Seongkyuun Cho and Gle enn Ballard (2

2011) Last Planner

and Integgrated Projectt Delivery Lean Constructioon
Journal 2
2011 pp 67-78 - Lean and In ntegrated Projject
Delivery SSpecial issue www.leancon
w structionjourn
nal.org

Last Planner and In

ntegrate
ed Project De
elivery
y
Se Cho1 and Glenn Ballarrd2
eongkyun C

Abstrract
Researc ch Question ns: 1) Doess the use of Last Planner (LP) imp prove projec ct performaance?
2) D
Does Integrated Projecct Delivery (IPD) show different project
p perfformance? 3)
3 Do
IPD
D projects use LP?
Purpose e: The firstt objective iis to figure out the rellationship between
b IPD
D, LP, and project
p
perrformance.
Researc ch Method:: survey of ‘Lean’ projects known n to adopt LP,
L includin ng IPD proje
ects, to
dettermine the e correlation between LP impleme entation annd Project performanc
p e (cost
red
duction + tim me reductio on); and a T test betwween IPD and non-IPD projects.
p
Findinggs: 1) Theree is significa
ant correlattion betwee en the degrree of imple ementation of LP
andd project pe erformance e; 2) IPD proojects do no
ot show sign nificantly different
d
perrformance from
f that of others not adopting IPD; and 3) IPD projec cts do not sh
how
sign
nificantly different implementation of LP fro om that of others but their
impplementatio on is near too significan
nce
Limitations: Limitations in sa ample size and
a data qu uality reducce the crediibility of
genneralizationns.
Implica
ations: This exploratoryy research revealed in nteresting and
a importa ant relationships
bettween proje ect structurres and pracctices on thhe one hand d and proje ect performance on
thee other.
Value foor practitiooners: The findings fro om this papper can be used
u by industry practitioners
to ddesign proje ect deliveryy systems foor better performance e.
Keyworrds: Integra ated Projecct Delivery, Last Plannner, Lean Co onstruction, survey.
Paper type:
t Full paper.
p

Litera
ature Review
R
According to the American
A Insstitute of Architects
A (A
AIA), the In
ntegrated Prroject Delivvery
(IPD) co
ontract form
m includes:
ƒ Early involvement of Key particiipants;
ƒ Shared risk and rewa
ard;

1
PhD C
Candidate, Civvil and Env. Enngineering. De
epartment, 4007-A McLaughllin Hall, Univ. of California,
Berke
eley, CA 94720 0-1712, USA, Phone
P +1 510/
/725-7929, seoongKyuncho@ @berKeley.edu u
2
Directtor, Project Production System Laborato ory, http://p2sl.berKeley.edu, and Adjun nct Associate
Professsor, Civil and
d Env. Enginee
ering. Departm
ment, 215-A McLaughlin
M Ha
all, Univ. of Ca
alifornia, Berk
keley, CA
947200-1712, USA, Phone
P +1 415/
/710-5531, [email protected]
Lean Co
onstruction Journal 2011 p
page 67 www.leanconstructionjournal.org
http://c
creativecommons..org/licenses/by-nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

ƒ Multi partyy contract;

ƒ Collaborattive decisio
on making and
a control;;
ƒ Liability waivers
w amoong key participants; and
a
ƒ Jointly developed and validatedd project gooals (Cohen et al., 2010).
Similarlly, the Natioonal Association of Staate Facility Associationn (NASFA), Constructio on
Owners Association n of Americca, Associattion of Highher Educatioon Facilitiess Officers,
Associatted Genera al Contracto
ors in Ameriica, and Am merican Insttitute of Arcchitects defined
IPD as a project de elivery syste
em using a multi partyy contract that
t has moore than two o
parties selected byy qualification based procuremen
p nt, managed d/shared risk, compen nsation
based oon team perrformance without
w GM
MP, and open book acco ounting (NA ASFA et al 2010).
2
According to CMAA A, the purpoose of IPD iss to solve currently
c accknowledged problemss in the
construction industry such as low rates of o productivvity, high raates of ineffficiency annd
rework,, frequent disputes,
d exxcessive cosst, and exce essive dura
ation--all ca
aused by
organiza ational, commmercial, and
a operatiional proble ems in curre ent projectt delivery syystems
(Thomse en et al., 2009)
2
The Last Planner (LP) is a productionn planning and
a control system imp plemented on
construction proje
ects to imprrove plannin
ng and prodduction perfformance. It has four main
processes:
ƒ hedule;
Master sch
ƒ Phase scheedule;
ƒ Look aheaad Plan; and
d
ƒ Weekly Plaan (Hamzehh, 2009).
Many reesearchers have
h provedd reducing plan variab
bility helps increase
i prroductivity, such as
Liu et a
al (2008) sug
ggesting a regression
r l 3 between plan reliiability and
line d productivity, and
Alarcon et al (19977) showing difference
d in productivity before and after implementi
i ing LP.
Again, tthe LP has been
b createed to maxim
mize reliabiility of the work/mateerial/inform
mation
flow to minimize waste
w in tim
me/money in project pprocesses an nd to maximmize custommer
value (B
Ballard, 20000)
While IPD has tried
t to inte
egrate project particip pants’ roless and relatio
ons contracctually
in orderr to improvee project o
outcomes, LPL has enforrced system matic production contrrol
reducing plan variaability for the
t same pu urpose. Ourr question iss if having project
organizaation integrrated by ussing contrac
ctual alignmment, such as a IPD, is en
nough to maximize
desired outcomes, such as cost/time red duction. If it
i is not enoough, our next concern n is
whether the imple ementation of LP can achieve
a those outcome es. To find out
o the ansswers to
those questions, we
w did some e hypothesiss testing in this researrch.
Hypothesis testting regardiing project performance based on a large nu umber of prrojects
is a welll establishe
ed methodo ology. For example,
e Chhoi (2008) used
u one waay ANOVA (AAnalysis
of Varia
ance) to invvestigate if there is sig
gnificant diffference in schedule performance
p e among
three diifferent con ntract types, selected from a govvernment da atabase of more than 1,700
projectss. More simmilar to our research de esign, Sanvido et al (1998) made a survey
questionnaire, sent it to 7600 0 projects, and
a got 378 8 responsess on which they
t did
multivariate t-testt, chi squaree test, ANOOVA, and regression to identify peerformance
differen
nces among g three project deliverry systems.

3
Labou
ur Productivityy = 0.530 + 1.0
095*Weekly Plan Percent Completion
C
Lean Co
onstruction Journal 2011 p
page 68 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

Resea
arch De
esign
Resea
arch Hypo
othesis
Our reseearch assummption: prooject perforrmance variies with Lasst Planner (LP)
(
implementation. Based
B on thiis assumptio
on, we diaggnosed the degree of LP
L impleme ented in
Integratted Projectt Delivery (IPD) projectts to determ
mine the co
orrelation between
b LP
implementation an nd IPD projeects’ perforrmance. Thhis assumptiion must bee supported
d by
general hypothesiss testing. Thhus, our first research hypothesiss is:

If a pro
oject imple
ements Lastt Planner (LLP) more, itt achieves better
b
pro
oject perfo
ormance better than th hose emplooying LP lesss.

If the first hypothe

esis had nott been supp ported, it would
w be meeaningless to
t go furtheer
comparring IPD projjects with others
o in te
erms of LP and
a our reseearch would have been
redirectted to a quaalitative ex
xploration seeking
s whaat caused LP
P to fail. Hoowever, the
e first
hypothe esis was sup
pported, ma aking it meaningful to test the se
econd and third
t hypothhesis.
The seccond hypoth hesis is:

If a proje
ect adopts Integrated Project
P Delivery (IPD),, its perform
mance is
diffferent from
m those of other projeccts.

A
And the thirrd hypothessis is:

If a proje
ect adopts IPD,
I its deg
gree of impllementation
n of Last Pllanner is
diffferent from
m those of other projeccts.

This papper is devotted to the interpretati

i ion of the results
r from
m the first, the
t second and the
third hyypothesis te
esting.

Resea
arch Meassuremen
nt
The firsst thing thatt we have to
t do after forming hypotheses is to specify the measurrement
of varia
ables. We co onceptualizzed our variiables as sh
hown in Figu
ure 1, follow
wing Adcocck et al
(2001).

Lean Co
onstruction Journal 2011 p
page 69 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

Figure
e 1: Concep
ptualizatio
on and measurement: Levels and
d task (Adcock et al., 2001)
We struuctured the variables in the hypottheses so th
hey could be measured
d in the folllowing
parts.
The independe e of the firstt hypothesis is the deg
ent variable gree of implementation of
Last Pla
anner (LP). To measure e this abstrract concept, we developed indicators to be scored
based oon the follow
wing five ellements:

1 Pullin
ng productiion: each worker
w invesstigates the
e readinesss of the
next workers
w (im
mmediate cu ustomers) before
b execcution of ta
asks
(Tommmelein, 199
98)
2 Looka ahead proccess: each front
f line supervisor removes
r connstraints
(prere equisite woork, contracctual approovals, sequeential
inapppropriatenesss, insufficient resourrce as well asa labour &
equippment, inad dequate durration, fund ding probleem, problemms found
in firsst run studyy, etc) befo
ore executioon of its ta
asks. Constrrained
tasks are not elig gible for in
nclusion on daily or weeekly work plans
p
(Ballaard, 2000)
3 Learn
ning from breakdowns
b s: failures to complette planned tasks
t are
analyz
yzed to roott causes and
d actions arre taken to prevent
reoccurrence (Baallard, 2000
0)

Lean Co
onstruction Journal 2011 p
page 70 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

4 Phasee schedulinng: every haandoff in a phase shou uld be defin

ned by
collab
boration of all relevan
nt specialistts in the ph
hase before
e the
handooff is produ
uced (Ballarrd et al., 20
003)
5 Distriibuted control: Work is planned in greater detail as yo ou get
closerr to execution, and pla
anning is doone collabo
oratively byy those
who are
a to do th he work. (Baallard et all., 2003)

Th
he indicators in the bo
ox above are transform
med into surrvey questio
ons:
Table 2: Survey que
estions mea
asuring Lasst Planner
# Queestions Answer tyype & Scoring Rule
1 Whaat percentag
ge of specialtty contractors participatted in Percentagge ⇒
sche
eduling the project
p phase(s) in whichh they were to do their None: 1/66; 0-25%:2/66;
work? 25-50%:3//6; 50-75%:44/6;
75-100%:55/6; and All: 1
2 To w
what extent was
w the prinnciple followed that onlyy work that Frequencyy ⇒
was ready to be e performed could be plaaced on a weeekly work Never: 1//5; Rarely: 2/5;
plan
n? Bear in miind that work is ready to
o be performed when all Sometime es: 3/5; Ofte
en: 4/5;
consstraints are removed.
r And Alwayys: 1
3 To w
what extentt was the principle
p folllowed that work Frequencyy ⇒
shou
uld be done
e in responsse to a requuest from an Never: 1//5; Rarely: 2/5;
imm
mediate cusstomer, suchh as the nex xt trade? Sometime es: 3/5; Ofte
en: 4/5;
And Alwayys: 1
4 Did the project measure the e extent to which
w you ‘did what you Yes/No ⇒
said
d you were go oing to do?’ (The measurre is the perccentage of Yes: 1; an
nd No: 1/6
weeekly work pla
an tasks com mpleted as planned. If the ere were
100 tasks on weekly work pllans and 70 were
w compleeted as
plan
nned (no parrtial credit), the percenttage would beb 70%)
5 Howw often weree reasons for not completting planned d tasks (on Frequencyy ⇒
weeekly work pla
an) analyzed to root causses and actio on taken to Never: 1/
/5; Rarely: 2/5;
prevvent reoccurrrence? Sometimees: 3/5; Ofte
en: 4/5;
And Alwayys: 1

So far, we
w have specified the measurement of the independen
i nt variable in the first
hypotheesis. Next, we
w address the depend dent variab
ble of the sa
ame hypoth hesis, proje
ect
performmance. We decided
d to use the summ of cost re
eduction rattio (%) (actual cost under
final ap
pproved buddget) + dura
ation reduction ratio (%) (actual duration
d relative to fin
nal
approveed schedulee) as a meassure of projject perform
mance beca ause of the low probab bility of
getting good data on other peerformance e dimensions.
Thhe depende e of the second hypoth
ent variable hesis is the same as the e dependen nt
variablee of the firsst hypothesis. The deppendent varriable of the e third hypoothesis is sa
ame as
the indeependent variable of thet first hyp pothesis. Annd the indeependent va ariable of thhe
second hypothesis is the same e as the ind
dependent variable
v of the third hypothesis. Thus,
T
the lastt concept th hat we define is the in
ndependent variable off the second and the third t
hypothe esis; i.e., to
o what exteent a projecct adopts In
ntegrated Project
P Delivvery (IPD), or
whether a project adopts IPD D. We decide ed to take the binary variable, whether
w a prroject
adopts IPD, as the type of this variable because
b wee could not get enough h IPD projeccts to
measure e the exten nt of implem
mentation. In addition n, it would be difficultt for respon ndents
to scoree the degree of adopting IPD structures if we e had used continuouss variables.

Lean Co
onstruction Journal 2011 p
page 71 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

Hypotthesis tessting methodolog

gy
The hyppothesis tessting was pe erformed diifferently aaccording too the type of
o variable. The
indepenndent variab ble (degree e of Last Pla
anner impleementation) of the firsst hypothessis is a
quantitaatively conttinuous ord dinal variable because the sum off scores of the
t five que estions
in Table
e 2 is the to
otal degree of Last Pla anner imple ementation of a projecct, represen nted as
a real n
number. The e dependen nt variable (cost reducction + timee reduction)) of the samme
hypotheesis is a ratiio variable representeed as a real number. Th hus, regression betwee en the
two varriables is ap
ppropriate for
f testing the
t hypothe esis. Howevver, the inddependent variable
v
of the second
s and the third hypothesis iss a binary categorical
c variable, ‘w
whether or not a
project adopts IPD D’, for which h regression
n analysis iss not appro
opriate. In this
t case, we
w used
a T-testt, to determ
mine whether the cate egorization (IPD or othe erwise) hass a significa
antly
differen
nt influencee on depend dent variables: projectt performan nce in the second
s hypoothesis,
and thee degree of implementtation of Last Planner in i the third
d hypothesiss.

Samplling Strattegy
In commmon sense, the most appropriate form of sam mpling to suupport a hyypothesis is
randomized sampliing. Howevver, Last Pla anner (LP) is
i a very speecific tool for
f producttion
control so that we need the very v specificc respondents who can n determine e the degre ee of LP
implementation in their proje ects. Thus, we decided d to use a purposive
p saampling tak king
advantaage of e-ma ail lists in re
elevant groups such ass general IGGLC group in n Yahoo4, or
particip
pants in worrkshops succh as those sponsored by b the Project Producttion System m
5
Laborattory . The same applie es to selectiion of IPD p
projects. If we were to o select proojects
randomly from anyywhere in the world, very v few, if any, IPD projects wou uld be inclu
uded.
Purposivve sampling g is widely used in stud dying unusuual critical cases. For example,
e itt can be
used efffectively in
n identifying g communitties across the
t United States thatt have voted d for
the winner in the past,
p or it is used in se
electing keyy informantts for ethnoographic stuudies
such as one describ bing gangstter’s lives (BBernard, 20000)

Results
Regre
ession mo
odel from
m testing
g the firstt hypothesis
There iss a significa
ant correlattion betwee
en the implementation n of Last Pla
anner (LP) and
a
project performance—the sum m of cost an
nd schedulee reduction percentage es. That meeans we
have successfully supported
s t first hyp
the pothesis. Th
his is repressented as a regression model
in Table
e 3 in the Appendix.
A
Figuure 2 is a grraphical rep
presentationn including scatter plootting and a linear regrression
line. Evven though we w used a straight
s line
e, the scattter plot see
ems to show w a curve is more
approprriate in describing beh haviour of variables. Thhus, we trieed several linear
l regreessions,
whose independen nt variables are ‘squaree of indepeendent varia able (X) in Figure
F 2’ orr X2 and
‘cube of X’ or X3
The result is en are of X’ orr X2 is ‘Y(Sum
ncouraging.. The regresssion model with ‘squa m of
cost red
duction and 7371101×X2-3.89088’ with
d time reducction) = 0.7 w its P<||t(2.98)| is 0.005,

4
http:/
//finance.dir..groups.yahoo
o.com/group/iiglc/message/
/677
5
http:/
//p2sl.berkeleey.edu/
Lean Co
onstruction Journal 2011 p
page 72 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

which iss less than 0.009,

0 the P<|t(2.71)| in Table 3 with mere e X. The lesss p value off t
(P<|t|) means there is greate er significan
nce in the coefficient
c of
o the regre ession line.
3 3
Furtherrmore, the regression
r m
model with X is ‘Y = 0.1484254×
0 ×X -1.617307’ with its p value
of t is 0
0.004, which an 0.005 in the regresssion model with X2. Bu
h is less tha ut, X to the fourth
does no ot show morre significan
40
30
20
10
0
-10
-20 nce than X3.

.5 1 1.5
5 2 2.5 3 3.5 4 4.5 5
Implem
mentation of Las
st Planner (X)

Sum
m of cost reducttion and schedu
ule reduction(Y) Y=4.141356**X-9.003641

Figu
ure 2: Regre
ession of Last
L Planner on Projec
ct Performa
ance
The finaal regressio
on line with X cubed, saying that the
t projectt performan nce is
proporttionate to thhe degree of
o Last Plan
nner’s impleementation n cubed, is visually
v
represe
ented as blu ue diamond type plots in Figure 3. We decide ed to call itt ‘Cho-Balla
ard
curve’, which show ws that Projject Perform
mance (sum m of cost reeduction and d schedule
3
reductioon) = 0.1484254 ×(Imp plementatioon of Last Planner)
P -1..617307.

Lean Co
onstruction Journal 2011 p
page 73 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

Cho-Ballard Curve
C
5
50
"more reduction of cost and schedule" Bottom to Top
Sum of Cost reduction and Schedule reduction: Y (%),

4
40

3
30

2
20

1
10

0
0 1 2 3 4 5 6

-1
10 Y=0.14
484254*X^3-1.6
617307

-2
20
X: Score
S of Last Planner Impllementation (LPI), "more regorous LPII" L->R

Figure 3: Cho-Ballard
C d Curve b/w nner)3 and Project pe
w (Last Plan erformance
e

Summ
mary of Hypothesi
H g6
is testing
The folllowing box summarizes the resultts of hypoth
hesis testing
g so far.

Hy
ypothesis 1
If a pro
oject implem
ments Last Planner (LP) more, itt achieves project
p
performaance betterr than those
e employingg LP less
=> Stron
ngly supporrted by the regression model: Prooject Perforrmance
(sum
m of cost redduction and
d schedule reduction) = 0.14842554 ×
3
(Imple
ementation of Last Plaanner) -1.61
17307
Hy
ypothesis 2
If a proje
ect adopts Integrated Project
P Delivery (IPD),, its perform
mance is
diffferent from
m those of other projeccts.
=> Failed
F to be supported
d definitive
ely
Hy
ypothesis 3
If a proje
ect adopts IIPD, its deg
gree of impllementation of LP is different
d
from thosse of other projects.
=> Fa
ailed to be supported.
s However, IPD
I projectts in our sam
mple
implemented LP to a certaiin degree evven though h the level is
i not
significcant statistically.

6
For de
etail of hypothesis testing, please see Ap
ppendix
Lean Co
onstruction Journal 2011 p
page 74 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

Conclusion
We founnd in this re
esearch tha
at project performance
p e improves with the im
mplementattion of
Last Pla
anner. Howe ever, we diid not find a strong rellationship among
a Last Planner, Prroject
Perform
mance, and Integrated Project De elivery (IPD)).
Thiss research does
d not pre
event us froom believinng that if IPD, aligning goals of
particippants, and LP,
L reducingg project va ariability, a
are combine ed, any pro
oject can acchieve
better pperformanc ce. Indeed, this is the claim
c put fo
orward by Lean
L Constrruction adh
herents,
criticizing forms off IPD that rely only on alignment of commerrcial interessts and
organiza ational inte
egration, whhile neglectting the leaan ‘operatinng system’, which addresses
how the e work is acctually donee. Future re
esearch is needed
n to validate
v thiss claim.

Appe
endix
Detaill of the first
f hypo
othesis te
esting
Table 3 is the resu
ult produced d by STATA v.10, a staatistics packkage, using data from the 49
projectss. Simply, we
w need to see the ‘co oefficient’, written onn the right side
s of ‘Y in
n Figure
2’ in Ta
able 3. This is the graddient of the regression line. Y is ‘ssum of costt reduction and
durationn reductionn’ and X is ‘the degreee of implementation off Last Plann ner’. The
significa
ance of thiss coefficien
nt is determ
mined by P > |t|, 0.009 9 (red-undeerlined number in
Table 3). Usually, if P>|t| is less
l than 0..05, we can
n say this co
oefficient (tthe regression
model) is significannt. In our case, the regression mo odel is Y=4..141356×X--9.003641
Table 3: Re
esult of Re
egression fo
or the first hypothesiss
Sourc
ce SS D
DF M
MS Number
N of object
o = 49
Mode
el 543.294059 1 543.294059 F(1, 47) = 7.36
Residu
ual 3467
7.23372 47 73.7709302 P
Probability > F = 0.0093
Tota
al 4010
0.52778 48 83.552662 R-squared = 0.1355
Adjjusted R-squ
uared = 0.117
71
Root Mean Squarre Error = 8.859
Y in Figu
ure2 Coefficient S Errors
Std T P>|t| 95% confidence
c In
nterval
X in Figu
ure2 4.1
141356 1
1.526046 2.71 0.009 1.071347 7.211366
Consta
ant -9.0
003641 5
5.279548 -1.77 0.095 -19.62
2472 1..61744

Detaill of the second

s hy
ypothesiis testing
g
The seccond hypoth hesis is < If a project adopts
a Integ
grated Proje ect Deliveryy (IPD), its
performmance is diffferent fromm those of other
o projeccts>. Beforee T test, we e needed to o see if
the twoo groups (IPD and Non IPD) have significantlyy different variance
v in project
performmance becau use generall T test is performed
p b
based on eqqual varianc ce. If not, T test
should be
b performed under th he unequal variance co ondition. Taable 4 is the variance ratio
test, na
amed as “sd dtest” in STTATA v.10. The
T f value stands for the ratio between
b the
e
variancee of IPD andd that of No on IPD, whiich is expreessed as ‘Raatio’ in Tablle 4. When the
probabiility, expresssed as p (FF<f), p (|F|>|f|), and p (F>f) in Table
T 4, is less than 0.005, the
alternattive hypoth
hesis, locate ed right aboove the pro obability, is chosen. In this test, the
t

Lean Co
onstruction Journal 2011 p
page 75 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

target a
alternative hypothesis is Ha: ratio o!=1. The probability right
r under the alterna
ative
hypotheesis is 0.084
43, which iss bigger thaan but near to 0.05 so that we cam
me to decidde to do
anotherr T test with
h unequal variance
v forr more assu
urance
Table 4: Variance
V Ra
atio Test on
n performance between IPD and
d otherwise
e
Grou
up Ob
bs. M
Mean Std. Err. Std. Dev. [95% Con
nf. Interval]
Non IPD 40 5
5.105027 1.55351 9.825258 1.962757
7 8.2472
297
IPD 9 4
4.160776 1.822258 5.466773 -.041357
77 8.3629
909
Com
mbined 49 4
4.931593 1.305816 9.140715 2.306073
3 7.557113
Ratio = standardd deviation (Non f = 3.230
02
IPD))/standard deviation (IPD
D) degrees of freedom = 39
Nulll hypothesis: Ratio =1 (=40-1), 8 (=9-1)
Alte
ernative Hypoothesis (Ha):: ratio <1 Ha: ratio !=
=1 Ha: Ratio > 1
Prob
bability: p (F
F<f)=0.9578 2*p (F>f) = 0.0843 P (F > f) = 0.0422

Table 5 is the resu

ult of T-testt with equal variance of
o STATA v.10. The ‘t’ value stands for
‘the rem
mainder of the Perform mance mean of Non IPD after sub btracted by the Mean ofo IPD’,
which iss expressed
d as ‘Differeence’ in Tabble 5. When n a probability, expresssed as p (T
T<t), p
(|T|>|tt|), and p (T
T>t) in Table 5, is lesss than 0.05,, the alternative hypotthesis, loca
ated
right ab
bove the proobability, iss chosen. In
n our case, the target alternativee hypothesiss is Ha:
Differen
nce!= 0, a different
d ex
xpression bu ut one having the same e meaning as that of our
o
second hypothesis.. p (|T| >|tt|) right beelow the altternative hyypothesis, Ha:
H Differen nce!=0,
is 0.782
28, much bigger than 0.05
0 so thatt we cannott choose the e alternativve hypothessis, our
second hypothesis
Ta
able 5: T-te
est with eq
qual varianc
ce on perfo
ormance be D and Non IPD
etween IPD
Group Obs. Mean
M Std. Err. Std. Dev. [95% Co
onf. Interval]
Non
n IPD 0
40 5
5.105027 1.55351 9.825258 1.962757
7 8.2472
297
IPD 9 4
4.160776 1.822258 5.466773 -.041355
57 8.3629
909
Com
mbined 9
49 4
4.931593 1.305816 9.140715 2.306073
3 7.557113
Diffe
erence .6551067 3.15756 -5.90614
44 7.7946
646
Diffeerence = Mean (Non IPD))-Mean t = 0.277
73
(IPD
D) degrees of freedom = 47
Nulll hypothesis: Difference = 0
Alte
ernative Hypo othesis (Ha):: Ha: Difference !=0 Ha: Diffeerence > 0
Diffe
erence < 0 p (|T|>|t|)) = 0.7828 P (T > t)) = 0.3914
Probbability: p (T
T<t)=0.6086

As we mmentioned, we did anoother T – tesst with une

equal variannce, whose result is sim
milar to
that of equal varia
ance. Unequ
ual T test saays the probability p (|T|>|t|) iss 0.6972, much
bigger than
t 0.05.

Detaill of the third

t hyp
pothesis testing
t
The thirrd hypothessis is <If a project
p adopts IPD, its degree of implementa
i ation of Lasst
Plannerr (LP) is diffferent from
m those of other projeccts> The varriance test said there is no
significa
ant differennce betwee en the varia
ance of the two groupss (IPD and Non
N IPD) in LP
implementation byy showing th he probabillity, used in
n determiniing whetherr to choose the
alternattive hypothhesis (standa ard deviatio
ons of the two
t groups are differe
ent), is 0.19
948,
Lean Co
onstruction Journal 2011 p
page 76 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

bigger than
t 0.05.
Table 6 shows the result ofo t test witth equal variance in te esting our third hypothhesis.
Similar to Table 5, if a probab bility right under the alternative
a hypothesis, representted as p
(T<t), p (|T|>|t|), and p (T>t), is less than 0.05, we w can choo ose the alteernative hypothesis,
located right above e the probaability. Our alternative e hypothesis is ‘Differeence (betwe een
means oof IPD and Non
N IPD)!=0 0’, a differeent expressiion but onee having thee same mea aning as
that of our third hyypothesis. Even
E though P (|T| >|t|), 0.074 is i bigger th
han 0.05, it is not
clear fo
or us whethe er to discarrd our thirdd hypothesiss. As for seccond hypothesis, it is clear
c in
that the
e probabilitty, P (T<t), is 0.7828, much
m bigge
er than 0.05 5. But, the third hypotthesis is
at the border.
b In short, even though Inte egrated Prooject Deliveery projectss do not shoow
implementation off Last Plann ner significaantly differe
ent from ottherwise, it seems to employ
e
Last Pla
anner to a certain
c degrree
Ta
able 6: T-te
est with eq
qual varianc
ce on perfo
ormance be D and Non IPD
etween IPD
Grou
up Obs. Mean Std. Err. Std. De
ev. [95% Conf. Interval]]
Non IPD 40 3.266667 .13213122 .835671
14 2.9994066 3.53927
7
IPD 9 3.801471 .1802989
9 .540896
68 3.385701
1 4.21724
42
Com
mbined 49 3.364896 .116053 .812370
07 3.131556
6 3.59823
36
Diffe
erence -.5348048 .2926632
2 -1.12356
67 .053957
76
Diffe
erence = Meaan (Non IPD)-Mean (IPD) t = -1.82
274
Null hypothesis: Difference = 0 degrees of freedom = 47
Alterrnative Hypo
othesis (Ha): Difference <0
< Ha: Diffe
erence !=0 erence > 0
Ha: Diffe
Prob
bability: p (T<t)=0.0370 p (|T|>|
|t|) = 0.0740
0 P (T > t) = 0.9630

Referrences
Adcock,, R., and Co ollier, D. (2
2001). Measurement va alidity: A shhared standard for qua alitative
an
nd quantita ative research. American Politicall Science Re eview, 95(3
3).
Alarcon, L. F., and d Cruz, J. C. (1997). “T The impact of project planning and Perform mance on
prroject outco omes”, Ten nth World Productivity
P y Congress. Santiago, Chile,
C October
Arbulu, R., Tomme elein, I., Waalsh, K., an
nd Hershaue er, J. (20022). “Contrib
butors to Lead time
in
n Constructiion Supply Chains:
C Casse of Pipe Suupports use ed in powerr Plants.”
Prroceedings of the 2002 2 Winter Sim mulation Co onference. Sandiego, CA, USA, De ec 8-11
Ballard,, G. (2000).. “The last planner sysstem of production con ntrol” PhD. Diss. Schoool of
Ciivil Eng., Unniv. of Birm
mingham, UK. Available e at www.le eanconstruction.org (2 26 Mar
20
010)
Ballard,, G. (2006).. “Rethinkinng project definition
d in
n terms of target
t costiing”. Proce
eedings
off the 16th annnual confeerence, Inteernational Group
G for Lean
L Construuction, San
ntiago,
Chhile.
Ballard,, G. (2008).. “LPDS Upd date”. Lean n constructiion journal,, 2008, pp. 1-19.
Ballard,, G., Iris To
ommelein, Lauri
L Koskela, and Greeg Howel (2002). “Lean n Constructtion
Toools and Te echniques”, In Rick Besst and Gera ard De Valen nche (Eds), Design and d
Coonstruction n building inn Value. Wooburn, MA: Elsevier Science. Pp. 227-255
Ballard,, G., and Ho owell, G. (22003). “An update on last planner.” Proceed dings of thee IGLC
1th , BlacKsb
11 burg, VA, USA,
U July 20
003, pp. 329 9-341.
Ballard,, G and Gre egory A. How well (2004). “Competiing Construcction Manag gement
Paaradigms”, Lean Consttruction Jou urnal, 1(1)338-45.
Bernardd, R. (2000). The socia al research methods-qu
m ualitative and quantita ative approaches.
CAA: Sage pub blication.
Lean Co
onstruction Journal 2011 p
page 77 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/
 C & Ballarrd: Last Plan
Cho nner and Inte
egrated Project Deliverry

Cho, Seeongkyun an nd Ballard, G. (2010), building

b an
nd argument with hypo othesis testting,
prroceedings IGLC-18, Ju uly 2010, Te echnion, Haaifa, Israel
Choi, Kuunhee (2008). “A new Decision Su upport Model for Innovvative Contracting Stra ategies
th
hrough a Qu uantitative Analysis
A on Aspects off Project Pe erformance” ” PhD. Diss. Civil
annd Environmmental Eng.., Univ. of California,
C B
Berkeley. C USA.
CA,
Chuck Thomson,
T Jooel Darringtton, Denniss Dunne, annd Will Lichttig (2009). Managing
In
ntegrated Prroject Delivvery. Consttruction Man nagement Association
A of America a.
Avvailable at http://cma aanet.org/ffiles/shared d/ng_Integrrated_Proje ect_Deliveryy__11-
199-09__2_.pd df (9 Jan 20011)
Cohen, Jonathan anda FAIA (20 010). Integrrated Projeect Deliveryy: Case stud dies. Americcan
In
nstitute of Architects
A C
California Council, Sacramento, CA,C USA.
Hamzeh h, Farook, R.R (2009). “Improving
“ C
Constructioon Workflow w-the role ofo Productio on
Pllanning andd Control” PhD.
P Diss. Civil
C and Envvironmenta al Eng. Univv. of Califorrnia,
Beerkeley, CAA, USA.
Liebermman.E. (2005). Nested analysis as a mixed method strattegy for com mparative
re
esearch. Am merican Poliitical Science review, 99(3)
Liu, M.,, and Ballarrd, G. (2008
8). “Improviing Labor Productivity
P y Through Production
Coontrol.” Prooc 16th Ann.. Conf. Int’l Group of Lean Consttr., IGLC 16 6, 16-18 Julyy,
Ma anchester, UK.
Nationaal Associatioon of State Facility Asssociation (N
NASFA), Con nstruction Owners
O Assoociation
off America, Association
A of Higher Education
E F
Facilities Offficers, Asso
ociated Gen neral
Coontractors ini America,, and Ameriican Institutes of Archiitects (2010 0). “Integraated
Prroject Delivvery for Pubblic and Privvate Ownerrs” Availablle at
htttp://www.agc.org/ga alleries/proojectd/IPD%%20for%20Pu ublic%20and d%20Private e%20Ow
neers.pdf (10 Jan 2010)
Sanvidoo, Victor, annd Konchar,, Mark (1998). Project Delivery Syystems: CM M at Risk, De esign-
Buuild, and De esign-Bid-B
Build. Consttruction Industry Institute Design Build Resea arch
Teeam 133, Austin, Texa as, USA.

Lean Co
onstruction Journal 2011 p
page 78 www.leanconstructionjournal.org
http://creativecommonss.org/licenses/by--nc-nd/3.0/