can better serve m arkets by u sin g data m in in g an d predictive an alytics
tools to au tom atically set plan s for placin g in ven tory in to in dividu al
retail location s. Th e key is pu ttin g th e n eeded produ cts in fron t of
poten tial cu stom ers, su ch as by kn owin g th at sn ow sh ovels will n ot sell
in Florida an d th at su n tan lotion sells poorly in Alaska.
An oth er poten tial way to foster an an alytics bu sin ess cu ltu re
with in an organ ization is to set u p a dedicated data an alytics grou p. An
an alytics grou p with its own director cou ld develop an an alytics
strategy an d project plan , prom ote th e u se of an alytics with in th e

c04 22 October 2012; 17:56:34

 BUI L DI NG THE BI G DATA TEAM 35

com pan y, train data an alysts on an alytics tools an d con cepts, an d work
with th e IT, BI, an d data wareh ou sin g team s on deploym en t projects.

GA UGIN G SUCCESS

Su ccess h as to be m easu red, an d m easu rin g a team s con tribu tion to

th e bottom lin e can be a dif cu lt process. Th at is wh y it is im portan t
to bu ild objectives, m easu rem en ts, an d m ileston es th at dem on strate
th e ben e ts of a team focu sed on Big Data an alytics. Developin g
perform an ce m easu rem en ts is an im portan t part of design in g a bu si-
n ess plan . With Big Data, th ose m etrics can be assign ed to th e speci c
goal in m in d.
For exam ple, if an organ ization is lookin g to brin g ef cien cy to a
wareh ou se, a perform an ce m etric m ay be m easu rin g th e am ou n t of
em pty sh elf space an d wh at th e cost of th at em pty sh elf space m ean s to
th e com pan y. An alytics can be u sed to iden tify produ ct m ovem en t,
sales prediction s, an d so forth to m ove produ ct in to th at sh elf space to
better service th e n eeds of cu stom ers. It is a sim ple com parison of th e
percen tage of space u sed before th e an alytics process an d th e per-
cen tage of space u sed after th e an alytics team h as tackled th e issu e.

c04 22 October 2012; 17:56:34