You are on page 1of 56


F OU R T H ANNUAL • AU GU S T 2 0 1 3

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 1
!"#$% %'() *+)+,-.'

Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'"/&4 sLudy bullds on prevlous Comp1lA research ln Lhe
cloud compuLlng space, furLher explorlng Lrends, challenges, and opporLunlLles. 1he ob[ecLlves of Lhls
sLudy lnclude:

- CollecL lnformaLlon on end-user purchase drlvers and/or lnhlblLors
- Assess depLh of cloud usage and ways ln whlch lL ls changlng l1 behavlor
- 1rack Lhe experlence of channel parLners offerlng cloud soluLlons, bulldlng cloud pracLlces
- Cauge channel percepLlons of end-user purchase drlvers and or lnhlblLors
- Measure lmpacL of cloud compuLlng on channel roles/vendor relaLlonshlps
- Assess Lhe need for Lralnlng and/or sLandards

1hls sLudy conslsLs of slx secLlons, whlch can be vlewed lndependenLly or LogeLher as secLlons of a
comprehenslve reporL. 1he lndlvldual secLlons and full reporL can be vlewed aL Lhe cloud compuLlng
research page on Lhe Comp1lA webslLe.

SecLlon 1: MarkeL Cvervlew
SecLlon 2: Channel 8uslness Model Analysls
SecLlon 3: Channel 8elaLlonshlps: vendors, CusLomers, and ulsLrlbuLors
SecLlon 4: Lnd user 8uslness Model Analysls
SecLlon 3: Lnd user Cloud MlgraLlon lssues
SecLlon 6: Cloud Challenges and CpporLunlLles

1hls sLudy was conducLed ln Lwo parLs.

ÞarL l: quanLlLaLlve onllne survey of 301 l1 and buslness professlonals ln Lhe unlLed SLaLes lnvolved ln l1
declslon-maklng (aka end users). uaLa collecLlon occurred durlng !une 2013. 1he margln of sampllng
error aL Lhe 93° confldence level for Lhe u.S. resulLs ls +/- 4.3 percenLage polnLs. Sampllng error ls larger
for subgroups of Lhe daLa.

ÞarL ll: quanLlLaLlve onllne survey of 400 l1 flrms ln Lhe unlLed SLaLes (aka channel). uaLa collecLlon
occurred durlng !une 2013. 1he margln of sampllng error aL Lhe 93° confldence level for Lhe u.S. resulLs
ls +/- 3.0 percenLage polnLs. Sampllng error ls larger for subgroups of Lhe daLa.

lor boLh surveys, Comp1lA worked wlLh Lhe research flrm 8esearch now Lo faclllLaLe daLa collecLlon
uslng an lndependenL panel. As wlLh any survey, sampllng error ls only one source of posslble error.
Whlle non-sampllng error cannoL be accuraLely calculaLed, precauLlonary sLeps were Laken ln all phases
of Lhe survey deslgn, collecLlon and processlng of Lhe daLa Lo mlnlmlze lLs lnfluence.

Comp1lA ls responslble for all conLenL conLalned ln Lhls serles. Any quesLlons regardlng Lhe sLudy should
be dlrecLed Lo Comp1lA MarkeL 8esearch sLaff aL research[

Comp1lA ls a member of Lhe MarkeLlng 8esearch AssoclaLlon (M8A) and adheres Lo Lhe M8A's Code of
MarkeL 8esearch SLandards.

F OU R T H ANNUAL • AU GU S T 2 0 1 3

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 2
/+0 1#(2%)

• As cloud componenLs are becomlng more prevalenL ln l1 archlLecLures, more companles are relylng
on cloud compuLlng for buslness processes such as sLorage (39°), buslness conLlnulLy/dlsasLer
recovery (48°), and securlLy (44°). 1hls sLrong usage and sLrong markeL lndlcaLors show LhaL cloud
compuLlng ls becomlng a defaulL parL of Lhe l1 landscape.

• AlLhough adopLlon raLes are hlgh and markeL numbers are poslLlve, Lhere ls sLlll confuslon relaLed Lo
cloud compuLlng. 1hls confuslon wlll hlnder end users and channel flrms from fully Lransformlng
Lhelr l1 pracLlces and offerlngs. Cnly 46° of channel flrms wlLh cloud offerlngs descrlbed Lhelr cloud
buslness as compleLely maLure-an esLabllshed, sLraLeglc parL of buslness plans.

• 1he lmpacL of cloud compuLlng wlll conLlnue Lo drlve companles Lo clarlfy lssues around usage and
move Lowards greaLer maLurlLy wlLh Lhe model. Lnd users sLlll reporL a wlde range of beneflLs, led
by Lhe ablllLy Lo cuL cosLs. ApproxlmaLely half of channel flrms see fasLer revenue growLh and larger
proflLs from cloud offerlngs Lhan from LradlLlonal offerlngs.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 3
34#$5 6+.'2#4#70 8#9(27 :2%# %'+ 8,(2)%-+,;

ln Comp1lA's 5* 5&-'."+6 7'")118 9:;< reporL, Lhe Lrend for cloud compuLlng was LhaL companles would
begln Laklng Lhe Lechnology for granLed. 1hls pro[ecLlon meanL LhaL companles would begln movlng
away from cloud LranslLlons as sLandalone pro[ecLs and begln vlewlng cloud componenLs as pleces
wlLhln Lhelr overall l1 archlLecLure.

uaLa from Lhe !
%&&'() *+,&-. /& 0)1'- 0123'"/&4 sLudy seem Lo conflrm Lhls behavlor. Whlle 80° of
end user companles ln lasL year's survey clalmed Lo use some form of cloud compuLlng, LhaL number
cllmbed Lo 90° Lhls year. lurLhermore, companles are esLlmaLlng larger percenLages of Lhelr
lnfrasLrucLure and appllcaLlons belng comprlsed by cloud sysLems.

As cloud componenLs are becomlng more prevalenL ln l1 archlLecLures, more companles are relylng on
cloud compuLlng for buslness processes such as sLorage (39°), buslness conLlnulLy/dlsasLer recovery
(48°), and securlLy (44°). 1hls ls belng accompllshed Lhrough use of all Lhree models for cloud
sysLems-SofLware as a Servlce (SaaS), ÞlaLform as a Servlce (ÞaaS), and lnfrasLrucLure as a Servlce
(laaS). ?ear over year, adopLlon raLes for all Lhree models has lncreased, especlally among medlum-slzed
buslnesses (see Appendlx for deLalls).

Company l1 SysLems lncreaslngly Cloud-based
!#0)"# )!#0("# ("#0!""#
&"!& &"!)
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 301 u.S. l1 and buslness execuuves (end users)

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 4
Whlle fuLure ouLlooks on Lhe use of laaS/ÞaaS lndlcaLe LhaL momenLum ls slowlng, Lhls ls lndlcaLlve LhaL
Lhe markeL ls reachlng a saLuraLlon plaLeau, noL a slgn LhaL cloud compuLlng ls falllng ouL of favor. Cnly 3
ouL of 10 buslnesses LhaL are noL currenLly uslng Lhose models plan Lo do so ln Lhe fuLure, a 10° drop
from lasL year. 1he drop comes mosLly among small and medlum buslnesses-large buslness lnLeresL
remalned flaL aL 33°. 1hese companles have exLenslve lnfrasLrucLure and greaLer requlremenLs for
LesL/dev envlronmenLs, so conLlnued lnLeresL here shows LhaL laLe adopLers clearly see poLenLlal ln
movlng Lo a cloud model.

MarkeL slzlng esLlmaLes also lndlcaLe sLablllLy and maLurlLy. 1he followlng daLa polnLs all suggesL robusL
spendlng on cloud servlces:

• CarLner esLlmaLes LhaL Lhe global publlc cloud servlces markeL wlll accounL for $131 bllllon
ln revenue ln 2013, an 18.3° lncrease from 2012. 1hls lncludes dlrecL spendlng on Lhe Lhree
prlmary cloud models as well as lndlrecL spendlng on buslness process ouLsourclng LhaL ls
offered from a cloud foundaLlon.

• Accordlng Lo luC, worldwlde spendlng on hosLed prlvaLe cloud soluLlons wlll grow aL a
compound annual growLh raLe of over 30° durlng 2012-2016, evenLually LoLallng $24 bllllon
ln revenue by 2016. 1he soluLlons lnclude boLh dedlcaLed prlvaLe clouds, where physlcal
lnfrasLrucLure ls asslgned Lo a slngle user, and vlrLual prlvaLe clouds, where prlvaLe clouds
for a slngle user are bullL vlrLually wlLhln a publlc cloud provlder.

• Analysys Mason-a research consulLancy focused on Lelecom, medla, and Lechnology-
predlcLs LhaL enLerprlse cloud servlces revenue wlll reach $31.9 bllllon by 2017. 1hey lnclude
only SaaS and laaS ln Lhls pro[ecLlon. ln Lhelr sLudy year-over-year growLh slows from 17° ln
2013 Lo less Lhan 3° by 2017, buL emerglng markeLs wlll experlence greaLer growLh,
accounLlng for sllghLly more Lhan 10° of Lhe overall cloud spend by 2017.

• lndlvldual cloud provlders dlffer ln Lhelr approach Lo releaslng flnanclal lnformaLlon as Lhey
conLend for markeL poslLlon, buL cloud servlces are clearly playlng slgnlflcanL roles ln overall
operaLlons. Amazon chooses Lo place Amazon Web Servlces under ºoLher operaLlng
segmenLs" raLher Lhan break ouL Lhe deLalls separaLely, and McCuarrle CaplLal esLlmaLes
LhaL AWS wlll generaLe $3.8 bllllon ln revenue ln 2013. ln conLrasL, MlcrosofL publlcly LouLed
Lhelr cloud servlces as a $1 bllllon buslness, and l8M reporLed LhaL cloud revenue grew over
70 percenL from 1C 2012 Lo 1C 2013 (wlLhouL dlscloslng Lhe acLual revenue number).

lL seems, Lhen, LhaL cloud compuLlng has become well esLabllshed as an opLlon for l1 operaLlons.
1wenLy-seven percenL of companles wlLh cloud as parL of Lhelr archlLecLure have been uslng cloud
soluLlons for Lhree years or more, and Lhey have moved pasL experlmenLs and Lowards a vlew of cloud
Lechnology as a glven plece of Lhelr overall lnfrasLrucLure.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 3
34#$5 <$)(2+)) 8#5+4) =%(44 >4$)(9+

1he lmpresslve markeL numbers and hlgh
adopLlon raLes may noL Lell Lhe whole
sLory, Lhough. 1here are many facLors
LhaL could sLlll be conLrlbuLlng Lo
confuslon abouL cloud compuLlng (see
sldebar). Companles may noL acLually be
uslng cloud soluLlons Lo Lhe exLenL Lhey
belleve, and Lrue buslness
LransformaLlon wlll cerLalnly lag slmpler
Lechnology adopLlon. Companles LhaL
are compleLely comforLable wlLh cloud
compuLlng concepLs such as self-servlce
and dynamlc resource allocaLlon wlll
adopL operaLlonal models LhaL dlffer
greaLly from models LhaL rely on more
LradlLlonal l1 sysLems. Movlng Lowards
Lhese new models ls more of a challenge
Lhan LranslLlonlng selecLed sysLems lnLo
cloud Lechnology.

Comp1lA's daLa from channel flrms
demonsLraLes Lhls dlfflculLy. Cf 400 flrms
who offer cloud soluLlons, only 46°
descrlbed Lhelr cloud buslness as
compleLely maLure-an esLabllshed,
sLraLeglc parL of buslness plans. 1he
remalnder descrlbed Lhelr cloud buslness
elLher as a porLlon of Lhe buslness
runnlng alongslde oLher offerlngs or as
plloL/evaluaLlon lnlLlaLlves. As boLh
channel flrms and end users Lruly
embrace cloud operaLlons, Lhere wlll
be less dlsLlncLlon beLween cloud
offerlngs and non-cloud offerlngs, and
Lhe flnal soluLlon wlll slmply be
produced regardless of Lhe
lnfrasLrucLure used.

Confuslon among end users wlll delay
Lhelr move Lo a compleLe cloud
buslness model. lf a company's l1
soluLlons are meeLlng Lhelr needs, lL
may noL be crlLlcally lmporLanL Lo know
Lhe dlfferences beLween hosLed
offerlngs vs. cloud offerlngs. lf Lhe
=#$-.+) #? 3#2?$)(#2 (2 %'+ 34#$5 8,-@+%

• 34#$5A,)'(27B As producLs and servlces are
lncreaslngly labeled wlLh ºcloud" ln order Lo
capLure Lhe growlng markeL, Lhe characLerlsLlcs of
a Lrue cloud offerlng become less dlsLlncL.

• C#A2D4,0(27 %'+ .4#$5 %-+25B Many l1
professlonals or lndusLry waLchers are qulck Lo
dlsmlss cloud as ºnoLhlng new." Whlle Lhere are
aspecLs of cloud LhaL can be found ln compuLlng
models daLlng back Lo Lhe lnLroducLlon of Lhe
malnframe, Lhere are also unlque aspecLs LhaL may
geL overlooked wlLhouL Lhe proper educaLlon.

• E0D+ ,25 ?,%(7$+B Some end users may be overly
eager Lo descrlbe Lhelr lnfrasLrucLure as cloud-
based Lo have Lhe appearance of keeplng up wlLh
Lhe Lrend. Slmllarly, some users may slmply apply
Lhe label Lo Lhelr sysLems wlLhouL much LhoughL ln
an efforL Lo avold any furLher dlscusslon on Lhe

Confuslon 8eLween Cloud/PosLed Models ÞerslsLs
no dlñerences/uon'L know
Mlnor dlñerences
Ma[or dlñerences
Source: !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 300 u.S. l1 and buslness execuuves (end users)

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 6
company decldes Lo pursue cloud-speclflc beneflLs, Lhough, Lhey may be surprlsed Lo flnd LhaL Lhelr
sysLems are noL as cloud-enabled as Lhey mlghL have lmaglned.

1hls confuslon ls also presenL ln Lhe general ecosysLem, as evldenced by cloud provlder brand
awareness. More Lhan 9 ln 10 end users belleve Lhere ls a hlgh or moderaLe degree of dlfferenLlaLlon
among cloud provlders, buL Lhe percepLlon of Lop provlders favors LradlLlonal l1 companles (such as
MlcrosofL, l8M, and PÞ) over companles LhaL are generally regarded as leaders wlLhln Lhe cloud lndusLry
(such as Amazon and 8ackspace). Whlle Lhe LradlLlonal companles do offer cloud soluLlons (and Lhe
awareness levels may be a good lndlcaLor for Lhelr fuLure prospecLs), lower awareness of leadlng cloud
provlders llkely lmplles lower awareness of leadlng-edge lnnovaLlons.

1o beLLer descrlbe Lhe many nuances of Lhe cloud Lrend, boLh from Lhe end user perspecLlve and Lhe l1
soluLlon provlder/cloud provlder perspecLlve (aka ºLhe channel"), Lhls reporL wlll use Lhe followlng
models. See SecLlons 2 and 4 for addlLlonal deLalls and analysls.

!""# % &""# % #""#
&'()"*+ % ,-.'(/ 0123/4
&"-5+6* 72/+14
8/39":26 ;'"64(:26
=2'>?2@ 0A"6<+
&'22E 2E 9269+F*
0123/ G/2F:26 &'2<'+44(26 72/+1 E2' 86/ H4+'4
&3.1(9 0123/4
&'2)(/+' #@(*9A(6<
'()*#%+ ,
/.01$+ ,
50.04+ ,
!""# % &""# % #""#
&'()*+ % &,*-"./ % 01(,*2 3)4'25
3)4'2 677,/7"849 % :,4;/,"7/ #/,-*+/5
&,4+',/ 0< =
#<> "22
/?@/,85/ .4
('*)2 +)4'2
045897 D E*,/+.
.4 +'5.4F/,
3495')897 % 62-*5*97 % !G #4)'8495
I"9"7/2 #/,-*+/5
3'5.4F*J"849 D
6,+B*./+.',/ D
3)4'2 :'5*9/55 I42/)5 H,"F/K4,; L4, .B/ 3B"99/)
3',22+4 1+-)D+.%(9+
>25 F)+- 1+-)D+.%(9+

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 7
:;D,.% #? 34#$5 3#;D$%(27

As companles galn more clarlLy abouL cloud sysLems and move Lowards a malnsLream vlew of cloud as
an l1 opLlon, Lhe lmpacLs of movlng Lo a cloud model are belng seen across Lhe Lechnology and buslness
landscape. ConLlnulng a Lrend from prevlous years, many respondenLs vlew cloud as Lhe besL opLlon for
a varleLy of reasons, wlLh no slngle reason sLandlng ouL as Lhe lone drlver or beneflL. Also, Lhe reallzed
beneflLs are LlghLly clusLered LogeLher, lendlng welghL Lo Lhe argumenL LhaL cloud compuLlng offers
many advanLages.

8esearch from cloud managemenL
company 8lghLscale shows LhaL cloud
beneflLs are more wldely experlenced
as companles galn maLurlLy wlLh Lhe
cloud model, and Comp1lA's daLa
corroboraLes Lhls flndlng. Looklng aL
Lhe beneflL of ºbeLLer opLlon for
mulLlple reasons" ln Comp1lA's
survey, 30° of laLe adopLers (less
Lhan one year of cloud experlence)
rank Lhls as a beneflL. 1hls number
rlses Lo 39° wlLh mlddle adopLers
(beLween 1 and 3 years of cloud
experlence) and [umps Lo 36° for
early adopLers (more Lhan 3 years of
cloud experlence). Clearly, more Llme
spenL uslng cloud sysLems leads Lo a
wlder appreclaLlon of Lhe advanLages
Lhey offer.

lrom a channel perspecLlve, Lhe mosL relevanL lmpacL ls Lhe effecL cloud ls havlng on Lhe boLLom llne. As
expecLed, Lhe maLurlLy of a flrm's cloud offerlngs has a dlrecL correlaLlon Lo Lhe flnanclal lmpacL Lhey are
seelng. 1wenLy-nlne percenL of flrms LhaL descrlbed Lhelr offerlngs as ºnoL maLure" saw fasLer cloud
revenue growLh compared Lo 34° of ºmaLure" companles. lor proflL margln, Lhe dlfference ls even
more drasLlc-21° of ºnoL maLure" companles see larger cloud proflL marglns compared Lo 38° of
ºmaLure" companles.
8eneñLs Cloud AdopLers Pave Lxperlenced
ÞredlcLable prlclng
8educuon ln lnLernal l1 headcounL
Slmple/fasL lmplemenLauon
8educe operauonal complexlLy
Creauon of new oñerlngs or servlces
8uslness unlLs operaLe more freely
8euer model for llcenslng/upgrades
Add new capablllues or feaLures
8educe caplLal expendlLures
Modernlzauon of legacy l1
8euer opuon for muluple reasons
AblllLy Lo cuL cosLs
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 449 u.S. end users wlLh cloud soluuons

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 8

lor boLh end users and channel flrms, one of Lhe largesL challenges ln movlng Lowards more robusL
cloud archlLecLure ls flndlng or bulldlng Lhe rlghL skllls. Accordlng Lo a [olnL sLudy by luC and MlcrosofL,
1.7 mllllon cloud-relaLed [obs worldwlde wenL unfllled ln 2012. Plrlng managers clLed dlfflculLy ln flndlng
hlgher-level skllls such as bulldlng mlgraLlon plans or assesslng Lhe rlsk of a cloud LranslLlon. uaLa from
8urnlng Class 1echnologles Labor lnslghLs also shows demand for cloud skllls growlng: from 2011 Lo
2012, Lhe number of posLlngs referenclng cloud grew by 30°. Cver LhaL same Llme perlod, Lhe number
of core l1 [ob openlngs acLually decllned, showlng how lmporLanL cloud compuLlng ls ln Lhe modern l1

l1 workers are aware of Lhls skllls gap. uaLa from Comp1lA's 9
%&&'() 5* 0(+,,+ 5&./4#". sLudy shows
LhaL cloud compuLlng ls Lhe Lhlrd-hlghesL sklll area LhaL l1 professlonals would llke Lo develop. 1he Lop
areas are neLworks/lnfrasLrucLure and servers, boLh of whlch are cerLalnly lnvolved ln cloud soluLlons. As
cloud compuLlng becomes even more malnsLream, Lhe demand for Lhe proper skllls wlll conLlnue Lo rlse
as companles seek Lo galn Lhe beneflLs of Lhls powerful model.

!" $%&''() *+,-. /(( 01.+23( 4(3('5(
!-6&78 9,1- $)15: ;<(,+'=.
!"#$% '() *(+,-".(/0% 12#%$3+(4
502'.30( 60'07$0 82#9+/
=,1@+'= 9&.8(,
A,1@+'= &8
.&-( ,&8(
$)15: =,1@+'=
/15,7(B $1-6"!CD. !
%&&'() *+,&-. /& 0)1'- 0123'4&5
E&.(B FGG 7%&''() H,-. @+8% 7)15: 1<(,+'=.
!"#$% '() *(+,-".(/0% 12#%$3+(4
502'.30( 12#?+ @,2A.7(
415=%)I 8%(

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 1 9

use of laaS/ÞaaS
!"##$%&'( *+,%- .//012//0
2'/% &8 *+$ ,% 9$:& ;$/#
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 301 u.S. l1 and buslness execuuves (end users)
Cverall = 48°
Cverall = 29°
8uslness producuvlLy
43° 43° 48°
31° 39° 49°
Web presence
48° 33° 46°
vlrLual deskLop
34° 43° 47°
36° 32° 48°
Analyucs 34° 43° 47°
32° 49° 44°
P8 managemenL
28° 43° 32°
Lxpense managemenL
23° 47° 39°
Pelp desk
26° 38° 33°
llnanclal ManagemenL
36° 43° 33°
23° 37° 37°
Call cenLer
13° 33° 31°
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 301 u.S. l1 and buslness execuuves (end users)
use of Cloud-8ased Appllcauons
F OU R T H ANNUAL • AU GU S T 2 0 1 3

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 2
/+0 1#(2%)

• Palf of channel flrms Loday descrlbe Lhelr cloud buslness as ºmaLure," whlle anoLher 4 ln 10 clalm Lo
have reached aL leasL a degree of cloud servlces' maLurlLy alongslde Lhelr more esLabllshed llnes of
buslness. 8y conLrasL, [usL 13° sald Lhelr currenL cloud buslness ls noL maLure aL all Loday.

• Comp1lA has ldenLlfled four prlmary cloud buslness models for Lhe channel: 8ulld,
Þrovlde/Þrovlslon, Lnable/lnLegraLe and Manage/SupporL. Seven ln 10 soluLlon provlders lnvolved ln
cloud sLarLed wlLh Lhe 8ulld buslness model buL are now lnvolved wlLh one or more of Lhe oLher
Lhree frameworks. 1wenLy-slx percenL of Lhem are conducLlng buslness across all four areas.

• Slx ln 10 channel flrms lnvolved ln Lhe Manage/SupporL buslness model have creaLed l1 dashboards
LhaL allow cusLomers Lo Lrack Lhelr cloud uLlllzaLlon, cosLs and oLher meLrlcs Lo undersLand Lhelr

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 3
3$)(2+)) 4-,2)5#-6,%(#2 7(%'(2 %'+ 8',22+9: 1,%' %# %'+ 89#$;

lor Lhe channel's parL, Lhe Loplc of cloud buslness models has been one of Lhe mosL heavlly debaLed
among lndusLry flrms. 1he dlscusslon has moved from early-day concerns LhaL shlfLlng l1 Lo Lhe cloud
would LhreaLen Lhe channel's very exlsLence Lo Loday's reallzaLlon LhaL cloud has ln facL spawned a wlde
varleLy of buslness models from whlch Lo galn a fooLhold ln Lhls arena.

Þerhaps Lhe spaLe of buslness model opLlons ls helplng Lo drlve cloud's conLlnued acceleraLlon across
Lhe l1 channel boLh ln Lerms of adopLlon and sLraLeglc value. Comp1lA's 4
Aooool 1teoJs lo clooJ
compotloq sLudy found LhaL half of soluLlon provlder flrms Loday descrlbe Lhelr cloud buslness as
ºmaLure," whlle anoLher 4 ln 10 clalm Lo have reached aL leasL a degree of cloud servlces' maLurlLy
alongslde Lhelr more esLabllshed llnes of buslness. 8y conLrasL, [usL 13° sald Lhelr currenL cloud buslness
ls noL maLure aL all Loday. Smaller flrms (47°), lnLeresLlngly, were more llkely Lo self-selecL as havlng a
cloud operaLlon LhaL ls maLure and esLabllshed, compared wlLh Lhe largesL companles (40°). 1hls ls
llkely due Lo Lhe facL LhaL many of Lhe smaller flrms could very well be ºborn ln Lhe cloud" versus larger
and more LradlLlonal soluLlon provlders LhaL wlll need Llme Lo plvoL Lo a new model.
1he growlng buslness acumen around cloud also correlaLes Lo Lhe
level of lmporLance buslnesses are placlng on Lhe model as a revenue
opporLunlLy. Cn average, a llLLle more Lhan half of channel flrms ln
Lhe Comp1lA sLudy descrlbe Lhe cloud revenue Lhey generaLed ln Lhe
pasL year as sLraLeglc and core Lo Lhelr buslness, as opposed Lo
slmply opporLunlsLlc dollars. ln pasL years, cloud revenue was
Lyplcally of a more LacLlcal naLure, an ad[uncL Lo more esLabllshed
sLreams of buslness such as producL LransacLlons or servlces
engagemenLs. Conslder Lhe slLuaLlon ln 2010, Lhe flrsL year Lhe
Comp1lA sLudy was conducLed, when barely a bllp of channel
companles reporLed heavy cloud lnvolvemenL and ln facL, Lhe ma[orlLy reporLed none aL all.
Clven Lhe more advanced lnvolvemenL ln cloud, Lhe 2013 research capLures addlLlonal nuLs and bolLs
lnformaLlon on channel flrm buslness model cholces, cloud-based proflL marglns, cusLomer and vendor
relaLlonshlps and lnlLlal enLry polnLs Lo Lhe cloud world. Cn LhaL lasL polnL, 3 ln 10 respondenLs descrlbe
Lhemselves as ºborn ln Lhe cloud," sLarLlng from scraLch as cloud-based enLlLles. 1haL revelaLlon
suggesLs LhaL a number of new channel companles have come lnLo exlsLence ln Lhe lasL few years,
perhaps offseLLlng Lhe wave of aLLrlLlon among esLabllshed soluLlon provlder flrms, some of whom are
reachlng a naLural buslness end of llfe and oLhers LhaL have been forced Lo shuLLer due Lo poor buslness
declslons, such as falllng Lo dlverslfy.
Whlle ºborn ln Lhe cloud" companles are mulLlplylng, Lhe facL remalns LhaL mosL channel flrms LhaL have
adopLed cloud have done so from Lhelr exlsLlng poslLlon reselllng l1 producLs, servlces and consulLlng.
nearly half (43°) of Lhese LradlLlonal soluLlon provlder companles descrlbed Lhelr LranslLlon Lo cloud as
rapld, whlle anoLher fourLh characLerlzed Lhelr addlLlon of cloud soluLlons as havlng happened gradually.
no maLLer Lhe pace of change, all of Lhese flrms have had Lo make sLraLeglc and LacLlcal declslons abouL
whlch cloud buslness model or models Lo pursue. 1hey base Lhese dlrecLlonal cholces on a number of
facLors, lncludlng Lhe company's currenL operaLlng model, speclflc capablllLles and sLafflng, as well as
focus areas ln Lhe markeL. 1o capLure lnformaLlon abouL Lhelr cloud buslness, Lhe Comp1lA sLudy
presenLed respondenLs wlLh four broad Lypes of cloud operaLlng models and asked whlch, lf any, Lhey
have engaged ln durlng Lhe pasL year.
<#7 =%,>+ #5 89#$; ?,%$-(%0:
@',% () %'+ A$%$-+B
39° - ConfldenL on Lhe rlghL paLh
31° - SomewhaL confldenL
8° - noL LhaL confldenL

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 4
8uslness model declslons pose a Lrlcky Lask for channel flrms maklng - or havlng made - a cloud
LranslLlon. Þrlmary conslderaLlons depend on where a company wanLs Lo go wlLh cloud: 8esell a
vendor's cloud soluLlons? AggregaLe and broker cloud servlces from a varleLy of dlfferenL sources?
lnLegraLe and cusLomlze cloud-based apps and servlces, or slmply sell Lhe lnfrasLrucLure Lo an end user
and provlde consulLlng? Lach of Lhese paLhs and more are posslblllLles, as are varylng revenue models
avallable for all. lorLy-flve percenL of channel companles from Comp1lA's 4
Aooool 1teoJs lo clooJ
compotloq say deLermlnlng Lhe approprlaLe buslness model around cloud presenLed a slgnlflcanL
challenge ln Lhe pasL year. 1haL asserLlon falls [usL behlnd Lhe mosL dlfflculL challenge clLed ln Lhe
flndlngs: ueveloplng cloud experLlse across boLh Lechnlcal and sales arms wlLhln a company, a Lask LhaL
loglcally flows oftet Lhe lnlLlal buslness model declslon ls made.

1he above dlagram deflnes Lhe four buslness model areas, along wlLh Lhe subcaLegorles of work acLlvlLy
LhaL are assoclaLed wlLh Lhem. 1he longer bars, such as laaS/ÞaaS/SaaS and ConsulLlng/Advlslng/l1
SoluLlons represenL acLlvlLles LhaL can accompany any of Lhe four maln buslness models. See subsequenL
pages for addlLlonal deLalls.

lL should be noLed LhaL Lhese four buckeLs are by no means reflecLlve of every posslble Lype of channel
model appllcable Lo cloud, however Lhey do represenL a falrly compleLe plcLure of Lhe Lype of cloud
work belng done across Loday's channel.

'()*#%+ ,
/.01$+ ,
50.04+ ,
!""# % &""# % #""#
&'()*+ % &,*-"./ % 01(,*2 3)4'25
3)4'2 677,/7"849 % :,4;/,"7/ #/,-*+/5
&,4+',/ 0< =
#<> "22
/?@/,85/ .4
('*)2 +)4'2
045897 D E*,/+.
.4 +'5.4F/,
3495')897 % 62-*5*97 % !G #4)'8495
I"9"7/2 #/,-*+/5
3'5.4F*J"849 D
6,+B*./+.',/ D
3)4'2 :'5*9/55 I42/)5 H,"F/K4,; L4, .B/ 3B"99/)

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 3
1he charLs below provlde a gllmpse lnLo Lhe level of acLlvlLy each cloud buslness model has garnered Lo
daLe ln Lhe channel, along wlLh Lhe relaLlve value belng placed on each of Lhe four areas. Many soluLlon
provlders are lnvolved wlLh more Lhan one of Lhe models below, wlLh 26° of Lhem conducLlng buslness
across all four areas. AddlLlonally, Lhe second charL lndlcaLes Lhe order ln whlch Lhose channel flrms
lnvolved ln Lwo or more of Lhese buslness models goL sLarLed. 1he ma[orlLy launched lnLo cloud wlLh
offerlngs and work ln Lhe 8ulld caLegory.

Cloud 8uslness Model AcuvlLy/lmporLance
!"#$% '( )$*+,%** -'.%#
/012+1%* 3' 45",,%# 6+78*
8ulld 46° 37°
Þrovlde/Þrovlslon 49° 38°
Lnable/lnLegraLe 33° 33°
Manage/SupporL 38° 32°
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
48° 49°
8ulld Þrovlde/
4#'$. )$*+,%** -'.%# >,2'#2%8%,3
"8',: 45",,%# ?"73,%7*
?#", 3'
!" $%&'
(%)*+,+-./// 0'1 2-, 3%, 415
71° 13° 3° 9°
30° 33° 29° 7°
32° 32° 23° 10°
Manage SupporL
29° 34° 17° 20°
8ank Crder ln Whlch Channel llrms AdopLed Cne or
More of Lhe varlous Cloud 8uslness Models
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
Among Lhose channel ñrms provldlng servlces ln all four cloud areas Loday, 647
sLarLed Lhelr cloud pracuce ln a 8ulld mode. 8uL whlle 8ulld ls Lhe mosL common
sLarung polnL for cloud adopuon, Lhe Manage/SupporL buckeL ls Lhe mosL
prevalenL among ñrms Loday.

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 6
89#$; ?#;+9 C9+6+2% D: 3$(9;

A sLrong sLarLlng polnL for Lhe cloud for many flrms ls Lhe caLegory of 8ulld, whlch ls deflned ln Lhe sLudy

1yplcolly tbls lovolves fltms ptocotloq veoJot-boseJ botJwote ooJ softwote ptoJocts to coosttoct ptlvote
ooJ/ot bybtlJ clooJs fot costomets. 1bey moy olso offet coosoltloq qolJooce oo tbe best l1 otcbltectote,
cooflqototloo ooJ ptoJoct cbolces fot tbe ptoject.

8oughly half (48°) of channel flrms Loday reporL currenLly offerlng some form of 8ulld servlces, wlLh
anoLher Lhlrd plannlng Lo add Lhls flavor of cloud Lo Lhelr llneup ln Lhe comlng year. 8ulld represenLs a
blL of a cloud launchlng pad, of Lhose flrms LhaL are also operaLlng one of Lhe oLher Lhree cloud buslness
models, 7 ln 10 sLarLed wlLh a wlLh 8ulld pracLlce before addlng Lhe oLhers. 1hls makes sense lf you Lhlnk
back Lhree Lo flve years ago as Lhe lndusLry and channel began lLs cloud LranslLlon. AL Lhe Llme, many
channel flrms were looklng for Lhe mosL obvlous place Lo plug ln, wlLh many of Lhem admlLLedly worrled
LhaL cloud would dlslnLermedlaLe Lhem ln Lhe markeL. So where dld Lhey Lurn ln con[urlng a buslness
model LhaL made sense? 1hey leveraged LradlLlonal soluLlon provlder sLrengLhs ln lnfrasLrucLure
producLs sales and lmplemenLaLlon work. ln dolng so ln a cloud conLexL, Lhese companles poslLloned
Lhemselves as bullders of prlvaLe clouds wlLhln exlsLlng cusLomers' on-premlse envlronmenLs.

AddlLlonally, a number of vendors lncludlng PÞ, vMware, CA and oLhers have broughL Lo markeL ºcloud-
ln-a-box" Lype producLs LhaL ouLflL channel parLners wlLh a Lurnkey way Lo sell and deploy prlvaLe cloud
soluLlons for Lhelr end cusLomers. Large-scale enLerprlse prlvaLe cloud deploymenLs are noL Lhe
malnsLay of mosL of Loday's channel, whose bread and buLLer cusLomers Lyplcally reslde ln Lhe SM8
space. So Lhese cloud-ln-a-box soluLlons and LoolseLs have allowed many of Lhem Lo provlde 8ulld-Lype
cloud servlces wlLhln Lhe conflnes of Lhelr sklll seLs, cusLomer needs and flnances. 1hey also allow for
much fasLer Llme Lo compleLlon Lhan a ma[or prlvaLe cloud enLerprlse deploymenL from scraLch.

lurLher evldence LhaL 8ulld acLlvlLles seemed a loglcal [ump-on polnL for channel flrms mulllng cloud
appears when you look aL whlch Lypes of companles have adopLed Lhls model. Slx ln 10 channel flrms
LhaL ldenLlfy as resellers are currenLly offerlng 8ulld servlces vs. 47° of channel companles LhaL allgn
more ln Lhe servlces buckeL (break/flx, managed servlces, consulLlng). lL's noL an overwhelmlng
dlfference, buL Lhe resellers, wlLh Lhelr fooLhold ln lnfrasLrucLure producL sales, would flnd 8ulld a
naLural enLry polnL Lo cloud.

ln Lerms of focus areas wlLhln Lhe 8ulld framework, 31° of respondenLs characLerlzed whaL Lhey do as
mosLly procurlng hardware and sofLware Lools, Lhen reselllng Lo cusLomers Lo bulld prlvaLe clouds.
AnoLher 2 ln 10 say Lhelr 8ulld buslness ls prlmarlly abouL leveraglng an exlsLlng cusLomer's on-premlse
hardware and layerlng sofLware and managemenL Lools Lo bulld a prlvaLe or hybrld cloud soluLlon.
WhaL's more common Lhan elLher of Lhese acLlvlLles ln lsolaLlon ls a model ln whlch channel flrms do
boLh, whlch roughly half of respondenLs sald Lhey dld.

1hlrLy-elghL percenL of small flrms parLlclpaLed ln boLh Lypes of 8ulld acLlvlLles compared wlLh 33° of
medlum and large companles LhaL dld.

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 7
89#$; ?#;+9 C9+6+2% E: 1-#F(;+G1-#F()(#2

1hls buslness model for cloud poslLlons Lhe soluLlon provlder as Lhe hub for provlslonlng varlous vendor-
based and homegrown cloud servlces Lo Lhe end cusLomer. Palf of channel flrms dolng cloud Loday are
parLlclpaLlng ln Lhe Þrovlde/Þrovlslon model, wlLh a Lhlrd of all respondenLs ln Lhe sLudy bellevlng LhaL
Lhls buslness model for cloud has Lhe mosL growLh poLenLlal ln Lhe nexL Lwo years - regardless of
wheLher Lhey are lnvolved ln lL or noL Loday. 1haL compares wlLh roughly a quarLer of flrms LhaL deem
each of Lhe oLher Lhree buslness models as Lhe fasLesL growLh drlvers.

A qulck deflnlLlon:

1yplcolly tbese fltms ote tesellloq ooJ ptovlslooloq loo5 offetloqs socb os stotoqe ooJ compote copoclty
ftom Amozoo web 5etvlces ot 5oo5 offetloqs socb os Cooqle Apps. 1bey ote olso evolootloq ooJ
oqqteqotloq votloos clooJ setvlces fot costomets.

WlLhln Lhls model, Lhere are several subcaLegorles of acLlvlLles LhaL channel flrms pursue Loday. Many
parLlclpaLe ln a number of Lhese areas. Slx ln 10 channel flrms lnvolved ln cloud reporL whlLe labellng a
vendor's cloud offerlngs, ln effecL leadlng wlLh Lhelr own brand on whaL ls Lyplcally a SaaS deal. AnoLher
slx ln 10 sald Lhey operaLe Lhelr own daLa cenLer and sell homegrown cloud servlces, whlle [usL more
Lhan half (34°) are reselllng vendor-based cloud servlces LhaL can lnclude SaaS and laaS offerlngs ln
mosL cases. 1hlnk MlcrosofL Cfflce 363 or Coogle Apps, as menLloned above.

noL surprlslngly, resellers are much more llkely among channel
flrms dolng cloud Lo sell vendor-based cloud offerlngs. SevenLy-
Lwo percenL of channel flrms LhaL ldenLlfy as resellers dld so ln
Lhe lasL year, as opposed Lo half of channel companles LhaL are
malnly sLeeped ln Lhe servlces buslness. 1he Þrovlde/Þrovlslon
model has perhaps Lhe lowesL barrler Lo enLry of all four of Lhe
cloud buslness frameworks. lL's noL a blg leap Lo resell a vendor's
SaaS appllcaLlons lf you are already reselllng LhaL vendor's oLher
producLs and servlces. Powever, Lhe marglns on vendor SaaS
sales Lend Lo be on Lhe lower slde, whlch requlres channel flrms
Lo flnd ways Lo add value Lo lncrease proflLablllLy. Some avenues
Lo pursue: lndusLry secLor speclallzaLlon (e.g. selllng SaaS-based
LlecLronlc Medlcal 8ecords (LM8) sysLems Lo Lhe healLhcare
lndusLry), or developlng capablllLles ln Lhe Lhlrd buslness model
area, Lnable/lnLegraLe (dlscussed below) Lo lncrease marglns.

1hose flrms farLhesL along wlLh cloud are far more llkely Lo be
operaLlng Lhelr own cloud and selllng homegrown servlces dlrecL Lo cusLomers. Seven ln 10 companles
LhaL clalm a maLure and sLraLeglc cloud buslness or have drlven sLakes ln all of Lhe four cloud buslness
models sald Lhey are operaLlng Lhelr own cloud daLa cenLer operaLlon and servlces. 1hls compares wlLh
less Lhan half of companles LhaL are elLher less maLure overall wlLh cloud or parLlclpaLlng ln fewer Lhan
four of Lhe ouLllned buslness models.

Whlch cloud servlces are Lhey selllng? 1he usual suspecLs of sLorage, emall and vlrLual deskLop are
among Lhe mosL popular cloud servlces movlng Lhrough Lhe channel Loday. nearly Lwo-Lhlrds of channel
4#H 89#$; =+-F(.+) =#9; "0
8',22+9 1,-%2+-) <,)% I+,-
! Lmall
! vlrLual deskLop
! Pelp desk
! SLorage
! AnalyLlcs
! ÞroducLlvlLy apps
! 8uslness conLlnulLy (u8)
! llnance-relaLed apps
! CollaboraLlon Lools
! C8M
! Call cenLer

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 8
flrms have Lhese servlces ln Lhelr cloud porLfollo. Cne surprlse? SlxLy-Lwo percenL also sald Lhey sold
cloud-based buslness lnLelllgence and daLa analyLlcs ln Lhe pasL year, a ma[or leap from prevlous years
of Lhls sLudy when LhaL Lech caLegory landed low on Lhe channel offerlngs llsL. ln facL, 73° of reseller-
based channel flrms are offerlng 8l and daLa analyLlcs as a cloud servlce Lo cusLomers. now lL should be
noLed LhaL Lhe sample slze for Lhe 8l responses ls smaller Lhan oLher areas of Lhe Comp1lA sLudy, buL
dlrecLlonally speaklng Lhe rlse ln cloud-based analyLlcs ln Lhe channel lndlcaLes a new level of
sophlsLlcaLlon among Lhese flrms and qulLe llkely belles Lhe lnfluence of 8lg uaLa. WlLh cloud verslons of
analyLlcs Lools now avallable, soluLlon provlders have an opporLunlLy Lo sell Lhelr predomlnanLly SM8
cusLomers soluLlons LhaL may have been beyond flnanclal reach ln Lhe on-premlses form.

89#$; ?#;+9 C9+6+2% J: C2,"9+GK2%+>-,%+

1he Lnable/lnLegraLe buslness framework for cloud has been a sweeL spoL for channel flrms over Lhe
pasL several years. lL makes sense. 8eyond Lhe reselllng of producLs, channel flrms have made Lhelr
name provldlng lnLegraLlon and lmplemenLaLlon work LhaL llnks cusLomers' varlous hardware and
sofLware soluLlons, as well as offerlng verLlcal lndusLry and oLher Lypes of cusLomlzaLlon. Cloud soluLlons
and mulLlcloud envlronmenLs only add more layers of Lechnology Lo connecL lnLo LhaL lnfrasLrucLure.
Conslder Lhe followlng deflnlLlon for Lhe model:

1yplcolly tbey ote ptovlJloq loteqtotloo ooJ lmplemeototloo setvlces tbot moy locloJe tyloq o costomet´s
oo-ptemlse l1 solotloos to lts clooJ-boseJ solotloos ot costomlzloq clooJ-boseJ solotloos to flt o
pottlcolot bosloess oeeJ ot vettlcol.

lor Lhe pasL Lhree years of Comp1lA cloud sLudles, Lhe number one source of posL-sale dollars has been
lnLegraLlon work. 1hls area has rouLlnely been a place where Lhe channel cushlons lLs overall proflL
marglns. Slnce mosL soluLlon provlders charge cusLomers on a recurrlng revenue basls for cloud
soluLlons (by consumpLlon or by number of users eLc.), Lhe pro[ecL work assoclaLed wlLh Lhe
Lnable/lnLegraLe caLegory allows Lhem Lo add revenue noL lncluded ln Lhe base conLracL.

1he buslness model encapsulaLes a
number of acLlvlLles. 8oughly Lwo-
Lhlrds of companles lnvolved ln
Lnable/lnLegraLe are currenLly
provldlng Lhe followlng:
deploymenL of cloud soluLlons,
lnLegraLlon of on-premlse Lools and
lnfrasLrucLure Lo cloud soluLlons
and cusLomlzaLlon, ofLen Lo meeL
verLlcal lndusLry needs. AnoLher 6
ln 10 are engaged ln cloud-based
appllcaLlon developmenL, mosL
llkely bulldlng SaaS apps or
exLenslons Lo apps LhaL Lle cloud ln
wlLh an end cusLomer's currenL
envlronmenL. Many of Lhe ÞaaS
plaLforms ln Lhe cloud slmpllfy Lhls
!"#$"%&'( "*
12344 %/50*6%%-7
40° 36° 49°
44° 38° 30°
49° 66° 30°
Manage SupporL
33° 67° 33°
Medlum-Slzed Channel llrms MosL Llkely Lo Conslder
Cloud Acuvlues Across all Models SLraLeglc
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
8 ln 10 medlum-
slze channel ñrms
are lnvolved wlLh
acuvlues compared
wlLh [usL more Lhan
half of Lhelr smaller

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 9
Lype of work. Meanwhlle, a smaller percenLage (34°) of companles have gone down Lhe road of moblle
appllcaLlon developmenL Lo synch and lnLeracL wlLh cloud soluLlons. As moblle conLlnues Lo grow ln
sLaLure across all Lypes of appllcaLlons, expecL LhaL caLegory Lo acceleraLe. Conslder resellers alone: Lhe
daLa flnd LhaL 73° of Lhem are dolng some work ln Lhls moblle appllcaLlon developmenL versus 37° of
channel flrms LhaL ldenLlfy more sLrongly as servlce provlders.

ln Lerms of oLher segmenLaLlon dlfferences, flrms LhaL conslder Lhemselves maLure wlLh cloud are more
llkely Lhan Lhose wlLh moderaLe Lo no maLurlLy Lo provlde work ln Lhe Lnable/lnLegraLe caLegory. nearly
8 ln 10 companles wlLh sLraLeglc cloud buslness acumen are planLed ln Lhls caLegory compared wlLh [usL
29° of Lhose wlLh llLLle cloud experlence.

89#$; ?#;+9 C9+6+2% L: ?,2,>+G=$HH#-%

1he glue blndlng a channel flrm Lo lLs cusLomers ofLen hlnges on Lhe managemenL of l1 afLer Lhe lnlLlal
sale and lmplemenLaLlon. Many cusLomers need posL-sales aLLenLlon frankly, especlally smaller
buslnesses wlLh no l1 sLaff. 1hls ls one reason LhaL managed servlces ls vlewed as a musL-enLer markeL
for Loday's soluLlon provlders, noL only does lL provlde a more predlcLable sLream of recurrlng revenue,
buL also creaLes a conLracLual arrangemenL for ongolng l1 servlce and supporL. 1he managemenL and
supporL of cloud soluLlons ls no dlfferenL and offers a compelllng buslness model for Lhe channel, as
deflned below:

1yplcolly tbey ote Jellvetloq tbe ooqoloq moooqemeot ooJ soppott of clooJ-boseJ setvlces os ptoject
wotk ot lo o coottoctool, tecottloq teveooe moJel. 1bey ote olso oJJloq, scolloq ot ttooblesbootloq clooJ
setvlces os oeeJeJ.

ManagemenL of cloud soluLlons lncludes Lhe baslc servlces, such as LroubleshooLlng or repalrlng of
cloud-based l1 on a pro[ecL basls, whlch a Lhlrd of channel flrms ln Lhls model dld lasL year. AnoLher 6 ln
10 are conducLlng remoLe monlLorlng of cloud soluLlons for cusLomers and/or managlng soluLlons LhaL
reslde ln a mulLlcloud envlronmenL. MulLlcloud managemenL ls a solld opporLunlLy area for Lhe channel
as myrlad cloud apps and oLher soluLlons mushroom ln Lhe markeL. CusLomers are accesslng soluLlons
from an array of dlfferenL provlders and daLa cenLer locaLlons, ofLen wlLh llLLle real handle on Lhe source
of Lhese servlces. Channel flrms are opLlmally poslLloned Lo serve as a gaLekeeper. Along a slmllar veln,
cloud-Lo-cloud mlgraLlon ls becomlng more prevalenL as cusLomers experlmenL wlLh dlfferenL cloud
provlders Lo gauge performance levels and saLlsfacLlon raLes. !usL more Lhan half of channel flrms are
engaged ln Lhls Lype of acLlvlLy Loday.

Llkewlse, Lhe channel ls wlsely developlng ways Lo demonsLraLe cloud 8Cl Lo cusLomers. ln facL, 6 ln 10
channel flrms lnvolved ln Manage/SupporL have creaLed l1 dashboards LhaL allow cusLomers Lo Lrack
Lhelr cloud uLlllzaLlon, cosLs and oLher meLrlcs Lo undersLand Lhelr lnvesLmenL. SevenLy-Lwo percenL of
medlum-slzed soluLlon provlders are bulldlng Lhese Lypes of dashboards, a clear value-add.

AnoLher reason LhaL Lhe Manage/SupporL model ls llkely Lo flourlsh ls Lhe facL LhaL whlle cloud ofLen
absLracLs complexlLy from Lhe cusLomer's vlew, Lhe soluLlons Lyplcally enable more connecLlons, more
end polnLs and rellance on neLworks. 1hese are all areas needlng ongolng supporL.

Aooool 1teoJs lo clooJ compotloq: SecLlon 2 10

Channel ÞarLner Cloud CrowLh LxpecLauons

23° - 49°
30° - 73°
Crow modesLly
Crow slgnlñcanLly
!"#$"%&'(" *+ #","%-" +#*.
$/*-0 1'/"1 2% /'1& 34 .*%&51
'.*%( $5'%%"/ 6'#&%"#1
7/*-0 #","%-" (#*8&5
"96"$&':*%1 '.*%( $5'%%"/
6'#&%"#1 *,"# %"9& 34 .*%&51
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
F OU R T H ANNUAL • AU GU S T 2 0 1 3

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 2
/+0 1#(2%)

• 1wo Lhlrds (63°) of channel flrms characLerlze cusLomer demand for cloud-based l1 soluLlons and
servlces as elLher very hlgh or hlgh, wlLh anoLher 3 ln 10 descrlblng demand as somewhaL hlgh. !usL
9° sald demand for Lhese soluLlons was low.

• Slx ln 10 channel flrms say LhaL cloud has generally sLrengLhened Lhelr cusLomer relaLlonshlps, wlLh
[usL 13° clalmlng lL has weakened Lhem and roughly a quarLer LhaL sald LhaL Lhelr cllenL bonds have
remalned Lhe same. 1haL sald, noL all channel flrms LhaL adopL cloud wlll engender more good wlll
wlLh cusLomers, some may slmply have a cusLomer seL LhaL ls noL cloud-frlendly, oLhers may noL
galn sufflclenL experLlse Lo provlde value.

• lour ln 10 channel flrms sald Lhey experlenced cases where cusLomer demand for cloud soluLlons
ouLsLrlpped Lhelr capaclLy Lo dellver, whlle anoLher 20° losL a deal because a cusLomer deslred a
cloud soluLlon Lhey dld noL offer.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 3
34#$5 6+7,25 ,25 %'+ 3$)%#7+- *+4,%(#2)'(8

CemenLlng a successful cusLomer relaLlonshlp requlres channel flrms Lo possess a keen undersLandlng of
Lhelr cusLomer's buslness goals, needs and paln polnLs, and also have Lhe ablllLy Lo apply LhaL knowledge
Lo Lhe soluLlons recommended and sold. 1hls baslc buslness reallLy ls no dlfferenL wheLher you are
selllng LradlLlonal on-premlse hardware and sofLware or cloud-based soluLlons. ln facL, Lhe opLlons LhaL
cloud provldes have only upped Lhe anLe when lL comes Lo channel buslness relaLlonshlps wlLh all of
Lhelr consLlLuenLs: cusLomers, vendors and dlsLrlbuLors. ln Lhls secLlon, we wlll explore Lhe lmpacL cloud
has had on all Lhree afflllaLlons.

llrsL leL's Lake a look aL cusLomer demand for cloud, whlch ln many ways drlves channel behavlor.
Conslder Lhe followlng:

Cloud demand acceleraLlon exLends Lo Lhe channel as well, accordlng Lo Lhe Comp1lA sLudy. 1wo Lhlrds
(63°) of channel flrms characLerlze cusLomer demand for cloud-based l1 soluLlons and servlces as elLher
very hlgh or hlgh, wlLh anoLher 3 ln 10 descrlblng demand as somewhaL hlgh. !usL 9° sald demand for
Lhese soluLlons was low.

Among channel flrm Lypes, resellers ln Lhe sLudy clLed Lhe hlghesL demand. SevenLy-four percenL of
Lhese companles sald demand was elLher very hlgh or hlgh, compared wlLh 62° of servlces-orlenLed
channel flrms LhaL dld so. Cne explanaLlon ls LhaL resellers as a group Lend Lo be lnvesLed heavlly ln
selllng vendor-based cloud soluLlons, ln parLlcular SaaS appllcaLlons such as MlcrosofL Cfflce 363 and
Coogle Apps. 1hls ls an acLlvlLy LhaL flLs lnLo Lhe cloud buslness model framework called
Þrovlde/Þrovlslon, and across all frameworks ln Lhe Comp1lA sLudy Lhe Þrovlde/Þrovlslon cloud model
was predlcLed Lo experlence Lhe greaLesL growLh poLenLlal for Lhe channel ln Lhe nexL Lwo years. (5,,
5,6"/1& 7 18 "#/. +,31+" 81+ 21+, -,"(/). 1& "#, 9(+/1'. 6)1'- :'./&,.. 21-,) 8+(2,;1+ks)

lndeed, channel flrms reporLed hlgh demand for cloud servlces ln general. Conslderlng Lhe four buslness
models ouLllned ln Lhls reporL, l1 companles wlLh a 8ulld pracLlce reporL Lhe hlghesL levels of cusLomer
lnLeresL ln cloud soluLlons. Þresumably, Lhese are cusLomers engaglng ln 8ulld pro[ecLs, such as
deploylng a prlvaLe cloud, buL slnce many channel parLners are lnvolved ln mulLlple cloud offerlngs, lL
could be demand sLemmlng from a range of cusLomer cloud needs.
AddlLlonally, Lhe demand levels for each dlsLlncL buslness model also map Lo Lhe rank order ln whlch
many channel flrms enLered Lhese four spaces. MosL of Lhe companles LhaL are offerlng Lwo or more
Lypes of cloud frameworks sLarLed wlLh Lhe 8ulld caLegory, Lhen moved along addlng Þrovlde/Þrovlslon,
<(+"&,+ 81+,6(.". "#, 3':)/6 6)1'- .,+9/6,. 2(+=," "1 4+1; >?@A 3,+6,&" /& 7B>C "1 "1"() D>C>
:/))/1& ;1+)-;/-,E '3 8+12 D>>> :/))/1& /& 7B>7@

FG0 +,.,(+6# .#1;. "#(" ;1+)-;/-, .3,&-/&4 1& 3':)/6 F* 6)1'- .,+9/6,. /. ,H3,6",- "1
(33+1(6# D>BB :/))/1& /& 7B>I@ J9,+ "#, 7B>7K7B>I 81+,6(." 3,+/1-E 3':)/6 F* 6)1'- .,+9/6,.
;/)) ,&L1M ( 61231'&- (&&'() 4+1;"# +(", 18 7I@!NE 8/9, "/2,. "#(" 18 "#, F* /&-'."+M 19,+())@E
(&- :M 7B>AE 1&, 18 ,9,+M .,9,& -1))(+. .3,&" 1& 3(6=(4,- .18";(+,E .,+9,+E (&- ."1+(4,
188,+/&4. ;/)) :, "#+1'4# "#, 3':)/6 6)1'- 21-,)@

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 4
Lnable/lnLegraLe and flnally Manage/SupporL. lL's unsurprlslng Lhen LhaL demand ls reporLed Lo be
hlghesL ln Lhe flrsL Lwo caLegorles, as Lhey are Lhe mosL prevalenLly pracLlced across Lhe channel Loday.
LxpecL cusLomer demand for Lhe managemenL and supporL of cloud servlces - Lhe fourLh buslness
model caLegory -- Lo grow ln comlng years as more end users mlgraLe from Lhe purchase of cloud
soluLlons Lo Lhe need Lo Lend and monlLor Lhe servlces over Lhelr llfecycle.

WlLh cusLomers curlous abouL cloud and demand up, fresh conversaLlons have been sparked beLween
channel flrms and Lhelr exlsLlng cllenLs, whlle doors are openlng Lo poLenLlal new cusLomers seeklng, ln
some lnsLances, a beLLer mouseLrap Lo meeL Lhelr Lechnology needs.

Case ln polnL? Slx ln 10 channel flrms say LhaL cloud has generally sLrengLhened Lhelr cusLomer
relaLlonshlps, wlLh [usL 13° clalmlng lL has weakened Lhem and roughly a quarLer LhaL sald LhaL Lhelr
cllenL bonds have remalned Lhe same. 1hls ls encouraglng news glven Lhe facL LhaL many ln Lhe channel
have feared publlcly LhaL cloud would drlve a wedge beLween Lhem and Lhelr cusLomers. 1here's been
rampanL apprehenslon abouL such lll effecLs as a resurgence ln vendor dlrecL sales and end user
cusLomers chooslng a self-servlce model for Lhelr l1 soluLlons, l.e. procurlng SaaS appllcaLlons over Lhe
lnLerneL. And whlle boLh of Lhese Lrends are happenlng Lo a cerLaln exLenL, Comp1lA daLa suggesL noL aL
such dlre expense Lo mosL of Lhe channel, especlally Lhose LhaL have reached a hlgh level of cloud
maLurlLy Loday and lnLend Lo remaln commlLLed. 1haL sald, noL all channel flrms LhaL adopL cloud wlll
engender more good wlll wlLh cusLomers, some may slmply have a cusLomer seL LhaL ls noL cloud-
frlendly, oLhers may noL galn sufflclenL experLlse Lo provlde value, eLc.

CusLomer uemand for Cloud Soluuons
!"#$%&'( *'&+,- .%( !/%"- 0%/"1%,#
! CusLomer deslre Lo reduce complexlLy
! CusLomer deslre Lo lncrease moblle/
remoLe access Lo company daLa
! CusLomer eyelng cosL reducuons or
Lhey are prlce sensluve
! CusLomer has buslness ob[ecuves besL
me wlLh cloud soluuons
! AblllLy Lo on-ramp and scale new
feaLures/servlces fasLer
! CusLomer has no ln-house l1
2+3, 4'+#%,# $% 4'5%&&',-
!/%"- %6'( 7,89('&3#' 0%/"1%,#
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
very Plgh
:;< =>?
Plgh or Low
4+1,A %. *'&+,- .(%&
!B+,,'/ C+($,'( C'(#9'516'
Channel parLners
wlLh a cloud 6'/)-
pracuce or a
pracuce are mosL
bulllsh on cusLomer
demand for cloud

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 3
ln facL, channel flrms can play a crlLlcal role ln deLermlnlng when cloud versus on-premlse works besL for
Lhelr cusLomers. 1he savvy 'LrusLed advlsor' should be movlng far beyond a producL speeds-and-feeds
sales plLch Lo a buslness-orlenLed consulLlng approach Lo Lhe engagemenL. 8espondenLs Lo Lhe Comp1lA
sLudy sald Lhe prlmary reasons LhaL Lhey are recommendlng cloud Lyplcally map Lo a cusLomer's
buslness prlorlLles. 1hey also fall ln llne wlLh macro reallLles such as Lhe growLh of Lhe remoLe work
force. As one example, Lhe percenLage of channel flrms LhaL sald Lhelr cusLomers' deslre Lo lncrease
moblle/remoLe access Lo company daLa sparked Lhe reason Lo recommend cloud [umped Lo 46° of
respondenLs compared wlLh 38° lasL year. 1hls underscores Lhe surge ln moblllLy soluLlons, as well as
8?Cu and LelecommuLlng Lrends LhaL are happenlng ln Lhe markeLplace. lL also demonsLraLes Lhe
channel's ablllLy Lo Lle Lhe value of cloud lnLo Lhese burgeonlng areas.

A cusLomer's deslre Lo reduce complexlLy across Lhelr l1 envlronmenL remalns Lhe no. 1 reason LhaL
soluLlon provlders recommend cloud offerlngs ln place of on-premlse. Slx ln 10 medlum-slzed channel
flrms clLed Lhls as Lhe prlmary cloud recommendaLlon drlver. Cver Lhe nexL several years, more proof
polnLs wlll emerge on 1CC and Lrue cosL savlngs enabled by cloud, however. 8uL Loday, cusLomers
lncreaslngly are open Lo explorlng cloud soluLlons ln lleu of spendlng on new hardware, packaged
sofLware and oLher ln-house lnfrasLrucLure. 1he challenge for Lhe channel ls golng Lo be demonsLraLlng
Lhose cosL savlngs over Llme Lo Lhe cusLomers whlle wardlng off compeLlLlon from glanLs llke Amazon,
whose cloud prlclng has conLlnued Lo drop quarLer over quarLer.

AnoLher challenge lles ln Lhe ablllLy Lo dellver cloud servlces ln locksLep wlLh demand. lL sounds llke a
problem you shouldn'L mlnd havlng: Loo much demand for your servlces. 8uL unless a channel flrm has
Cloud and Lhe CusLomer 8elauonshlp
!"# %&"'( )*+
,-.*/01( %'+0"-12 341+
CusLomer demand ouLsLrlpped
capaclLy Lo dellver cloud soluuons
LosL cloud sale Lo vendor, dlsLy or
oLher non-soluuon provlder
LosL a sale because cusLomer
asked for soluuon we don'L oñer
819*:;1 <*&1+ ='0/"-1+
#40) %'+0"-12+
no change
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
Channel companles wlLh a hlgh degree of cloud
maLurlLy and lnvolvemenL are 21+, )/6,)7 Lo
reporL some Lype of negauve cusLomer ouLcome
Lhan channel parLners wlLh lower levels of cloud

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 6
adequaLe capaclLy and skllls Lo meeL LhaL demand, Lhey wlll have Lo elLher Lurn away or lose buslness. ln
facL, 4 ln 10 channel flrms sald Lhey experlenced cases where cusLomer demand for cloud soluLlons
ouLsLrlpped Lhelr capaclLy Lo dellver, whlle 20° losL a deal because a cusLomer deslred a cloud soluLlon
Lhey dld noL offer. AnoLher negaLlve sales ouLcome lnvolved loslng a cloud deal Lo anoLher soluLlon
provlder, vendor or dlsLrlbuLor, whlch a Lhlrd of respondenLs sald occurred lasL year. 1hls laLLer lssue ls
noL conflned Lo cloud obvlously, and happens wlLh regularlLy ln Lhe on-premlse sales arena as well.

lnLeresLlngly, channel flrms wlLh hlgh levels of maLurlLy wlLh cloud and Lhose LhaL are lnvolved ln four of
Lhe ouLllned cloud buslness models were more llkely Lo experlence capaclLy or lnablllLy Lo dellver lssues
around cloud. More Lhan half of Lhese Lypes of companles, for example, experlenced demand for cloud
soluLlons beyond whaL Lhey were able Lo execuLe on versus an average around 4 ln 10 of all oLher
caLegorles of channel flrms. 1hls ls mosL llkely explalned by Lhe facL LhaL Lhey are so heavlly lnvolved ln
cloud - cloud experLs, as lL were - LhaL demand naLurally gravlLaLes Lhelr way, and ln some cases ls
beyond Lhelr means.

1he emphasls on demand should noL overshadow Lhe facL LhaL noL all end user companles are buylng
lnLo cloud yeL. 1oday's channel flrms conLlnue Lo encounLer some dlfflculLy selllng Lo cerLaln cusLomers.
A perennlal concern ls securlLy warlness, whlch conLlnues Lo rank as Lhe chlef roadblock Lo cloud
accepLance based on whaL channel companles of all slzes say Lhey are hearlng from cusLomers. Cn
average, more Lhan half of respondenLs clLed securlLy as Lhe chlef hurdle Lo a cloud sale. Concern abouL
lnLerneL downLlme moved up Lhe llsL of cusLomer ob[ecLlons Lo cloud Lhls year, whlch may be Lhe resulL
of Lhe pasL year's very publlc ouLages. 1here are many examples, some lnvolvlng flnanclal lnsLlLuLlons as
well as one LhaL Look down porLlons of MlcrosofL's CuLlook servlces for several days or a more recenL,
albelL very brlef, ouLage aL Coogle.

9#8 3$)%#7+- :";+.%(#2) %# !5#8%(2< 34#$5

=>? SecurlLy concerns
@@? lnLerneL downLlme
@A? lnLegraLlon concerns wlLh on-premlse lnfrasLrucLure
BC? uaLa porLablllLy lssues
BD? 1oLal cosL of ownershlp concerns
B>? Lack of undersLandlng of opLlons/Lradeoffs of cloud
BE? 8eslsLance Lo change by lnLernal l1

1wo of Lhe maln cusLomer ob[ecLlons acLually pose poLenLlal opporLunlLles for channel flrms lf Lhey are
handled well. lnLegraLlon concerns abouL Lylng cloud lnLo exlsLlng lnfrasLrucLure and worrles abouL daLa
porLablllLy need noL be deal breakers - lnsLead Lhey provlde a chance for Lhe soluLlon provlder Lo flex
Lhelr value, knowledge and sklll seL. 8elng able Lo explaln Lo a poLenLlal cusLomer ln deLall whlch parLy
ulLlmaLely ºowns" daLa placed ln Lhe cloud, parLlcularly ln a slLuaLlon where a cloud provlder mlghL go
ouL of buslness or Lhe cusLomer falls behlnd on paymenLs, demonsLraLes Lhe channel flrm's knowledge
of cloud-based models. uaLa porLablllLy moves from a sales obsLacle Lo overcome Lo a value-added
servlce Lo sell. Llkewlse wlLh lnLegraLlon. lor channel flrms selllng cloud Loday, Lhe greaLesL source of
revenue afLer Lhe sale lles ln lnLegraLlon work - cloud Lo on-premlse and cloud Lo cloud. A proven Lrack
record here wlLh exlsLlng cusLomers can serve as a blueprlnL or proof polnL Lo persuade more relucLanL
cusLomer prospecLs, much llke case sLudles are used.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 7
1oLal cosL of ownershlp ls creaLlng some doubL ln Lhe mlnds of cusLomers lnlLlally looklng aL cloud as a
cosL-reducer of l1. Whlle Lhe barrler Lo enLry for cloud - dependlng on Lhe soluLlon - wlll ofLen appear
far more affordable Lhan Lhe expense for new l1 equlpmenL or sofLware on-premlse, Lhe longer-Lerm
cosLs assoclaLed wlLh cloud soluLlons and Lhelr managemenL are noL enLlrely LransparenL.

F+25#-) ,25 6()%-("$%#-)

CusLomers aren'L Lhe only relaLlonshlp LhaL channel flrms need Lo reconslder afLer dlvlng lnLo cloud.
vendors and dlsLrlbuLors, key elemenLs Lo Lhe l1 supply chaln and fundamenLal buslness parLners Lo Lhe
channel, musL also be consldered.

WlLh respecL Lo vendors, channel flrms are evaluaLlng Lhem on Lhelr cloud sLraLegles, lncludlng such
conslderaLlons as a vendor's cloud offerlngs, buslness model, rouLes Lo markeL (dlrecL or lndlrecL) and
compensaLlon plan for parLners selllng cloud. 1hls evaluaLlon could resulL ln conLlnued allgnmenL wlLh
exlsLlng vendors LhaL have added cloud, or lL could mean a swlLch Lo vendors elLher more lnvesLed ln
cloud or more channel-frlendly ln how Lhey dellver cloud offerlngs.

Cn whaL basls are channel companles chooslng cloud vendors Lo work wlLh? LasL year, vendor brand
repuLaLlon was Lhe maln reason, desplLe predlcLlons LhaL Lhe cloud model wlll erase Lhe lmporLance of
vendor brandlng ln Lhe sales process because servlces are lnLanglble and lnvlslble Lo Lhe cusLomer. noL
so Lhls year. 1he chlef reason Lo selecL a vendor Lo work wlLh relaLed Lo cloud ls cusLomer preference.
lacLors aL Þlay ln Chooslng Cloud vendor ÞarLners
8evenue/comp model provlded
CommlLmenL Lo cloud
SupporLs all four cloud buslnes models
vendor brand repuLauon
1ype/varleLy of cloud servlces oñered
CusLomer preference/demand
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs
lrom perspecuve of channel parLners wlLh a cloud pracuce

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 8
1hls a hlghly Lelllng flndlng, demonsLraLlng [usL how much leverage cusLomers Loday have ln Lhe l1
markeLplace, even wlLh someLhlng as new as cloud. WhaL's unclear ls wheLher cusLomers are
demandlng Lo work wlLh a parLlcular cloud vendor or lf Lhey are zerolng ln on a speclflc Lype of cloud
servlce LhaL ln Lurn drlves Lhe channel parLner Lo seek ouL a vendor maLch.

8egardless of wheLher Lhe cusLomer ls demandlng lL or noL, channel flrms selecLlng a vendor Lo work
wlLh are also lnLeresLed ln Lhe vendor goods. llndlngs show LhaL Lhey place greaL emphasls on Lhe Lype
and breadLh of cloud offerlngs LhaL Lhe vendor has ln lLs qulver, especlally among Lhose wlLh a sLeep
lnvesLmenL ln Lhelr own cloud pracLlce. Slx ln 10 flrms LhaL have hlgh levels of cloud maLurlLy, hlgh cloud
proflL marglns or parLake ln all four cloud buslness models sald a vendor's cloud porLfollo ls paramounL
ln Lhe selecLlon process. Llkewlse, Lhose channel flrms wlLh sLrong cloud lnvesLmenLs also value Lhe
vendor's level of commlLmenL Lo Lhe model as selecLlon crlLerla. More Lhan half of Lhese Lypes of flrms
versus, for example, 2 ln 10 low cloud maLurlLy companles, say Lhey are keen on allgnlng wlLh vendors
LhaL Lake cloud and Lhelr channel's lnvolvemenL ln lL serlously.

Companles wlLh Lhe deepesL cloud experlence and lnvolvemenL were also Lhe mosL llkely Lo have
experlenced ma[or change ln Lhe makeup of Lhe vendors Lhey work wlLh compared wlLh Lhose wlLh less
cloud experlence. 1hls could slgnal Lhe facL LhaL Lhose flrms have moved away from some of Lhe more
LradlLlonal l1 manufacLurers Lo [oln forces wlLh some of Lhe newer, cloud-focused vendors LhaL have
emerged on Lhe scene.

Cloud lmpacL on ulsLrlbuuon
!"#$$%& ()*+ ,-.%/0#12$3 42* 5)30*)6702*
8%&#12$3"). 29%* :%-0 ; <%#*3
ulsLrlbuLors are servlng as an
aggregaLor of cloud servlces LhaL
soluuon provlder can Lhen resell
ulsLrlbuLors sull ñgurlng ouL where
Lhey ñL ln Lhe cloud world
ulsLrlbuLors playlng a hosung role
for cloud servlces ln Lhelr daLa
ulsLrlbuLors are compeung agalnsL
soluuon provlders wlLh Lhelr cloud
!"#$$%& @#*0$%* @%*/%.12$3 24 82&% 24
5)30*)6702*3 )$ # !&27A B2*&A
uon'L know
LxpecL decreaslng
no change
LxpecL lncreaslng
lnvolvemenL wlLh
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 400 u.S. l1 channel ñrms wlLh cloud oñerlngs

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 9

MeanLlme, Lhe oLher plece of Lhe value chaln Lo conslder ls dlsLrlbuLlon. MosL channel flrms have well-
esLabllshed relaLlonshlps wlLh broadllne dlsLrlbuLors such as 1ech uaLa, lngram Mlcro and Synnex, as
well as worklng wlLh value-added dlsLrlbuLors such as u&P. !usL as channel flrms have wondered whaL
role Lhey would have ln a cloud-based world, dlsLrlbuLors - long Lhe fulflllmenL arms for Lhe channel ln
Lerms of producLs - have had Lhelr share of angsL abouL where Lhey flL ln.

8ased on whaL Lhe channel has Lo say, dlsLrlbuLors wlll assume [usL as cenLral of a role ln Lhe dellvery
and packaglng of cloud servlces as Lhey do ln Lhe producL-based world. ln facL, half of channel flrms
expecL Lo see lncreased lnvolvemenL wlLh dlsLrlbuLors as a resulL of offerlng cloud, whlle only 7° plan Lo
decrease Lhelr alllances. Where mosL channel flrms see value comlng from dlsLrlbuLlon ls ln Lhelr ablllLy
Lo serve as aggregaLors of cloud servlces, effecLlvely packaglng up offerlngs and soluLlons from varlous
vendors ln an easlly consumable way for soluLlon provlders Lo resell. lL's a loglcal exLenslon of whaL
dlsLrlbuLlon already does for producL-relaLed soluLlons.

1haL sald, Lhe poslLlve daLa mlghL belle some poLenLlal reallLles around dlsLrlbuLlon and Lhe cloud. llrsL
off, channel flrms could be hedglng Lhelr beLs ln predlcLlng a sLronger engagemenL wlLh dlsLrlbuLors
broughL on by cloud. lL doesn'L cosL soluLlon provlders much Lo work wlLh a dlsLrlbuLor, so lL's noL a
sLreLch for Lhem Lo wax poslLlvely abouL Lhe fuLure poLenLlal around cloud and Lhese alllances. 8uL from
Lhe dlsLrlbuLor perspecLlve, success ln cloud ls golng Lo requlre some reenglneerlng of operaLlons and
flnanclal lnvesLmenL - llkely slgnlflcanL. Pow nlmbly and successfully dlsLrlbuLors execuLe on Lhls
LranslLlon ls whaL wlll ulLlmaLely deLermlne how well Lhelr channel parLners vlew Lhem. And wlll Lhelr
lnvesLmenL ln cloud dellver enough revenue Lo offseL losses ln more esLabllshed parLs of Lhelr buslness?

F OU R T H ANNUAL • AU GU S T 2 0 1 3

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 2
/+0 1#(2%)

• As end users maLure ln Lhelr adopLlon of cloud Lechnologles, Lhey wlll progress Lhrough phases of
lncreaslng compeLence and complexlLy. 8eglnnlng wlLh an LxperlmenL phase where companles wlll
learn abouL cloud concepLs and bulld proof-of-concepL models, flrms wlll pass Lhrough sLages of
non-crlLlcal use and lull ÞroducLlon whlle almlng Lo have fully 1ransformed l1 pracLlces.

• ln movlng Lowards 1ransformed l1, buslnesses wlll experlence many changes Lo Lhelr l1 operaLlons.
Þrlmary examples lnclude changes Lo pollcles and procedures regardlng Lechnology (whlch requlre
some level of enforcemenL Lo be effecLlve) and changes Lo Lhe l1 deparLmenL (whlch may lnvolve
reLralnlng or hlrlng for speclflc skllls).

• Surprlslngly, Lhe number of companles LhaL conLracLed wlLh ouLslde flrms for cloud work dropped
from 21° ln 2012 Lo 11° ln 2013. Whlle Lhls ls parLly due Lo large companles bulldlng Lhelr own
capablllLles ln-house, channel flrms should also be alerL Lo ways of descrlblng Lhe value of Lhelr
cloud servlces and bulldlng capaclLy for lncreased cloud demand.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 3
324 5)+- 1-#6-+))(#2

WlLh aL leasL flve years of cloud adopLlon hlsLory, dlsLlncL phases are beglnnlng Lo emerge as buslnesses
become more comforLable wlLh Lhe cloud model. ln each phase, Lhere are cerLaln concepLs LhaL musL be
learned or acLlvlLles LhaL musL be masLered ln order Lo galn Lhe full beneflLs avallable Lhrough cloud

ln Lhe 378+-(9+2% phase, companles are slmply bulldlng famlllarlLy wlLh Lhe cloud model, lncludlng
Lermlnology and baslc worklng prlnclples. Companles LhaL were early cloud adopLers passed Lhrough
Lhls phase several years ago, buL lL ls lmporLanL Lo remember LhaL Lhose companles represenL a small
percenLage of overall buslness adopLlon, and many flrms are sLlll enLerlng Lhls sLage as mass adopLlon
Lakes place.

uurlng Lhls sLage, a company may LesL ouL cloud sysLems by bulldlng sample vlrLual lnsLances or uslng
sofLware on a free Lrlal basls. 1hese proof-of-concepL underLaklngs wlll mosL ofLen be performed on
publlc cloud sysLems, slnce Lhey are readlly avallable and requlre mlnlmal lnvesLmenL. Companles may
lnvesLlgaLe Lhe pros and cons of prlvaLe clouds, buL very few wlll begln bulldlng ouL Lhose sysLems aL Lhls

!""# % &""# % #""#
&'()"*+ % ,-.'(/ 0123/4
&"-5+6* 72/+14
8/39":26 ;'"64(:26
=2'>?2@ 0A"6<+
&'22E 2E 9269+F*
0123/ G/2F:26 &'2<'+44(26 72/+1 E2' 86/ H4+'4
&3.1(9 0123/4
&'2)(/+' #@(*9A(6<

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 4
nexL, companles wlll move lnLo a :#2;.-(%(.,< 5)+ sLage. Pere, cloud sysLems wlll acLually be used for
operaLlonal workflow, Lhough buslnesses wlll noL choose Lo mlgraLe Lhelr mosL lmporLanL sysLems or
mosL senslLlve daLa. 1yplcally, a perlpheral sysLem wlll be chosen or LesLlng and developmenL work wlll
be moved, wlLh Lhe full producLlon sysLems sLlll malnLalned ln-house.

8y movlng an operaLlonal sysLem or appllcaLlon Lo Lhe cloud, companles can learn Lhe flne deLalls
behlnd a cloud LranslLlon and also galn a flrsL-hand appreclaLlon for Lhe lnLegraLlon challenges. 1hrough
regular use, Lhe Lrue cosLs of a cloud soluLlon wlll become apparenL, and companles may flnd areas of
unexpecLed expendlLure, such as neLwork upgrades. 1hose buslnesses LhaL plan on pursulng prlvaLe
clouds may begln bulldlng Lhose aL Lhe end of Lhls sLage ln order Lo galn experlence wlLh managlng cloud

Cnce a cerLaln level of comforL wlLh Lhe
cloud model has been achleved,
companles wlll move lnLo =$<< 1-#4$.%(#2.
8uslnesses aL Lhls sLage have undersLood
and mlLlgaLed Lhelr securlLy concerns and
vlew cloud sysLems as a vlable opLlon for l1
operaLlons. Þollcles and procedures wlll be
bullL and modlfled as flrms change Lhe way
Lhey procure and uLlllze Lechnology.

AnoLher acLlvlLy LhaL beglns Laklng place ln
Lhls sLage ls cloud provlder swlLchlng. lor a
varleLy of reasons, companles may move
from one publlc cloud provlder Lo anoLher,
move from a publlc cloud lnLo a prlvaLe
cloud, or even move from a publlc cloud
back Lo an on-premlse sysLem. As more
organlzaLlons begln Lo employ Lhls
ºmulLlcloud" approach, Lhere wlll be new
areas of opporLunlLy. SecLlon 3 of Lhls
reporL examlnes mlgraLlons ln more deLall.

Many companles wlLh legacy l1
archlLecLure are currenLly ln Lhe full
producLlon phase, wlLh very few havlng
moved Lo a model of >-,2)?#-9+4 @>.
1hose companles LhaL are born ln Lhe
cloud, havlng compleLely bullL Lhelr
buslness sysLems around cloud soluLlons,
are Lhe prlmary resldenLs of Lhls caLegory
aL Lhls polnL ln Llme. Pere, companles are
noL slmply movlng exlsLlng sysLems or
appllcaLlons lnLo Lhe cloud, Lhey are
changlng Lhe way Lhey work ln order Lo reap Lhe
full beneflL.

A8%(#2) ?#- 1-(B,%+ C<#$4 C#2)%-$.%(#2
As lnLeresL grows ln bulldlng prlvaLe clouds as parL of an
overall l1 lnfrasLrucLure, Lhere are several opLlons LhaL can be
consldered for Lhe underlylng Lechnology. 1hese sofLware
packages provlde Lhe mechanlsms for Lurnlng hardware lnLo
cloud componenLs, lncludlng vlrLuallzaLlon supporL, AÞls for
end users, and fabrlc conLrollers LhaL manage aspecLs such as
dynamlc allocaLlon. 1hese examples demonsLraLe Lhe range
of opLlons avallable.
• A8+2D%,.EF Crlglnally a collaboraLlon beLween nASA
and 8ackspace, Lhls open source sofLware has
backlng from companles such as PÞ, l8M, and A1&1,
who have all sLaLed LhaL Lhey wlll use Lhls plaLform
for Lhelr prlvaLe cloud offerlngs.
• C<#$4D%,.EF ClLrlx acqulred Lhls Lechnology as parL
of Lhelr 2011 purchase of and released lL
as an open source pro[ecL ln 2012. 1hanks Lo Lhe
prevlous user base, CloudSLack clalms LhaL $1 bllllon
worLh of buslness LransacLlons Lake place annually
on Lhelr plaLforms.
• 3$.,<08%$)F Whlle CpenSLack and CloudSLack boLh
provlde some level of compaLlblllLy wlLh Amazon
AÞls Lhrough LranslaLlon mechanlsms, LucalypLus
has llcensed Lhe AÞls dlrecLly. 1hls provldes Lhe mosL
dlrecL llnk beLween a prlvaLe cloud and Amazon
Web Servlces, Lhough lL also lncreases Lhe rlsk of
vendor lock-ln.
• BC<#$4F 1hls offerlng from vMware may hold appeal
for cusLomers who have heavlly vlrLuallzed uslng
vMware Lools, buL Lhey may flnd LhaL Lhere ls a
Lradeoff ln seamless lnLeroperablllLy wlLh a varleLy
of oLher cloud vendors.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 3
1he key Lo 1ransformed l1 ls opLlmlzaLlon. 8aLher Lhan uslng Lhe same workflow and lncorporaLlng
cloud-based sysLems, workflows wlll be alLered Lo Lake advanLage of unlque cloud characLerlsLlcs.
Slmllarly, exlsLlng appllcaLlons LhaL do noL have bullL-ln monlLorlng or securlLy feaLures wlll be rebullL or
replaced wlLh sofLware LhaL ls more cloud-enabled. 1hls ls noL Lo say LhaL a company wlll have Lhelr
enLlre l1 archlLecLure ln Lhe cloud, 1ransformed l1 wlll usually sLlll employ on-premlse models along wlLh
publlc cloud and prlvaLe cloud opLlons, and workflows may be lndlvldually opLlmlzed as a company
moves Lowards a hollsLlc LransformaLlon.

1he LranslLlon Lo 1ransformed l1 ls challenglng, and Lhe converslon may Lake place over several years
slnce many companles are reallzlng beneflLs [usL by reachlng Lhe lull ÞroducLlon sLage. As compeLlLlon
lncreases, Lhough, buslnesses wlll need Lo clalm any advanLage Lhey can galn and wlll become more
lnLeresLed ln Lhe lnvesLmenL requlred Lo fully uLlllze cloud compuLlng.

:+G C<#$4 H#4+<) I-(B+ :+G J+',B(#-

As companles progress from experlmenLaLlon wlLh cloud models Lo a full LransformaLlon of enLerprlse l1
pracLlces, Lhey wlll lncreaslngly flnd LhaL Lhere musL be changes Lo Lhe overall l1 approach. 8aLher Lhan
slmply Lransferrlng a glven l1 sysLem Lo Lhe cloud and allowlng work pracLlces Lo conLlnue as before, a
buslness can galn furLher beneflLs and more properly secure lLs asseLs by underLaklng some behavloral

lor cerLaln use cases and early sLages of adopLlon, cloud compuLlng may presenL only a sllghL change Lo
Lhe operaLlons of an l1 Leam. lor oLhers, Lhe change can be much larger, lmpacLlng noL only Lhe l1 Leam
buL also Lhe buslness processes of an enLlre company. 1hls ls Lhe reason for Lhe momenLum behlnd l1
becomlng more of a parLner Lo Lhe llnes of buslness raLher Lhan slmply a suppller. 1he ablllLy Lo use
Lechnology Lhrough Lhe cloud can sLreamllne Lhe way l1 ls provlded, buL lL can also sLreamllne general
workflows. 8y comblnlng Lhorough Lechnlcal knowledge wlLh operaLlonal procedure and sLraLeglc
ob[ecLlves, l1 can be a ma[or player ln movlng a buslness forward.

1haL LransformaLlon does noL come from a buslness slmply applylng Lhe currenL knowledge and skllls of
an l1 deparLmenL Lo Lhe Lask aL hand. 1he l1 deparLmenL lLself should be carefully examlned Lo
deLermlne how well lL ls equlpped Lo undersLand and uLlllze cloud Lechnology. ln addlLlon,
organlzaLlonal pollcles and procedures may have Lo be modlfled or creaLed Lo deflne Lhe proper
lnLeracLlon wlLh cloud provlders or Lhe l1 Leam.

Small buslnesses especlally should Lake care Lo examlne Lhelr l1 and buslness operaLlons. 1wenLy-Lhree
percenL of small buslnesses (1-99 employees) uslng cloud soluLlons have noL underLaken any operaLlonal
changes aL Lhls polnL, compared Lo [usL 7° of medlum-slzed buslnesses (100-499 employees) and 7° of
large buslnesses (300+ employees). ÞarLly, Lhls ls due Lo Lhe amounL of Llme spenL ln Lhe cloud so far-
operaLlonal changes are less common among laLe adopLers, where Lhere ls a heavler concenLraLlon of
small buslnesses. lL ls also llkely LhaL many small buslnesses have less bandwldLh Lo revlew Lhelr
operaLlonal procedures. WlLh cloud compuLlng belng a less dlrecL meLhod for bulldlng and uslng l1, lL
may be advlsable for smaller companles Lo begln bulldlng more operaLlonal apLlLude ln order Lo
opLlmlze Lhelr workflow wlLh cloud sysLems.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 6

1here ls an unforLunaLe slde Lo Lhls dlsrupLlon and sLreamllnlng. Cloud compuLlng has caused some
amounL of LrepldaLlon among l1 sLaff Lhanks Lo Lhe posslblllLy of ouLsourclng compuLe resource (and Lhe
subsequenL malnLenance). nearly one ln flve companles reporLs reduclng Lhe number of l1 sLaff,
showlng LhaL Lhls may be a loglcal sLep for some buslnesses. Powever, lL should be noLed LhaL Lhls ls
near Lhe boLLom of Lhe llsL of changes a company may have underLaken, [usL as lL ls near Lhe boLLom of
Lhe llsL of beneflLs llsLed by cloud users. As Lhe [ob markeL ad[usLs Lo Lhe reallLles of cloud, proacLlve l1
workers can geL lnvolved ln a varleLy of oLher acLlons Lhelr buslness may be Laklng Lo beLLer uLlllze cloud

1#<(.(+) ,24 1-#.+4$-+)

Changes Lo pollcy or procedure are Lhe mosL common acLlon Laken by companles who have moved lnLo
a cloud envlronmenL. WheLher lL ls redeflnlng sLeps LhaL Lhe l1 Leam musL Lake as Lhey admlnlsLer l1
soluLlons or bulldlng new processes for maklng declslons on Lechnology, acLlons relaLed Lo pollcy and
procedure occupy Lhree of Lhe Lop four spoLs on Lhe llsL of changes drlven by cloud compuLlng.

lor Lhe second year, securlLy Lops Lhe llsL of lLems LhaL companles are addresslng as Lhey modlfy
pollcles. 1hrough Lhe early days of cloud adopLlon, securlLy has been commonly clLed as Lhe prlme facLor
prevenLlng companles from movlng Lo cloud soluLlons. WlLh Lhe vasL ma[orlLy of companles havlng
sLarLed some form of cloud adopLlon, lL ls apparenL LhaL companles are elLher accepLlng some level of
rlsk or flndlng ways Lo mlLlgaLe cloud securlLy concerns-aL leasL for Lhe appllcaLlons Lhey are placlng ln
l1-8elaLed Changes urlven by Cloud
ConLracLed wlLh ouLslde company
8educed number of l1 sLañ
new monlLorlng/managemenL Lools
8esLrucLured l1 deparLmenL
new pollcles for l1 declslon maklng
AdapLed monlLorlng/managemenL Lools
Changed exlsung pollcles/procedures
8ullL new pollcles/procedures
Source: !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 449 u.S. end users wlLh cloud soluuons

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 7
Lhe cloud. Powever, Lhe
dlscusslon around securlLy has
now shlfLed from addresslng
baslc concerns (whlch drlves
lnlLlal adopLlon) Lo focuslng on
flner deLalls (whlch drlves full
producLlon use).

AnoLher area frequenLly
addressed ln pollcy changes ls
Lhe proper sLorage of company
daLa. Comp1lA's sLudles on Lhe
8lg uaLa Lrend show LhaL many
companles feel LhaL Lhelr use
of daLa ls less Lhan opLlmal, ln
parL because Lhey do noL have
sLrong overarchlng daLa
managemenL pracLlces. 1he
use of cloud sysLems for sLorage and backup only compllcaLes an already complex slLuaLlon, and many
companles may beneflL from a daLa audlL and educaLlon on besL pracLlces ln managlng daLa ln a cloud

SecurlLy and daLa managemenL are examples of pollcles LhaL are broader ln naLure and apply Lo Lhe
enLlre company. 1hese may be less frequenL Lhan changes LhaL apply only Lo an l1 deparLmenL, buL
organlzaLlon-wlde changes are more llkely Lo be Lruly dlsrupLlve and may be more necessary. As llnes of
buslness galn more undersLandlng of cloud soluLlons and Lhe poLenLlal beneflLs, Lhere ls an lncreased
chance LhaL Lhey wlll use Lhelr own budgeL for procuremenL. 1he lnLenL may noL be mallclous, buL Lhe
resulL can be deLrlmenLal lf Lhey puL company daLa aL rlsk or cause a work sLoppage when Lhe soluLlon
goes down.

8ulldlng pollcles LhaL address Lhese lssues wlll be a cross-deparLmenLal efforL, buL should sLlll lnvolve
slgnlflcanL lnpuL from Lhe l1 Leam. 1he l1 Leam musL learn how Lo meeL Lhe needs of Lhe buslness wlLh
mlnlmal delay, and Lhey also musL properly undersLand and convey Lhe need Lo malnLaln a company's
rlsk posLure. 1he besL pollcles wlll represenL an agreemenL aL all levels of Lhe company on how cloud
resources wlll be procured and used. 1hls wlll noL necessarlly mean LhaL all lnformaLlon flows Lhrough
Lhe l1 Leam for approval, buL Lhe l1 Leam should be comforLable wlLh Lhelr level of awareness and
responslblllLy ln Lhe process.

Þollcles are only effecLlve lf Lhey are enforceable, and lL appears LhaL companles are movlng more
Lowards LhaL model. Compared Lo lasL year, Lhere ls an even hlgher lncldence of companles clalmlng a
hlgh or moderaLe level of enforcemenL. Many companles enforce pollcy by performlng perlodlc revlews
Lo ensure compllance, and some also use monlLorlng Lools Lo Lrack l1 acLlvlLy across Lhe company or
requlre Lhe approval process for cloud resources Lo come Lhrough l1.

!"#$%$&' )*+ !,"%&+-,&' $* ) .#"-+ /*0$,"*1&*2
!"#"$ &' ()'&*+",")-
! 9"+:*6-;

! <*&="* >-&*?7" &' +&,=?); @?-?

! A>" &' ,&B6$" @"#6+">

! 9:==&*- '&* +$&:@ >"*#6+">
! C==*&#"@ #")@&*>
C*"?> C@@*">>"@ B; <&$6+;
3"-,%&4 ."156789' !
$%%&'( *+,%-. /% 0(1&- 0123&4%5 '2-+:
;)'&4 <=< %#"-+ &*+ -'&,' 2>)2 %>)*?&+ 5"#$%:@5,"%&+-,&

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 8
@2%+-2,< @> I+8,-%9+2% C',26+)

Aslde from pollcles and procedures LhaL geL creaLed or alLered wlLhln an l1 deparLmenL, oLher changes
may Lake place LhaL allow l1 Lo be beLLer equlpped Lo deal wlLh a cloud envlronmenL. 1hese changes
may be perlpheral Lo Lhe speclflc cloud soluLlons belng puL ln place, buL Lhey sLlll have an effecL on Lhe
overall lmplemenLaLlon cosL. Companles pursulng cloud pro[ecLs should conslder wheLher Lhese acLlons
would be a worLhwhlle parL of Lhelr cloud lnvesLmenL glven Lhelr buslness ob[ecLlves.

1he flrsL area Lo conslder ls Lhe need for new or lmproved Lools Lo conLrol or monlLor a cloud soluLlon. ln
some cases, Lhls could be a slmple exLenslon. lor example, a sysLem admlnlsLraLor may use sofLware
provlded by hls vlrLuallzaLlon vendor Lo spln up new vlrLual machlnes. ln a cloud slLuaLlon, Lhls wlll llkely
be handled Lhrough a web fronL end, and Lhere may be llLLle Lralnlng requlred Lo creaLe new servers.
1here may even be generlc machlne lmages LhaL have been saved by oLher users Lo ellmlnaLe Lhe more
Ledlous sLeps lnvolved ln geLLlng a vM or lnsLance up and runnlng.

CLher Lools wlll be new addlLlons Lo Lhe l1 Loolbox. 1hlrd parLy sofLware may be needed LhaL can
provlde a comprehenslve vlew of an archlLecLure, allowlng admlnlsLraLors Lo see vlLal lnformaLlon from
boLh cloud sysLems and on-premlse sysLems ln a slngle pane of glass. 1here may be a requlremenL Lo Lle
a company's flnanclal sysLem Lo a cloud provlder Lo deLermlne usage for bllllng purposes. 1hese are
examples of Lools LhaL can be ma[or pro[ecLs on Lhelr own, drlvlng cosLs hlgher and necesslLaLlng a heavy
Llme lnvesLmenL as well.

1ool leaLures needed for Cloud
Connecuon Lo ñnance sysLem
SupporL for mulucloud envlronmenLs
Capablllues for securlLy audlLs
Capablllues for auLomauon
SupporL boLh cloud and on-premlse
Þerformance monlLorlng
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5 sLudy
8ase: 234 cloud end users who have made Lool changes

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 9
1he oLher prlmary area Lo conslder wlLhln Lhe l1 deparLmenL ls Lhe personnel. As menLloned before,
sLaff reducLlon may be an afLereffecL of a cloud LranslLlon, buL Lhe more llkely scenarlo ls LhaL companles
wlll resLrucLure Lhe l1 deparLmenL ln some way. ln facL, 30° of companles LhaL have performed l1
resLrucLurlng have hlred new workers Lo bolsLer l1 efforLs-a sllghL bump from lasL year. 1hls may be
especlally Lrue where cloud compuLlng ls used Lo add new capablllLles Lo a buslness, and a slzable of
amounL of equlpmenL remalns on-premlse.

AL an organlzaLlonal level, Lhe movemenL Lowards
LranslLlonlng Lhe l1 deparLmenL from suppller Lo
parLner seems Lo be maklng good progress. Palf of
Lhe companles ln Lhe survey say LhaL Lhe l1 Leam has
been able Lo move away from malnLenance Lasks and
become lnvolved ln more lnnovaLlve work. An even
greaLer number (37°) say LhaL Lhe l1 deparLmenL has
become more lnLegraLed wlLh oLher llnes of buslness
Lo undersLand how Lo meeL Lhelr needs Lhrough
cloud soluLlons. 1he lmpllcaLlon ls LhaL Lhe Lechnlcal
knowledge held by Lhese workers ls sLlll valuable,
and new roles are belng developed LhaL requlre new

l1 sLaff are Laklng acLlon Lo bulld Lhese new skllls. 1wo Lhlrds of companles ln Lhe survey who reporLed l1
deparLmenL resLrucLurlng also reporLed LhaL Lhelr l1 sLaff had Laken Lralnlng Lo bulld new skllls. 1hls
Lralnlng may ofLen be lnlLlaLed by Lhe employees. uaLa from Comp1lA's 5"(", 16 "#, 7* 58/)). 9(3 sLudy
shows LhaL whlle 37° of companles Lraln or reLraln exlsLlng sLaff Lo bulld new skllls, 36° of companles
had no formal process for ldenLlfylng l1 skllls gaps. A fasL-movlng fleld such as cloud compuLlng requlres
a proacLlve approach, and employees may be more adepL aL undersLandlng whlch Lechnlcal skllls Lhey
need Lo learn. 1o make sure Lhese skllls are complemenLed wlLh Lhe buslness skllls Lo drlve prlorlLles, a
company musL Lake an acLlve role ln deflnlng Lhe skllls needed and supporLlng Lralnlng.

5)+ #? A$%)(4+ C#98,2(+)

Cne of Lhe blggesL surprlses wlLh l1-relaLed changes ls LhaL only 11° of companles have conLracLed wlLh
an ouLslde flrm for Lhelr cloud lnlLlaLlves. ln 2012, 21° of companles had conLracLed wlLh ouLslde flrms,
and Lhe expecLaLlon was for Lhls number Lo rlse as more buslnesses (especlally SM8s) adopLed cloud
soluLlons and became more maLure ln cloud usage.

Cne of Lhe facLors conLrlbuLlng Lo Lhe decllne ls LhaL large companles have experlenced a slgnlflcanL
dropoff ln Lhelr use of Lhlrd parLles. Whlle Lhese enLerprlses may have lnlLlally used a consulLlng flrm or
a speclallsL Lo geL Lhelr pro[ecL sLarLed, Lhey have llkely bullL Lhe skllls lnLernally LhaL are needed Lo
malnLaln Lhelr pro[ecLs and push lnLo new areas.

SLlll, Lhe number of small companles LhaL have conLracLed wlLh an ouLslde flrm has noL rlsen
dramaLlcally, and Lhere are several reasons LhaL Lhls parLnerlng has noL Laken off. Among companles
LhaL dld parLner wlLh a soluLlon provlder, Lhe Lop reason LhaL Lhey ellmlnaLed poLenLlal candldaLes was
LhaL Lhe cosLs were Loo hlgh. AlLhough lower cosLs ls sLlll Lhe Lop drlver for cloud soluLlons, many
companles flnd LhaL parLlcular appllcaLlons do noL acLually have lower cosLs, malnly due Lo unexpecLed
>08+) #? DE(<<)K*#<+) C#98,2(+)
L,B+ !44+4 %# @> I+8,-%9+2%)
MNO Cloud archlLecL
MPO Skllls Lo bulld prlvaLe clouds
QRO ueparLmenLal llalsons
SMO lnLegraLlon speclallsLs
NPO Compllance speclallsL

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 4 10
expenses. As companles more closely examlne Lhelr LranslLlon cosL, Lhe addlLlonal overhead requlred Lo
conLracL Lhe work mlghL be enough Lo deLer Lhem compleLely from ouLsourclng.

Looklng aL channel flrms LhaL are provldlng cloud servlces, 42° say LhaL Lhey experlenced a demand for
cloud servlces LhaL ouLsLrlpped Lhelr capaclLy Lo supply a soluLlon. 1hls could be a sLaLemenL of Lechnlcal
capablllLy or slmply raw bandwldLh. ln addlLlon, 33° sald LhaL Lhey losL a sale Lo a non-LradlLlonal l1
soluLlon provlder, such as a vendor, dlsLrlbuLor, or Lelecom carrler.

1aken LogeLher, Lhese daLa polnLs suggesL LhaL Lhere are some gaps-elLher real or percelved-ln Lhe
ablllLy of soluLlon provlders Lo adequaLely supporL cloud servlces. lor SM8s, Lhe hurdles of addlLlonal
overhead cosL and sLruggles wlLh soluLlon provlder capaclLy may be Loo large Lo overcome. As a resulL,
Lhese companles may work dlrecLly wlLh cloud provlders ln many lnsLances. Lven Lhough Lhey may
recelve servlces beyond Lhe acLual producL belng procured, Lhls ls llkely noL consldered ºconLracLlng" by
Lhe end user slnce lL all comes dlrecLly from Lhe provlder.

SoluLlon provlders should be aware LhaL Lhese percepLlons may exlsL and should Lake acLlon Lo address
Lhem. lor Lhe lssue of cosL, soluLlon provlders wlll need Lo be prepared Lo fully explaln Lhe value of a
cloud soluLlon and Lhe added beneflL LhaL Lhe provlder ls able Lo dellver. 1he lssue of capaclLy may be
more challenglng. lf a cusLomer deslres a soluLlon beyond Lhe Lechnlcal capablllLy of a provlder or lf Lhe
provlder ls fully booked, hlrlng or Lralnlng can solve Lhe problem buL are noL lmmedlaLe flxes. Cne
suggesLlon for soluLlon provlders ls Lo look for parLners wlLh speclallzed experLlse, Lhen acL as a focal
polnL wlLh Lhe cusLomer Lo provlde a compleLe soluLlon.
F OU R T H ANNUAL • AU GU S T 2 0 1 3

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 2
/+0 1#(2%)

• Cloud mlgraLlons are plcklng up sLeam as companles become more famlllar wlLh Lhe cloud model
and seek Lo opLlmlze Lhelr cloud usage. A healLhy percenLage of companles are elLher movlng from
one publlc cloud provlder Lo anoLher, movlng from a publlc cloud provlder Lo Lhelr own prlvaLe
cloud, or movlng appllcaLlons back on-premlse.

• SwlLchlng publlc cloud provlders could lnvolve elLher mlgraLlon of lnfrasLrucLure or of appllcaLlons.
1here are a wlde varleLy of reasons LhaL companles swlLch provlders, lncludlng securlLy concerns
(43°), cosLs (42°), and feaLure seLs (41°).

• Movlng Lo a Lrue prlvaLe cloud ls a compllcaLed Lask lnvolvlng speclallzed skllls. Many companles
elecL Lo work wlLh an ouLslde flrm, buL bulldlng lnLernal skllls ls also a popular opLlon. llfLy-one
percenL of large companles have reLralned sLaff compared Lo 31° of small companles. lor hlrlng,
Lhe gap ls even wlder-63° of large companles have hlred new skllls compared Lo [usL 38° of small

• Cloud mlgraLlon wlll become an even greaLer focus area ln Lhe fuLure, as 44° of companles LhaL
have noL yeL performed a secondary mlgraLlon plan Lo do so ln Lhe nexL 12 monLhs. 1hese
companles wlll wanL Lo carefully examlne Lhe efforL requlred, as Lhe daLa shows LhaL Lhey may be
underesLlmaLlng Lhe challenges.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 3
3(4-,%(#2 1,%%+-2) 5(42,6 7+8 !9#:%(#2 1',)+

Cnce companles hlL a sLage where Lhey are uslng cloud sysLems as a sLandard parL of l1 archlLecLure, lL ls
naLural Lo assume LhaL Lhey wlll begln welghlng Lhe pros and cons of varlous provlders and varlous
models and conLlnually shlfL Lo achleve Lhe opLlmal mlx. up Lo Lhls polnL, Lhere has been a greaL deal of
focus on Lhe hurdles Lo cloud adopLlon (such as securlLy) and Lhe challenges ln Lhe lnlLlal LranslLlon of
funcLlonal sysLems lnLo Lhe cloud. Many companles are sLlll concerned wlLh Lhese areas, buL early
adopLers are beglnnlng Lo explore furLher lssues relaLed Lo cloud mlgraLlon.

Comp1lA's sLudy found LhaL a healLhy percenLage of companles have begun shlfLlng lnfrasLrucLure or
appllcaLlons followlng Lhelr orlglnal LranslLlon Lo Lhe cloud. Whlle Lhere are some companles LhaL have
made mulLlple moves, approxlmaLely 10°-13° of Lhe sample have only made one of Lhe moves ln Lhe
charL above, so LhaL more Lhan slx ln Len cloud users have made a secondary move of some Lype.

lL should also be noLed LhaL mlgraLlons lndlcaLe a spaLe of accepLable opLlons for l1 sysLems. 1hls has
lmpllcaLlons for Lhe fuLure approach Lo bulldlng ouL archlLecLure. Whlle Lhe Lerm ºhybrld cloud" has
been ln use slnce Lhe early days of Lhe cloud era, lL has carrled a connoLaLlon of a slngle appllcaLlon LhaL
resldes ln boLh a prlvaLe cloud and publlc cloud. Lxamples mlghL lnclude a producLlon sysLem on a
prlvaLe cloud wlLh backup ln a publlc cloud or a prlvaLe cloud appllcaLlon LhaL uses a publlc cloud for
Lnd user Cloud Mlgrauon Þauerns
Þubllc Cloud
Þrovlder #1
Þubllc Cloud
Þrovlder #2
Company A8C
27° (lnfrasLrucLure)
Source: Comp1lA's !
*+,&-. /& 0)1'- 0123'4&5 sLudy
8ase: 432 cloud end users

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 4
Colng forward, Lhe reallLy ls LhaL companles wlll have Lhelr sysLems spread across mulLlple clouds,
chooslng Lhe opLlon LhaL besL sulLs Lhelr needs for a parLlcular appllcaLlon. 1hls ls no dlfferenL Lhan a
LradlLlonal daLa cenLer wlLh servers LhaL are conflgured for dlfferenL purposes, buL Lhe managemenL
challenges are much greaLer. As Lhe lndusLry moves Loward a mulLlcloud paradlgm, Lhere wlll be ample
opporLunlLy Lo help manage Lhe growlng complexlLy.

1$"6(. ! 1$"6(.

Lach of Lhe lndlvldual mlgraLlon
sLyles has parLlcular characLerlsLlcs.
1he mosL popular Lype of move ls
from one publlc cloud provlder Lo
anoLher. Pere, Lhere can be Lwo
dlfferenL componenLs lnvolved:
lnfrasLrucLure and appllcaLlons
(lnfrasLrucLure ls lmplled ln Lhe oLher
Lwo sLyles). 1here are many reasons
LhaL cusLomers are swlLchlng cloud
provlders, each of whlch ls worLh
conslderaLlon for boLh cloud vendors
and soluLlon provlders managlng
cllenLs' cloud sysLems.

• 5+.$-(%0; WlLh over 90°
of companles reporLlng
some form of cloud usage, Lhe hurdle of securlLy ln maklng an lnlLlal cloud LranslLlon seems
Lo be mosLly cleared. Powever, lL ls obvlously sLlll a general concern once Lhe LranslLlon has
been made. As companles deslre Lo have more crlLlcal and senslLlve daLa ln Lhe cloud, Lhey
may conLlnue Lo have quesLlons around securlLy and seek ouL provlders wlLh Lhe mosL up-
Lo-daLe Lechnologles and pracLlces.
• <#)%; Cne of Lhe prlmary characLerlsLlcs of Amazon Web Servlces has been frequenL prlce
cuLs-ln !une 2013, Amazon lowered prlces on dedlcaLed servers, racklng up Lhe 37
cuL across all servlces slnce Lhe AWS launch ln 2006. Amazon en[oyed a flrsL mover
advanLage ln prlclng for several years, buL new enLrles ln Lhe markeLplace such as Coogle
CompuLe Lnglne and MlcrosofL Azure are creaLlng downward pressure on prlces.
• =>>+-(24)?>+,%$-+)?#:+2 )%,29,-9); Large cloud provlders can provlde a wlde array of
servlces Lo sulL an equally wlde range of needs, buL Lhere can sLlll be lnsLances where a
cloud provlder sLands ouL. A provlder's cloud servlces may Lle very well Lo on-premlse
equlpmenL from Lhe same vendor. Conversely, open sLandards may be appeallng Lo
cusLomers wlshlng Lo avold vendor lock-ln.
• =$%,4+); Much ls made of cloud ouLages, buL Lhe reallLy ls LhaL sysLems fall. WheLher ln Lhe
cloud or ln an on-premlse daLacenLer, downLlme musL be accounLed for. Cloud ouLages Lend
Lo make headllnes Lhanks Lo Lhe wldespread lmpacL lf a popular servlce goes down. CuLages
are unllkely Lo be greaLly lmproved from one provlder Lo anoLher-a beLLer soluLlon ls Lo
bulld reslllency and redundancy dlrecLly lnLo appllcaLlons.
• @()),%()>,.%(#2 8(%' )+-A(.+; As wlLh ouLages, Lerms of servlce are unllkely Lo dlffer greaLly
from one cloud provlder Lo anoLher, especlally ln Lhe long Lerm. As for cusLomer servlce,
!"#$%&$ (%) *+,-./,&0 1234,. 54%26 1)%7,6")$
B,$$#C$D"6 +,-/ .2$-%E") $")7,."
B,$$#C$D"6 +,-/ -")E$ %( $")7,."
F2-#0"$ +,-/ %),0,&#4 G)%7,6")
H%7" -% E%)" %G"& $-#&6#)6$
I"J") %K"),&0$L("#-2)"$
5%$-$ -%% /,0/ +,-/ %),0,&#4 G)%7,6")
*".2),-M .%&.")&$ +,-/ %),0,&#4 G)%7,6")
*%2)."N !
%&&'() *+,&-. /& 0)1'- 0123'4&5
I#$"N @>9 .4%26 "&6 2$")$ -/#- /#7" $+,-./"6 G234,. .4%26 G)%7,6")$

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 3
one of Lhe draws of cloud compuLlng ls self-servlce capablllLy, whlch may explaln why Lhls ls
noL a ma[or drlver for swlLchlng. SLlll, wlLh one ln flve companles looklng for beLLer cusLomer
servlce, Lhere are opporLunlLles Lo provlde an lmproved experlence.

1$"6(. ! 1-(A,%+

!usL as cloudwashlng has caused confuslon beLween Lrue publlc cloud sysLems and hosLed sysLems,
many end users may glve Lhe label ºprlvaLe cloud" Lo a vlrLuallzed seL of resources. A Lrue prlvaLe cloud
employs sofLware LhaL manages resources auLomaLlcally, dynamlcally allocaLlng as needed wlLhouL
manual lnLervenLlon. A prlvaLe cloud should also offer Lhe same self-servlce capablllLy as found ln any
publlc cloud provlder. A sLandard vlrLual daLacenLer does noL lnherenLly have Lhese feaLures (see
SecLlon 4 for examples of sofLware plaLforms used Lo bulld prlvaLe clouds).

A move from Lhe publlc cloud lnLo a Lrue prlvaLe cloud may be Lhe mosL compllcaLed Lype of mlgraLlon,
Lhanks Lo Lhe Lechnlcal skllls necessary Lo bulld and operaLe Lhe prlvaLe cloud lnfrasLrucLure. 1he mosL
popular way Lo bulld ouL a prlvaLe cloud ls Lo look Lo a Lhlrd parLy. SevenLy percenL of companles have
elLher worked wlLh Lhelr exlsLlng l1 provlder or conLracLed wlLh a speclallsL ln bulldlng prlvaLe clouds.
Some companles have done boLh, posslbly beglnnlng work wlLh Lhelr orlglnal provlder and changlng Lo a
speclallsL lf Lhe work became Loo complex. 1he lncldence of conLracLlng wlLh Lhlrd parLles ls relaLlvely
conslsLenL from SM8s Lo large companles, lndlcaLlng LhaL Lhls speclallzed sklll wlLhln Lhe 8ulld channel
model ls ln hlgh demand.

1he oLher sLraLegy used when bulldlng prlvaLe clouds ls Lo reLraln exlsLlng sLaff or hlre Lhe approprlaLe
skllls. Pere, Lhere ls more dlfferenLlaLlon dependlng on company slze. llfLy-one percenL of large
companles have reLralned sLaff compared Lo 31° of small companles. lor hlrlng, Lhe gap ls even wlder-
63° of large companles have hlred new skllls compared Lo [usL 38° of small companles. Clearly, large
companles have Lhe capablllLy Lo Lraln or hlre whlle also malnLalnlng currenL operaLlons. lor Lhls reason,
Lhey may come Lo depend less on Lhlrd parLles for prlvaLe cloud supporL, whereas small companles may
sLlll requlre ouLslde experLlse.

1$"6(. ! =2B1-+C()+

lf a move Lo a prlvaLe cloud ls Lhe mosL
compllcaLed Lype of move, a move back
Lo on-premlse sysLems ls llkely Lhe
slmplesL. unless Lhe move lncludes
some addlLlonal archlLecLural changes,
Lhe company ls reverLlng Lo Lhe
meLhods used prlor Lo Lhe cloud
LranslLlon. 1here wlll sLlll be work
lnvolved, buL Lhe end resulL ls a known

lL ls more lnLeresLlng Lo look aL Lhe
drlvers for such a move. 1here were a
wlde range of opLlons for swlLchlng
publlc cloud provlders, buL by far Lhe
8easons for 8eLurnlng Lo an Cn-Þremlse Model
noL achlevlng cosL goals
uld noL have proper cloud skllls
ulssausñed wlLh cloud rellablllLy
ulssausñed wlLh performance
Could noL successfully lnLegraLe
needed Lo conLrol securlLy
Source: !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 93 cloud end users LhaL have swlLched Lo on-premlse sysLems

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 3 6
prlmary moLlvaLlon Lo move back Lo an on-premlse sysLem ls securlLy. 1hls does noL necessarlly mean
LhaL securlLy has caused companles Lo compleLely reLreaL from Lhe cloud. Companles could slmply have
moved a parLlcular appllcaLlon LhaL glves Lhem concern, and Lhey sLlll have less senslLlve appllcaLlons on
publlc cloud plaLforms. SLlll, lL ls anoLher lndlcaLlon LhaL cloud securlLy wlll remaln a hoL Loplc for Lhe
foreseeable fuLure.


AlLhough cloud mlgraLlons have rapldly become a parL of cloud adopLlon for many companles, many
flrms have noL yeL sLarLed Lhls Lype of acLlvlLy. Lven ln Lhe nexL year, Lhough, cloud mlgraLlon numbers
may approach overall adopLlon raLes: 44° of Lhose companles LhaL have noL yeL swlLched provlders or
moved Lo company-owned sysLems predlcL Lhey wlll do so ln Lhe nexL Lwelve monLhs.

As Lhese companles begln Lo perform mlgraLlons, Lhey may be ln for somewhaL of a surprlse. Compared
Lo companles LhaL have already swlLched from one publlc cloud provlder Lo anoLher, companles LhaL are
expecLlng Lo move underesLlmaLe Lhe efforL lL wlll Lake Lo Lransfer Lhelr cloud soluLlons. 1hls wlll be
dependenL on Lhe deLalls of Lhe Lwo provlders-for example, swlLchlng from one publlc CpenSLack
cloud Lo anoLher would probably enLall less efforL Lhan Lhe orlglnal LranslLlon. lf Lhe underlylng
plaLforms are dlfferenL, Lhough, Lhere wlll be dlfferences ln AÞls and ln Lhe overall sLrucLure of Lhe cloud
soluLlon LhaL wlll requlre changes Lo lmplemenLaLlon and lnLegraLlon.

LñorL of Cloud Change vs. Crlglnal 1ransluon



SlgnlñcanLly easler
ModeraLely easler
no dlñerence ln eñorL
ModeraLely more dlmculL
SlgnlñcanLly more dlmculL
LsumaLe from companles
plannlng Lo swlLch
AssessmenL from
companles LhaL have
swlLched provlders
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5 sLudy
8ase: 180 end users LhaL have swlLched provlders/148 end users plannlng Lo swlLch provlders
F OU R T H ANNUAL • AU GU S T 2 0 1 3

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 2
/+0 1#(2%)

• Whlle cloud compuLlng ls becomlng hlghly adopLed and well accepLed, Lhere are sLlll many
challenges Lo be dealL wlLh as companles Lransform Lhemselves. 1he flrsL hurdle ls procuremenL.
1he lncldence of rogue l1-where llnes of buslness are procurlng Lhelr own cloud soluLlons
wlLhouL necessarlly lnvolvlng Lhe l1 deparLmenL-has dropped sllghLly, buL companles musL sLlll
flnd ways Lo ensure LhaL boLh buslness needs and l1 concerns are belng addressed.

• CosL Lracklng ls Lhe second hurdle, wlLh 9° of companles noL even Lracklng cosLs and anoLher
12° flndlng savlngs Lo be lower Lhan esLlmaLed. Plgher usage cosLs, sLaff Lralnlng, and neLwork
upgrades can all conLrlbuLe Lo cosL surprlses.

• Lnd users conLlnue Lo have a spaLe of challenges ln Lhe lmplemenLaLlon sLages, led by
lnLegraLlon wlLh exlsLlng sysLems. Powever, companles LhaL have 3+ years of adopLlon hlsLory
flnd LhaL modlfylng pollcy has become a greaLer challenge Lhan lnLegraLlon, showlng LhaL
Lransformlng buslness pracLlces ls [usL as dlfflculL as a Lechnlcal mlgraLlon.

• 1ransformaLlon can be even more challenglng for channel flrms, whose llvellhood depends on
successfully LranslLlonlng Lo a new model for offerlng servlces. 1he Lop challenge clLed by
channel flrms ls developlng experLlse among Lhelr Lechnlcal and sales sLaff-clearly an lmporLanL
plece ln bulldlng servlce offerlngs and communlcaLlng value.

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 3
1-#.$-+3+2%4 5'+ 6(-)% 7#8()%(.,9 :$-;9+

Whlle buslness model LransformaLlon and mulLlcloud sLraLegles may slgnal Lhe shape of Lhlngs Lo come
ln Lhe cloud era, many companles are sLlll deallng wlLh more baslc loglsLlcal lssues. ÞarLly, Lhls ls due Lo
Lhe large number of companles LhaL are sLarLlng cloud adopLlon afLer walLlng for Lhe early adopLers Lo
blaze a Lrall. ÞarLly, lL ls due Lo Lhe facL LhaL Lhe loglsLlcs of cloud compuLlng LranslLlons are compllcaLed
and requlre slgnlflcanL Llme and energy from companles as Lhey sorL Lhrough Lhe lssues.

1he hurdles begln wlLh procuremenL.
Compared Lo lasL year, Lhe lncldence of
rogue l1-where llnes of buslness are
procurlng Lhelr own cloud soluLlons
wlLhouL necessarlly lnvolvlng Lhe l1
deparLmenL-has dropped sllghLly.
Companles may have experlenced lssues
wlLh sysLems procured wlLhouL l1
lnvolvemenL, such as securlLy breaches,
downLlme, or lnLegraLlon problems.
1hese compllcaLlons, whlch end up
cosLlng Lhe company more Lhan lnsLalllng
Lhe proper soluLlon Lo begln wlLh, could
be leadlng Lo mandaLes from execuLlve
managemenL Lo beLLer cenLrallze
Lechnology declslons.

CenLrallzlng Lechnology declslons does
noL necessarlly mean glvlng Lhe l1
deparLmenL full conLrol and flnal approval
ln every slLuaLlon. 1he appllcaLlons LhaL
are currenLly mosL llkely Lo be procured
by a llne of buslness are Lhose
appllcaLlons wlLh speclflc buslness use
cases-flnances, P8, or cusLomer
relaLlons. 1he buslness owners wlll have
Lhe besL sense of whlch Lechnology meeLs
Lhelr needs, and Lhey need Lo be able Lo
communlcaLe Lhose needs [usL as much as
l1 needs Lo be able Lo ralse a flag over
securlLy or lnLegraLlon concerns.

lor now, Lhe l1 deparLmenL ls prlmarlly
Lhe procuremenL agenL for appllcaLlons
LhaL have more company-wlde lmpacL,
such as emall, collaboraLlon, or buslness
producLlvlLy. As llnes of buslness become
more Lechnlcally savvy, Lhey may show
greaLer lnLeresL ln bulldlng ouL Lhelr own
lnvolvemenL of l1 1eam Where Llnes of
8uslness Þrocure Cloud Soluuons
6478 94:,+3./;
<,5 94-,7-./
Source: Comp1lA's !
%&&'() *+,&-. /& 0)1'- 0123'4&5 sLudy
8ase: 243 u.S. ñrms where buslness sLañ procured cloud soluuons
=,4>275./ ?95@,25
A4B7 B**+,-B7
!""#$%&'() +$), (-
!"#$#%"$& ($#$)*+*#,
-./ 01/
23 ($#$)*+*#,
-0/ 0-/
456*#7* ($#$)*+*#,
-8/ 0-/
9$&& 9*#,*:
-;/ 0</
1=/ 0=/
43> 1./ .1/
2*&6 ?*7@
1./ .1/
AB7"#*77 >:CDB%EF",G
1H/ .8/
I*J >:*7*#%*
18/ .H/
11/ .0/
11/ ../
L=/ .=/
M":,B$& ?*7@,C6
L./ <1/
N"#* CO AB7"#*77 >:C%B:*+*#,

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 4
verslons of Lhese appllcaLlons, leadlng Lo fragmenLaLlon across deparLmenLs.

Comlng up wlLh a procuremenL and usage plan LhaL meeLs Lhe needs of all parLles ls one of Lhe greaL
challenges noL [usL for cloud compuLlng, buL also for many Lechnologles LhaL are enLerlng buslnesses
Loday. Companles need Lo bulld a Lop-down undersLandlng of how needs wlll be presenLed, how
Lradeoffs wlll be made, and how rlsks wlll be accepLed or mlLlgaLed. 1he daLa shows LhaL organlzaLlons
are movlng ln Lhe rlghL dlrecLlon, buL Lhls remalns an area where besL pracLlces can be esLabllshed and

<#)%4 5'+ 6#99#=$> %# 1-#.$-+3+2%

WlLh a soluLlon procured, mosL buslnesses wlll wanL Lo Lrack cosLs ln order Lo [usLlfy Lhe lnvesLmenL,
especlally wlLh cosL savlngs belng a Lop drlver for adopLlon. 1hls Lracklng alone can presenL a challenge.
nearly one ln Len companles sLaLe LhaL Lhey have noL Lracked cosLs closely enough Lo valldaLe Lhe
orlglnal esLlmaLes. CosL Lracklng could be a lower prlorlLy for some flrms, especlally lf Lhey are
experlmenLlng or placlng low-lmpacL appllcaLlons ln Lhe cloud. CLher flrms may noL have Lhe proper
Lools needed Lo monlLor usage and cosLs.

!"#$%&'( *+,- +. */+01 2+/03+',
!""#""$#%& () *("& *($+,-#. &(
/-010%,2 3"4$,&#"
4"5#-5" -6#'
8' 9#" :&-6
;+:5" -6#'
;#,"(%" )(- *("& 3<=##.0%1 3"4$,&#
2+0"$5< !
%&&'() *+,&-. /& 0)1'- 0123'4&5
=#,5< >?@ $/+01 5'1 0,5",
A+-5< BC +.
6#D5 '+-
-"#$%51 $+,-,
• E$-0#/ 0,#(5 $+,-, 6&(65" -6#'

• E11&3+'#/ -"#&'&'( .+" ,-#F

• G9("#15, -+ '5-:+"%

• G'9/#''51 0,5 +. -6&"1 9#"-H
.+" $/+01 -"#',&3+'
• E1151 ,5$+'1 $/+01 9"+D&15"
.+" "5/&#I&/&-H

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 3
lor Lhose companles LhaL have Lracked cosLs, mosL say LhaL Lhe cosLs have meL or exceeded orlglnal
esLlmaLes. Powever, a number of companles have experlenced compllcaLlons or surprlses LhaL have led
Lo cosLs belng hlgher Lhan anLlclpaLed.

lL should also be noLed LhaL Lhe cosL equaLlon for many cloud soluLlons ls becomlng beLLer undersLood,
and many companles may be seLLlng Lhelr expecLaLlons based on whaL Lhey are seelng ln Lhe markeL.
lor example, an appllcaLlon LhaL has dramaLlc peaks and valleys ln lLs workload may lndeed be more
cosL-effecLlve ln a cloud seLLlng, whereas an appllcaLlon wlLh a relaLlvely sLeady workload may be cosL-
neuLral or even more expenslve. Powever, a company may sLlll prefer Lhe cloud soluLlon for Lhe sLeady
workload Lhanks Lo oLher beneflLs, such as self-servlce or auLomaLlc upgrades. Companles enLerlng Lhls
Lype of arrangemenL may be well aware of Lhe cosL lssues, so Lhey are noL surprlsed even Lhough Lhey
were noL necessarlly anLlclpaLlng huge savlngs.

?3>9+3+2%,%(#24 ! :#)% #@ <',99+28+) @#- A2; B)+-)

8eyond cosL Lracklng, Lhere conLlnue Lo be a number of lmplemenLaLlon challenges LhaL are more
Lechnlcal or buslness-relaLed ln naLure. AL a Lop level, Lhere ls noL much change year over year ln Lhe
varleLy or lmpacL of challenges. Powever, Lhere are dlfferences ln Lhe deLalls LhaL provlde lnslghL lnLo
Lhe dynamlcs behlnd Lhese lssues.

Challenges of uslng Cloud
Lower avallablllLy Lhan expecLed
Lower performance Lhan expecLed
Locked ln Lo cloud provlder
Þalnful Lransluon from legacy sysLems
ueñnlng/[usufylng 8Cl
CosLs hlgher Lhan orlglnally esumaLed
Learnlng curve for cloud model
Changes Lo l1 pollcy
lnLegrauon wlLh exlsung sysLems
Source: !
%&&'() *+,&-. /& 0)1'- 0123'4&5
8ase: 449 u.S. end users wlLh cloud soluuons

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 6

Cnce agaln, lnLegraLlon Lops Lhe llsL of challenges across all companles. 1he dlfflculLles ln enabllng
operaLlon beLween cloud sysLems and legacy sysLems lnvolve learnlng and flne-Lunlng AÞls, Lyplcally on
Lhe cllenL slde for SaaS appllcaLlons and boLh cllenL and server slde for laaS/ÞaaS. Lnd users musL learn
how Lo hook sysLems LogeLher and also reacL qulckly when Lhere are unexpecLed resulLs.

Powever, lnLegraLlon has sllpped Lo Lhe second largesL lssue for early adopLers, Lralllng changes Lo l1
pollcy. Cn one hand, lL ls somewhaL surprlslng Lo see how sLrongly lnLegraLlon sLruggles perslsL as
companles maLure ln Lhelr cloud usage. newer and more buslness-crlLlcal appllcaLlons cerLalnly brlng a
dlfferenL seL of challenges, forclng l1 Leams Lo consLanLly learn and lnnovaLe. Cn Lhe oLher hand, Lhls ls
an lndlcaLlon of Lhe dlfflculLy ln formlng new pollcles around cloud models. 1he sLep Lo a fully
Lransformed buslness appears Lo be aL leasL as dlfflculL as Lhe sLep Lo a flrsL-Llme cloud user.

Slmllarly, slgnlflcanLly more early adopLers experlence Lhe problems llsLed Lowards Lhe boLLom of Lhe
llsL: vendor lock-ln, avallablllLy, and performance. vendor lock-ln ls noL surprlslng, slnce early adopLers
have reached a sLage where Lhey are explorlng oLher cloud opLlons Lo flnd Lhe besL flL for Lhelr buslness
(see SecLlon 3 for more deLalls on cloud mlgraLlon). AvallablllLy and performance, Lhough, are sllghLly
more surprlslng. Cne mlghL expecL LhaL companles would seL Lhelr expecLaLlons ln Lhese areas early ln
Lhe adopLlon process and ad[usL as Lhey bulld ouL more robusL sysLems. lL seems LhaL Lhe opposlLe ls
happenlng-companles become more concerned as Lhey spend more Llme wlLh cloud sysLems and
conslder placlng more buslness-crlLlcal appllcaLlons ln Lhe cloud. Clearly, mlLlgaLlng rlsk around
downLlme or performance ls noL a Lask llmlLed Lo companles [usL beglnnlng Lhelr cloud [ourney.

5-,2)(%(#24 <',99+28+) 6,.+; "0 <',22+9 6(-3)

Whlle end users grapple wlLh procuremenL declslons and flnanclal proof polnLs around uslng cloud, l1
channel flrms slmllarly face a range of Lhelr own challenges ln elLher launchlng a cloud pracLlce or Lrylng
Lo scale and grow an exlsLlng one.

Cne of Lhe crlLlcal quesLlons for channel companles Lo answer ls wheLher or noL cloud makes sense from
an 8Cl perspecLlve and lf so, ln whaL capaclLy and ln whlch cusLomer scenarlos. 1hls baslc ºeconomlcs of
Lhe cloud" dlscusslon has been fronL-and-cenLer ln Lhe channel for Lhe beLLer parL of Lhe lasL Lhree Lo
flve years. 1he conversaLlon ls compllcaLed, due ln large parL Lo Lhe wlde varleLy of cloud buslness model
opLlons and poLenLlal revenue sLrucLures Lo explore as well as dlfferlng cusLomer needs. And yeL, we are
seelng soluLlon provlders move more declslvely. nearly 6 ln 10 sald Lhey proacLlvely pursued mulLlple
segmenLs of Lhe varlous cloud buslness models ln an aLLempL Lo qulckly and comprehenslvely enLer Lhe
cloud markeL, wlLh medlum and larger flrms more llkely Lo have gone Lhls rouLe Lhan Lhe smallesL
channel player (see SecLlon 3 of Lhls reporL for a deLalled dlscusslon of buslness models). As a resulL, a
segmenL of companles have assembled quanLlflable Lracklng meLrlcs on revenue and proflL margln,
whlch can serve as a guldeposL for channel companles movlng more slowly lnLo cloud.

Some deLalls: 8oughly 6 ln 10 channel flrms LhaL descrlbed Lhe sLaLe of Lhelr cloud buslness as maLure
and sLraLeglc sald LhaL cloud-based proflL marglns are generally hlgher Lhan Lhelr legacy-based proflL
marglns. AnoLher 6 ln 10 companles LhaL are lnvolved ln four dlfferenL cloud-based buslness models
(and Lherefore one can assume cloud ls sLraLeglc Lo Lhem) sald LhaL Lhelr revenue from cloud was
growlng fasLer Lhan Lhelr legacy sales. 1hese meLrlcs compare agalnsL [usL 2 ln 10 flrms wlLh ºnon-

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 7
maLure" cloud pracLlces LhaL sald cloud proflLs were hlgher Lhan proflLs from exlsLlng offerlngs, for
example. Clven Lhe facL LhaL many producL-focused vendors Loday are elLher experlenclng a demand
slowdown - uell and PÞ, for example - or llke Clsco forecasLlng dlfflculL markeL roads ahead, Lhe
declslon Lo place an emphasls on cloud and oLher Lypes of servlces by soluLlon provlders ls a parLlcularly
smarL one. So whlle cusLomers mlghL be more reLlcenL Lo spend caplLal on producLs, as some Lech
manufacLurers are seelng, Lhey sLlll need l1 servlces, wheLher lL ls managed, professlonal or cloud.

And yeL desplLe Lhe dlrecLlon of Lhe markeL or Lhe enLhuslasm of cusLomers, swapplng one model for
anoLher lsn'L always smooLh salllng for channel flrms. 1he facL ls LhaL cloud-based revenue models dlffer
greaLly from producL-based LransacLlonal sales operaLlons, whlch ln Lurn affecLs everyLhlng from
lncomlng cash flow Lo Lhe ways ln whlch sales reps are pald. WhaL was once a slgnlflcanL producL or
producL sale pald for ln one LransacLlon ls now a model based on small lncremenLs of recurrlng revenue
each monLh. 8uslness LransformaLlon - consldered a musL for long-Lerm channel flrm survlval - lsn'L for
Lhe falnL of hearL. Also ln play ls preclplLous prlce dropplng among cloud servlces provlders such as
Amazon, whlch puLs pressure on channel flrms.

Cn a general level, as evldenced ln Lhe charL below, mosL channel flrms lnvolved ln cloud, wheLher
maLure or oLherwlse, are beglnnlng Lo see real reLurns and a blL of a changlng of Lhe guard beLween
cloud- and legacy-based revenue and proflL levels.

1hese poslLlve growLh and margln raLes suggesL a decenL paLh for channel flrms LhaL make cloud parL or
all of Lhelr buslness. 8uL Lhere remaln slgnlflcanL challenges Lo dolng so. !usL as end users have hllls Lo
!" $%&''() *+,-. /(( 01.+23( 4(3('5(
!-6&78 9,1- $)15: ;<(,+'=.
!"#$% '() *(+,-".(/0% 12#%$3+(4
502'.30( 60'07$0 82#9+/
=,1@+'= 9&.8(,
A,1@+'= &8
.&-( ,&8(
$)15: =,1@+'=
/15,7(B $1-6"!CD. !
%&&'() *+,&-. /& 0)1'- 0123'4&5
E&.(B FGG 7%&''() H,-. @+8% 7)15: 1<(,+'=.
!"#$% '() *(+,-".(/0% 12#%$3+(4
502'.30( 12#?+ @,2A.7(
415=%)I 8%(

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 8
cllmb when Lhey declde Lo Lry cloud-based soluLlons for Lhelr l1 needs, soluLlon provlders face lnLernal
buslness challenges ln addlng cloud Lo Lhelr porLfollo. CperaLlonal hurdles exlsL alongslde exLernal
obsLacles such as cusLomer skepLlclsm and Lhe need ln many cases Lo reLhlnk relaLlonshlps wlLh vendors,
dlsLrlbuLors and oLher supply chaln parLners (see SecLlon 4 for deLalled analysls of Lhe cloud's lmpacL on
channel parLnershlps wlLh vendors, cusLomers and dlsLrlbuLors).

1he Lypes of lnLernal lssues LhaL channel flrms are navlgaLlng Lhrough have evolved somewhaL over Lhe
pasL year, reflecLlve of a growlng adopLlon raLe and maLurlLy of cloud use. new Lo Lhe Lop of Lhe llsL of
challenges lnclude Lwo hlghly speclflc enLrles perLalnlng Lo human resources. Palf of channel flrms clLed
Lhe need Lo develop cloud experLlse ln boLh Lechnlcal and sales areas as Lhelr prlmary challenge Lo
adopLlng a cloud focus, whlle anoLher half (49°) hlghllghLed Lhe efforL necessary Lo opLlmlze Lhelr sales
and markeLlng messages around cloud.

1he emphasls on sales LranslLlon ls an lmporLanL one. Cloud servlces can be an amorphous concepL for
cusLomers, whlch lends lLself Lo sales complexlLy. Companles LhaL have LranslLloned Lo a cloud model

ln parL or full have dlscovered LhaL Lralnlng ls paramounL Lo geLLlng an exlsLlng sales force or new hlres
Lo effecLlvely peddle cloud offerlngs. Slmllarly, soluLlon provlders are recognlzlng LhaL Lhey need Lo
make ad[usLmenLs Lo Lhelr markeLlng messaglng. 1yplcally Lhey have markeLed Lhelr buslnesses wlLh a
sLrong promoLlon of Lhe vendor brands Lhey speclallze ln. lL's noL uncommon for markeLlng maLerlal,
web slLes and oLher collaLeral Lo LouL a channel flrm's sLaLus as a MlcrosofL Cold CerLlfled or PÞ Alllance
parLner, for example. 8uL cloud servlces are dlfferenL ln LhaL Lhe vendor ldenLlflcaLlon ls less vlslble Lo
Challenges Channel laced ln Adopung Cloud
uecldlng vendors Lo work wlLh
lnlual sLarLup cosLs
8alanclng needs of legacy buslness
ueLermlnlng buslness model
Cash ßow/ñnanclal conslderauons
Cpumlzlng cloud markeung/sales messaglng
ueveloplng cloud Lech/sales experuse
Source: Comp1lA's !"# %&&'() *+,&-. /& 0)1'- 0123'4&5 sLudy
8ase: 400 u.S. l1 or 8uslness execuuves (aka end users) uslng cloud soluuons

%&&'() *+,&-. /& 0)1'- 0123'"/&4: SecLlon 6 9
Lhe end cusLomer. 1he soluLlon provlder ls Lhe vendor ln a sense, especlally ln cases where Lhe channel
flrm ls whlLe-labellng Lhe offerlng uslng lLs own name and brand. As a resulL, companles are Llnkerlng
wlLh boLh markeLlng and sales messaglng Lo reflecL a more accuraLe plcLure of Lhe value LhaL Lhey brlng
versus Lhe source of Lhe offerlng lLself.

nearly equal ln slgnlflcance level are Lhe resL of Lhe challenges clLed by channel flrms maklng Lhe move
Lo cloud, wlLh mosL of Lhose hurdles cenLered on flnanclal declslons. lnlLlal sLarL up cosLs, for example,
can be mlnlmal or qulLe large, dependlng on wheLher or noL Lhey lnvolved bulldlng a daLa cenLer Lo
provlde cloud servlces. lnLeresLlngly, Lhe largesL channel flrms clLed Lhls as a ma[or challenge, Lhough
Lhey are mosL llkely Lo have Lhe deeper pockeLs needed Lo ouLflL a new daLa cenLer lf Lhey don'L already
have one ln exlsLence. MeanLlme, cash flow and oLher flnanclal conslderaLlons ranked hlghesL among
channel flrms (63°) lnvolved ln all four Lypes of cloud buslness models ouLllned ln Lhls sLudy. 1hls
suggesLs LhaL Lhe level of commlLmenL Lhey have made Lo cloud has compllcaLed flnanclal fundamenLals,
one example would be Lhe effecLs of a decreased rellance on legacy sLreams of revenue, whlch ln Lhe
shorL-Lerm could creaLe cash flow concerns as Lhey ramp cloud sales.

Pow have channel flrms handled Lhese challenges? 1he resulLs are sLralghLforward:

CDE lnvesLed ln Lechnlcal, sales and buslness Lralnlng
CFE 1ranslLloned [usL a porLlon of Lhelr buslness Lo cloud, wlLh plans Lo scale laLer
CGE Þursued vendor-based buslness Lralnlng
HIE ÞarLnered wlLh oLher soluLlon provlders Lo comblne sklll seLs
HJE Plred new sales reps wlLh cloud experlence
HFE Þursued new llnes of caplLal/credlL

Channel flrms wlLh hlgh maLurlLy around cloud were mosL llkely Lo have parLnered wlLh anoLher soluLlon
provlder Lo Lake advanLage of complemenLary sklll seLs. lorLy-slx percenL of Lhese flrms enLered lnLo Lhls
klnd of collaboraLlve relaLlonshlp compared wlLh a Lhlrd of Lhose wlLh elLher some level of cloud
maLurlLy or Lhose wlLh a low degree of cloud experLlse. 1he same held for medlum and larger channel
flrms, nearly half of whlch parLnered wlLh anoLher company Lo enhance cloud skllls. 1hls compared Lo
less Lhan a Lhlrd of Lhe smallesL channel flrms LhaL have done so.

1hls parLnerlng ls an example of Lhe sLraLegy needed Lo geL on Lhe radar screen for more end users (see
SecLlon 4 for deLalls). As end users progress from Lhe LxperlmenL phase Lowards 1ransformed l1,
channel flrms of all models (8ulld, Þrovlde/Þrovlslon, Lnable/lnLegraLe, and Manage/SupporL) can asslsL
wlLh Lhe varlous sLeps ln Lhe process. lrom consulLlng Lo bulldlng prlvaLe clouds Lo enabllng new
workflows, channel flrms should Lake a proacLlve vlew of where Lhey can lead Lhelr cusLomers uslng
cloud Lools.

1he blggesL challenge may be ln cuLLlng Lhrough all Lhe confuslon surroundlng Lhls powerful new model.
lor many end users, hosLed compuLlng models may compleLely fulflll Lhelr needs Loday, and provlders
offerlng Lhese soluLlons under a cloud label may be experlenclng success ln Lhe markeLplace. As Lhe
focus shlfLs more and more from Lhe dellvery meLhods Lo Lhe servlces provlded, l1 flrms wlll wanL Lo
make sure Lhey are ahead of Lhe curve ln uLlllzlng Lhe Lechnology componenLs aL Lhelr dlsposal.
© 2013 CompTIA Properties, LLC, used under license by CompTIA Member Services, LLC. All rights reserved. All membership activities and offerings to members of CompTIA, Inc. are
operated exclusively by CompTIA Member Services, LLC. CompTIA is a registered trademark of CompTIA Properties, LLC in the U.S. and internationally. Other brands and company names
mentioned herein may be trademarks or service marks of CompTIA Properties, LLC or of their respective owners. Reproduction or dissemination prohibited without written consent of
CompTIA Properties, LLC. Printed in the U.S. August 2013 00304-US.