pull-driven construction of high-rise apartment …

PULL-DRIVEN CONSTRUCTION OF HIGH-RISEAPARTMENT BUILDINGS

Rafael Sacks1, Maxim Goldin2 and Zvika Derin3

ABSTRACT

Construction of high-rise apartment buildings is made complex by the myriad possibilities for clientsto adapt their apartments to suit their individual needs and preferences; traditional construction plan-ning practice of progressing upwards from floor to floor breaks down in the face of the arbitrary se-quence in which clients finalize their decisions. The results are long cycle times for delivery ofcompleted apartments and corollary high levels of work in progress (WIP), budget and schedule over-runs, and general dissatisfaction with the process on the part of the contractors, subcontractors and theclients. This paper presents a management model that applies lean thinking to this problem. The modelwas first formulated in theory, then tested using a management simulation game, and subsequently de-veloped for practical application by a dedicated team composed of university researchers and con-struction company personnel. It is now being tested in a large construction company.

KEYWORDS

Pull-driven flow control, High-rise buildings, Residential construction.

INTRODUCTION

The ability to customize apartments for individualclients is an absolute requirement in any free-market economy where developers compete forclients. In any housing construction project otherthan single-family dwellings, and particularly inhigh-rise construction, the building is usuallycommenced—and often completed—well beforethe last apartment is sold. Customization demandsthe ability to respond rapidly to late changes indesign, because home-owners’ design decisionsare only made relatively late in the productionprocess (always after the apartment is sold).

However, traditional construction managementscheduling and accounting practices are relativelyinflexible. Predetermined and optimized con-struction schedules and centralized control struc-tures result in significant and inevitable waste.Frustration for all involved is common, with highrates of litigation resulting. The conflict betweenthe need for customized products and rigid

approaches to production planning, with resultinginherent waste, is reminiscent of the problemsfaced by many manufacturing industries thatturned to lean principles in search of improvedmanagement approaches (Womack and Jones1998).

Housing forms a major part of the constructionindustry in most developed economies, and it hasbeen the focus of a number of studies that haveadopted a lean perspective. Gann (1996) com-pared industrialized housing with car manufacturein Japan, highlighting similarities and differencesin production and supply chain strategies,although none of the industrialized housing com-panies investigated constructed high-rise build-ings. Bashford et al. (2003) examined the impactof time-gating strategies adopted by US housingdevelopers, exposing very long cycle-times andlarge inventories of work in progress (WIP) thatresulted from extreme optimization of the individ-ual activities for single-family houses. The rela-tionship between cycle time and WIP predicted by

Production Planning and Control

217

1 Senior Lecturer, Faculty of Civil and Env. Eng., 840 Rabin Building, Technion – Israel Institute of Technology,Haifa 32000, Israel, Phone +972-4-8293190, Fax +972-4-8293190, [email protected]

2 Graduate Student, Faculty of Civil and Env. Eng., Technion – Israel Institute of Technology, Haifa 32000,Israel, [email protected]

3 Head, Project Planning and Control Dept., Danya-Cebus Construction Ltd., Tel Aviv, Israel.

Little’s Law (Hopp and Spearman 1996) has beenshown to hold at a macro project level for custom-ized housing projects (Bashford et al. 2005).Naim and Barlow (2003) proposed the applicationof lean and agile approaches to housing construc-tion in the UK, but focused on supply chains anddid not tackle the fundamental construction plan-ning and control practices. Ballard (2001) sug-gested multi-skilled teams as a key factor inachieving even-flow production for reducingcycle-times and enhancing stability forconstruction of single family houses.

In this project, a holistic lean construction man-agement strategy based on pull flow schedulingwas developed for high-rise customized apart-ment buildings. It is fundamentally different to thetraditional strategy common in most companiesactive in this sector. A set of tools, techniques andtechnological changes was then developed by acollaborative academic and industry team in orderto apply the strategy in practice. In this paper weanalyze the traditional approach and its problemsin terms of lean principles, and present the alter-native strategy. A management game was used toillustrate the problems, and it is presented briefly.Key components of the implementation plan arealso presented.

A LEAN ANALYSIS OF TRADITIONALPRACTICE

The main source of conflict and waste in tradi-tional high-rise residential construction arisesfrom the need to customize the apartments to thespecific design specifications of each home-owner (Rosenfeld and Paciuk 2000). Traditionalconstruction schedules call for progress of tradesthrough a building in a vertical ascending direc-tion, closely following the erection of the struc-ture itself. Conventional wisdom also dictatesthat, in order to optimize productivity, all of thework of any particular processing step on a floorshould be completed as a single work package.This is comparable to the desire to process largebatches on machines with setup times inmanufacturing settings.

However, apartments are not sold in thesequence of their location, and therefore the flowof information from client is out of sync with theneeds of the construction process. Most compa-nies deal with this situation by attempting tocoerce clients to make design decisions regardinglocation of partitions, selection of finishes of vari-ous types, etc., in accordance with the overall con-

struction schedule. When design decisioninformation is delayed or unavailable, assump-tions must be made for work to continue. This isof course inevitable where apartments are not soldahead of construction. The magnitude of theimpact is greatest for high-end apartments, wherethe degree of freedom for customization isgreatest.

Client-initiated changes are therefore executedlargely as change-orders. Change-orders havebeen shown to have a strong negative impact onlabor productivity (Moselhi et al. 2005), result inwasted materials. As a result, changes are invari-ably priced to clients at rates significantly higherthan standard construction costs (Rosenfeld andPaciuk 2000). In one reported case, a companycharged clients a demolition fee for partitions thatappeared on the standard plan even if the parti-tions were not yet erected! Where changes are sig-nificant and disrupt the standard process, clientsare dissuaded from making them irrespective ofthe price; this represents a lose-lose situation.Overall, clients’ perceptions are negative—theyare not receiving value.

At the outset of this study, statistics were col-lected from numerous projects of a large housingcontractor (2,300 units annually). The averagecycle times for the interior finishing works4 onindividual apartments, listed in Table 1, werefound to be very long when compared with the nettime required to perform the work (12 weeks). Inalmost every project, the WIP was 100% of thetotal production for a significant period of the pro-ject’s life, i.e. all of the apartments were workedon simultaneously. Batch size in every projectwas determined by the full number of apartmentson each floor, commonly 4.

Table 1: Average cycle times for individual apartmentsmeasured in high-rise housing projects.

Cycle-time MeasureAverage

(weeks)

StandardDeviation

(weeks)

Start of finishing works tohandover 49.1 8.4

First client change meeting tohandover 50.4 12.0

Last client change meeting tohandover 24.9 15.4

Duration of change definitionprocess—first to last clientchange meetings

25.5 17.5

Proceedings IGLC-13, July 2005, Sydney, Australia

218 Pull-driven construction of high-rise apartment buildings

4 Interior finishing works include, amongst others, acoustic insulation, flooring, electrical, plumbing and HVACsystems, gypsum board partitions and ceilings, waterproofing, painting, kitchens, ceramic tiling, and installationof doors.

Project managers on high-end projects reportedspending up to 60% of their time managing clientchanges and the resulting complexity of instruc-tions to specialty contractors. As changes becamedelayed due to late change orders, managementteams lost control of the flow of work; contrac-tors, always searching for high rates of productiv-ity (Sacks 2004), continually redirected theirefforts to those apartments in which larger quanti-ties of work were available, leaving many minordetails unfinished in incomplete apartments. Thisappeared to be a major source of high WIP, asapartments with little work left were simply notcompleted. Push scheduling was another impor-tant cause of high WIP, since work was begun inapartments that either had not been sold or whereclient changes had not been confirmed, with theresult that stoppages in their execution were inevi-table. In the latter stages of projects, contractorsreduced the sizes of their teams as instability andunpredictability in their workloads increased.

The state of affairs described above can becharacterized in terms of waste, value and flow.Firstly, all of the types of waste defined byWomack and Jones (Womack and Jones 1998) arepresent, as listed in Table 2.

An extensive survey conducted among clientsfrom the contracting company’s previous projectsrevealed that the existing process is not providingmaximum value. The range of changes isrestricted, prices for changes are considered to beexorbitant, decisions are demanded too early,multiple design consultations are required at dif-ferent locations, design choices and changes arenot accurately executed, and handover dates areunreliable if changes are made. From the point ofview of project developers (the company’s imme-diate clients), value is delivered when the designdecisions for unsold apartments are delayed as faras possible, extending the time during which theycan sell apartments and offer maximumcustomization flexibility to potential buyers.

Lastly, in terms of flow, work on apartments isnot continuous; WIP levels are high, often reach-ing the full complement of apartments in a build-ing, and cycle times are up to four times as long asthe net continuous work time required for com-pleting the finishing works in apartments.

PULL SCHEDULING MODEL FOR HIGH-RISE APARTMENTS

The goals of the process change initiative were a)to increase value for the clients by reducing thecost of customization while extending the time

during which design decisions could be made andreducing the cycle time for customizing eachapartment, and b) to reduce the direct costs andmanagement overheads for the general contract-ing company. Analysis of the existing manage-ment practice led to formulation of the followingprinciples for achieving improved productionflow:

• Abandoning conventional push schedulesbased on progress vertically up the floors ofthe building in favor of pull scheduling ofwork teams from apartment to apartmentdriven by completion of client change deci-sions,

• Reduction of handovers between work pack-ages5 and reduction in the number of distinctwork packages through increased use ofmulti-skilled teams and changes to construc-tion methods,

• Restructuring of the work to decouple stablework packages5 from those dependent on latechange information,

• Changes to the client change coordinationprocedures to improve communication andplanning reliability.

Once these principles were crystallized, a man-agement simulation game was devised to test thepotential benefits of a management model basedon the principles. The game was loosely based onthe airplane game (Verma 2003) which explainsthe impact of one-piece pull flow in a manufactur-ing setting. The simulation game was later used toaid in presenting the management model to con-struction industry professionals at all levels. Thegame and its results are the subject of thefollowing section.

APARTMENT CONSTRUCTIONSIMULATION GAME

The lean apartment construction simulation gamesimulates construction of an eight story buildingwith four apartments on each floor. Participantsare assigned the roles of project manager, a clientchange manager, a quality controller, a towercrane operator, and four specialty contractors.Their task is to carry out the interior finishingworks for all 32 apartments in as short a time aspossible. Execution of the finishing works is sim-ulated using assembly of small building modelsusing LEGO® bricks (see Figure 1); they areassembled in four distinct steps, each performedby one of the specialty contractors. Two addi-tional players represent the apartment clients—the first selects design variations from a prede-


Rafael Sacks, Maxim Goldin and Zvika Derin 219

5 A work package is defined as an activity performed in a defined area by a distinct work team. A stable workpackage is one that is not dependent on client preferences, i.e. it cannot be customized.

fined set of seven variations to the basic design,and the second checks completed apartments andpays $1,500 for each one that is free of defects.

In the first round of play, which lasts 11 min-utes, the project manager is provided with a sug-gested construction plan which calls for thespecialty contractors to progress up the buildingfloor by floor, one after the other, in logicalsequence according to the technological depen-dences of their work type. The batch size is four—only one contractor may work on a floor at a time.At set time intervals during construction, the cli-ents select design variations by pulling apartmentnumbers and variation codes at random from a hatand deliver them to the ‘company’. Design varia-

tions received before work on an apartment iscommenced are easily executed, although they dodisrupt steady flow because they require moretime. Apartments for which design variations arereceived after work has begun but before theapartment has been delivered to the client, must bechanged. The project manager must decidewhether to withdraw specialty contractors fromthe floors they are working on and send them tomake local changes, or to delay the changes. Thedriving factor is that the company is only paid forcompleted apartments, while they must investworking capital ($1,000) for any incompleteapartment. Play is stopped after 11 minutes, andthe team’s performance is assessed in terms ofapartments delivered, quantity of WIP, cash flow,defective apartments, and the time required todeliver the first non-standard apartment. Theresults measured through distinct executions ofthe game with seven different groups are providedin Table 3.

In the second round, the following changes aremade:

• pull scheduling replaces push scheduling—the work sequence is changed to match thesequence in which design variations are se-lected to be pulled by the clients’ selectionsof design variations (work is only begun onapartments once the design variations are se-lected),

• the four specialty contractors are replacedwith four multi-skilled apartment finishingteams that can perform all of the activitytypes.

All other conditions remain as before. Typicalresults for this round, obtained by the same sevengroups, are listed in Table 4. As can be seen, pro-ductivity is approximately doubled, WIP is drasti-cally reduced (as in all pull systems, it is explicitlycontrolled) as is cycle time, and as a result, cashflow turns from negative to positive.



Waste Observations

Undesiredproducts

Apartments built to standard designs areless attractive to potential buyers

Rework

Client changes performed as changeorders require demolition of workcompleted earlier; management effort isrequired to coordinate late change ordersand to control their execution

Inventories Inventories of completed (but not yetpurchased) apartments are accumulated;

Unnecessaryactivities

Unfinished apartments must be cleanedand repaired after periods during whichthey are not worked on.

Unnecessarymovement ofworkers and/ormaterials

Work stoppages are frequent whenapartments are sold during finishing toallow time for clients to reach designdecisions—specialty contractors areforced to move to other apartments andthen back again later.

Waiting formaterials orinformation

Delays due to unavailable informationreduce productivity;

Products thatdo not meetclients’ needs

Apartments built to standard designs donot fully meet clients’ needs; clients oftenforego customization where the cost ofchange-orders is considered prohibitive.

Table 2: Waste in traditional management ofconstruction of customized high-rise apartment buildings.

Figure 1: A standard apartment.Assembly steps represent flooring (orange), partitions (blue), electrical wiring (yellow) and acoustic ceilings (brown).

DEVELOPMENT OF ANIMPLEMENTATION PLAN

In this phase, the researchers were joined by con-struction company representatives to form a taskgroup charged with formulating specific steps forimplementation, first in a pilot project and laterfor all similar projects. The task group includedthree university researchers, the company vice-president for control systems and informationtechnologies, a senior company engineer, the headof the client change service department, a clientchange representative, and the pilot project sitestaff—project manager, construction engineerand the senior works supervisor. The team metweekly over an extended period to propose, evalu-ate and develop the implementation. The key fea-tures of the implementation decided upon areexplained in the following sections.

SEPARATION OF FIXED WORK PACKAGESFROM THOSE SUBJECT TO CHANGES

The first step is separation and disconnection ofwork packages with high degrees of certainty(structure, public areas and building facade) fromthose susceptible to change (partitions, systemsand finishes within apartments). This requiredchanges to basic features of building design, suchas removal of all water and electrical conduitsfrom flooring layers to walls and suspended ceil-ing systems, so that they could be installed as lateas possible in the process. In some cases, makingthese changes demanded cooperation of the devel-oper and the design team. The developer’s coop-eration was obtained in return for relaxation of thecontract requirement that the developer providedesign decisions as construction progresses up thebuilding regardless of whether apartments weresold or not (this requirement is standard practiceunder traditional contracts for high-rise apartmentconstruction). The potential of technologicalchanges to increase flexibility, through delayeddependence of the production process on informa-



Group Apartmentscompleted

Defectiveapartments

Time to firstapartment

(min.)W.I.P. Throughput

(units/min) Cash flow Cycle time(min.)

A 8 0 5:15 22 0.84 ($18,000) 5:30

B 6 0 7:30 14 0.80 ($11,000) 6:15

C 12 2 6:00 12 2.00 ($6,000) 4:00

D 7 1 7:30 10 0.86 ($6,500) 7:30

E 7 1 5:50 12 0.55 ($8,500) 5:50

F 8 0 8:00 13 2.00 ($9,000) 7:00

G 8 0 6:00 11 0.80 ($7,000) 6:00

Average 8.0 0.57 6:35 13.4 1.12 ($9,429) 6:00

Table 3: Lean apartment simulation game results—round 1.

Group Apartmentscompleted

Defectiveapartments

Time to firstapartment

(min.)W.I.P. Throughput

(units/min) Cash flow Cycle time(min.)

A 12 0 4:15 2 2.32 $4,000 2:00

B 18 0 4:00 4 3.58 $5,000 1:00

C 17 0 2:00 2 2.29 $6,500 1:00

D 15 0 2:30 3 1.87 $4,500 1:30

E 14 0 2:00 1 1.63 $6,000 1:00

F 16 0 2:30 2 2.31 $6,000 1:30

G 17 0 2:30 2 2.13 $6,500 1:30

Average 15.6 0.00 2:49 2.3 2.30 $5,500 1:21

Table 4: Lean apartment simulation game results—round 2.

tion, is measured with a simple ‘rigidity index’.The index is calculated using the formula:

( )RI N j Dii

n

i= −=

∑1

where i = client design decision index;n = number of client design decisions;ji = serial number of work package by whichdesign decision i must be made;N = total number of work packages;Di = relative impact of design decision i. A reduc-tion of the RI indicates that a process has beenmade more flexible. A hypothetical process withmaximum flexibility (i.e. all design decisions canbe made at the time of the last finishing workpackage would have RI = 0; a process thatdemanded all design decisions before its start

would have RI NDii

n

==∑

1

. In the pilot project, the

RI was decreased from 1258 to 979.

PULL SCHEDULING

The second step entails strict pull scheduling ofchange-susceptible work packages, controlledusing completed client change files as ‘Kanban’cards. A set of criteria were established for assess-ment of the ‘maturity‘ of the apartment files,including the degree of certainty that the clientchanges were final, a check of the availability ofall the materials selected, completeness of thebuilding frame, and availability of the tradesneeded. Apartments are released for constructionin order of their degree of maturity. The client-change representative was given the task ofmaking apartments ready by sorting clients into‘decisive’ and ‘indecisive’ groups, and schedul-ing their milestones for making product selectionsaccordingly. The degree of maturity of each apart-

ment was assessed in weekly meetings of the sitemanagement together with the client-change rep-resentative (and specialty contractor representa-tives where necessary), and the highest rankingapartments were selected for work to start.Figure 2 shows an excerpt of a status report thatwas implemented. It draws data from the com-pany’s existing client changes, project control andpurchasing software systems.

The TAKT time and the rate of release of apart-ments for finishing works depend on the quantityof apartments sold at any point in time. The basicrequirement is that all apartments sold should beavailable for occupation at the time that the build-ing as a whole receives a certificate of completionfrom the local building authority. Unsold apart-ments could be completed after that time (seebelow). The initial start date for the interior finish-ing works and the desired TAKT time are interde-pendent. The starting rate for the pilot project wasinitially set at 2 apartments per week, based on theneed to deliver 40 of the 52 apartments in the pro-ject within 32 weeks, with a cycle time of 12weeks.

CHANGES TO THE PRIME CONTRACT

The construction contracts between the developerand the general contractor (GC) were amended toincorporate pull flow driven by client changes.The team determined that value to the developercould be increased by allowing design decisionsfor unsold apartments to be postponed as far aspossible. In traditional construction, developerswere required to make decisions in time for apart-ments to be completed and the building handedover as a whole. In the lean implementation,where apartment finishing sequence is no longerconstrained to vertical progression of work teams,finishing of unsold apartments could be delayedbeyond initial occupation of the building. Con-


222 Pull-driven construction of high-rise apartment buildingsAp

t#

Owne

r'sNa

me

Stru

ctur

e

Layo

ut

Kitc

hen

plan

Plum

bing

cons

ulta

nt

Elec

tric

ityco

nsul

tant

HVAC

cons

ulta

nt

Door

sele

ctio

n

Floo

ring

sele

ctio

n

Fini

shm

ater

ials

avai

labl

e

Part

ition

plan

final

ized

Task

com

plet

ion

9 TS 100% Complete19/02/05

Complete19/02/05

Complete19/02/05

Complete12/03/05

Complete19/02/05

Complete12/03/05

88%

10 100% 0%11 CO 100% Complete

15/01/05Complete15/01/05

Complete15/01/05

Complete15/01/05

Complete15/01/05

Complete26/02/05

Complete26/02/05

Complete26/02/05

Complete26/02/05

100%

12 SH 100% Complete20/02/05

Complete20/02/05

Complete20/02/05

Complete7/03/05

Complete7/03/05

50%

13 100% 0%14 OF 100% Complete

26/02/05Complete26/02/05

Complete26/02/05

30%

Building: Savyonei Ramat Aviv 2 31/03/05Status report

Figure 2: An excerpt from a weekly apartment file status report.

struction methods that required fewer bulk mate-rials caused little noise, generated minimalmaterial waste, and required minimal materialdeliveries, were identified. The GC obtained thedeveloper’s agreement to cover additional costsfor these methods in return for the marketing flex-ibility it provided. Provisions were also made forthe GC’s release from obligation to complete inte-rior finishes for apartments whose completion thedeveloper preferred to postpone beyond the end ofthe major works.

WORK RESTRUCTURING

Both before project start and during execution,finishing works are restructured to enhance flow,though intrinsic improvement of the productionprocess itself. Changes in technology and/or workmethods aim to reduce the number of subcontrac-tors and the number of handover points betweenthem. A ‘complexity index’ was devised to enablecomparison of alternative process improvementstrategies. The measure is calculated as the prod-uct of the number of distinct work teams multi-plied by the number of interfaces between workteams. Figure 3 illustrates the concept with asimple example extracted from a full constructionprocess analysis, in which two alternatives for theprocess of flooring are compared. The lower theindex, the more streamlined the workflow.

REDEFINING MANAGEMENT ROLES

The role of the company’s ‘client-change repre-sentatives’ is upgraded to that of ‘interior-finishmanager’, which requires their relocation fromcompany headquarters to the construction site,retraining, and results in their becoming an inte-gral part of the site management team. They carryresponsibility for final construction coordinationand provide a direct link for information flowbetween the clients and the specialty contractors,

and play a key role in making apartment filesready. The project manager is relieved as far aspossible from dealing with client changes andinterior work, allowing him or her to concentrateon managing the project as a whole.

PULL-DRIVEN FLOW CONTROL

Once apartments have been introduced into theproduction system, flow is controlled using a pub-licly displayed electronic workflow progress chartto pull work teams (mostly specialty contractors)to the appropriate work packages, to prevent exe-cution of work out of sequence, and to preventcommencement of work that was not ready forrelease. A portion of the chart, termed the PullFlow Control (PFC) chart, is shown in Figure 4. Itshows the activities on the vertical axis; the activi-ties are drawn directly from an MsProject plancompiled for each apartment. As apartments areselected for execution, they are added to the hori-zontal axis. The state of each work package isevaluated and published using an appropriatesymbol. The chart has a number of benefits thataid the work teams and the project managers tomaintain steady flow:

• The path of the work teams through thebuilding, determined by the pull of ‘mature’client files, is clearly visible, as is the statusof work in every apartment.

• Bottlenecks can be identified easily—theyare activities behind which numerous ‘greenlight’ (ready for execution) activities accu-mulate horizontally.

• Work teams can clearly see the buffer ofwork being made available to them, as eachwork package ahead has its status clearly in-dicated, and so they can adjust their resourceassignments to the project appropriately inadvance and with a high degree of certainty.

• Work teams are prevented from enteringapartments before approval of completion of



Install gas line 1.GasContractor Install gas line 1.Gas

Contractor1 1Install waterpipes in floor

2.Plumber

ACTIVITY

CONTRACTOR

HANDOVERInstall

soundproofing2.Floorer

2Install

soundproofing3.Floorer

Before Pour Screed 2.FloorerAfter

3 Contractors: 4 Contractors: 3

Spread base 4.SandPump Handovers: 4 Install flooring 2.Floorer

Handovers: 24 2

Install flooring 3.FloorerComplexityIndex: 16 Install water

pipes in walls3.Plumber

ComplexityIndex: 6

Figure 3: Construction method ‘Complexity Index’.

the previous stage and before official ‘re-lease’ of the work by the ‘last-planning’team. This prevents ‘making-do’ (negativebuffering) (Koskela 2004) and enables main-tenance of single piece batch flow.

An additional benefit of the PFC chart reported bythe company was that the ‘red light’ was effectivein helping implement in-process quality control.

DISCUSSION AND CONCLUSIONS

Analysis of current construction managementpractice for high-rise apartment buildings in amajor Israeli construction company under themicroscope of lean thinking revealed significantwaste, client dissatisfaction with the value pro-vided in both the process and the product, highinventories of WIP and interrupted flow. The con-clusions of the analysis are supported by country-wide statistics and previous research that investi-gated the problems related to customization ofapartments. A new, lean construction manage-ment model was proposed by the researchers toreplace the existing practice. The various featuresof the model are designed to improve process flowand reduce cycle time. They represent a signifi-cantly different approach to scheduling the inte-rior works in high-rise apartment constructionthan the critical path method based approach. A

simulation game was devised to experiment withthe proposed method—its results proved that themodel had the potential to improve productionrates while at the same time delivering custom-ized apartments with no waste and reduced cost. Italso indicated that the company’s cash flowwould be drastically improved.

A joint university-industry team then devel-oped the model for implementation in a pilot pro-ject. The key principles adopted in theimplementation included pull-driven schedulingof apartment finishing works, technologicaldecoupling of stable work packages from workpackages subject to client changes, streamliningof work packages and handovers to improve flow,and redefinition of the role of the company’sclient change representatives. Not all of the prin-ciples of the theoretical model could be imple-mented; for example, the practice of extensivesub-contracting as much of the construction aspossible prevented employment of multi-skilledwork teams. The option of performing work usingsub-contracted multi-skilled teams was rejectedby the company based on the assumption that itwould reduce its own profit margin.

Despite initial skepticism at all levels, the per-sonnel involved became committed to the leanmodel once they had the time to think and reflecton existing conditions. Obstacles to implementa-tion were raised, but creative solutions were often



Figure 4: Electronic Pull Flow Control (PFC) chart.

found. For example, the ceiling work crew ques-tioned the practicality of moving their equipment(scaffolding etc.) from apartment to apartment,especially through finished public areas. Thesolution was to provide two sets of equipment andtransport them in specialized containers from bal-cony to balcony using the tower crane.

Three factors appear to be important inenabling this company to develop and implementa fundamental change in its management practicebased on the lean management model proposed bythe researchers:

• The specific techniques to be implementedwere crystallized over an extended period bya multidisciplinary team composed mainly ofcompany personnel, including the full sitemanagement team of the pilot project. Directparticipation of company personnel in devel-oping the procedures was essential not onlyfor application of the theoretical model underthe constraints of real project conditions, butalso to provide the team members themselveswith the opportunity for reflection and ‘un-learning’ of conventional wisdom. They alsogradually adopted ownership of the solutionsthat were reached and developed commit-ment to their success.

• No ‘complete’ or packaged solution was of-fered up front by the research team; rather, aset of principles for change was provided, to-gether with a set of measures that aided pre-diction of the outcomes of changes that wereproposed. The management simulation gameproved to be a powerful tool in this regard.The flexibility and complexity indices pro-vide a quick and structured method to deter-mine whether any proposed technical ororganizational change is likely to improveprocess flow.

• The tools for flow control were embedded inthe company’s computerized informationsystems. The first tool was a simple but com-prehensive summary report on the status ofeach apartment’s ‘maturity’ for execution,and the second was the pull-flow controlchart. Both were provided to all project par-ticipants; the company plans to make themavailable to subcontractors on the projectextranet, and to site supervisors on internet-enabled tablet PCs.

Construction of the pilot project building is nowunder way. The only aspect of the model imple-mented fully in this project is pull flow control,because detailed design was complete and spe-cialty contracts had already been awarded whenthe task group finalized its implementation plan.Detailed data are being collected, with the inten-tion of measuring the impacts of the method.

Early results suggest that substantial improve-ment requires holistic implementation, althoughmuch work is still necessary to establish whichaspects of the model can be implemented in prac-tice, what their impacts are, to what degree theyare interdependent, and where the hurdles tochange exist.

ACKNOWLEDGMENT

The authors gratefully acknowledge the contribu-tions of the joint Technion-Industry Lean devel-opment team whose professional knowledge,creative thinking and concerted teamwork con-tributed to refinement of the management meth-odology for application in practice.

REFERENCES

Ballard, G. (2001). “Cycle Time Reduction inHome Building.” 9th International Group forLean Construction, D. Chua and G. Ballard,eds., Singapore.

Bashford, H. H., Sawhney, A., Walsh, K. D., andKot, K. (2003). “Implications of Even FlowProduction Methodology for U.S. HousingIndustry.” J. Constr. Engrg. and Mgmt.,129(3), 330-337.

Bashford, H. H., Walsh, K., and Sawhney, A.(2005). “Production System Loading–CycleTime Relationship in Residential Construc-tion.” J. Constr. Engrg. and Mgmt., 131(1),15-22.

Gann, D. M. (1996). “Construction as a manufac-turing process? Similarities and differencesbetween industrialized housing and car pro-duction in Japan.” Construction Managementand Economics, 14(5), 437-450.

Hopp, W. J., and Spearman, M. L. (1996). FactoryPhysics, IRWIN, Chicago.

Koskela, L. (2004). “Making Do—The EighthCategory of Waste.” 12th Annual Conferenceon Lean Construction, C. T. Formoso and S.Bertelsen, eds., Elsinore, Denmark.

Moselhi, O., Assem, I., and El-Rayes, K. (2005).“Change Orders Impact on Labor Productiv-ity.” J. Constr. Engrg. and Mgmt., 131(3),354–359.

Naim, M., and Barlow, J. (2003). “An innovativesupply chain strategy for customized hous-ing.” Construction Management and Eco-nomics, 21, 593–602.

Rosenfeld, Y., and Paciuk, M. (2000). “Classifi-cation of client change-orders and analysis oftheir impact on construction efficiency.”Report #69.008, National Building ResearchInstitute, Technion, Haifa (in Hebrew).



Sacks, R. (2004). “Towards a lean understandingof resource allocation in a multi-project sub-contracting environment.” 12th Annual Con-ference on Lean Construction, Elsinore, Den-mark.

Verma, A. K. (2003). “Simulation Tools andTraining Programs in Lean Manufacturing—Current Status.” NSRP–ASE—0301–11, OldDominion University.

Womack, J. P., and Jones, D. T. (1998). LeanThinking: Banish Waste and Create Wealth inYour Corporation, Touchstone.



pull-driven construction of high-rise apartment …

Documents