Professional Documents
Culture Documents
Chang Ibbs 2006 System Model For Analyzing Design Productivity
Chang Ibbs 2006 System Model For Analyzing Design Productivity
Abstract: Many engineering design companies collect data such as person hours to manage projects. But the relationships between
Downloaded from ascelibrary.org by The City College of New York - CUNY on 12/14/23. Copyright ASCE. For personal use only; all rights reserved.
operational variables and performance are usually not thoroughly analyzed and interpreted. This paper proposes a system model and
procedure to relate influence variables to project productivity. The model was tested by analyzing 190 projects of an engineering
consulting company. The relationships between design productivity and various input and process variables were identified and inter-
preted. For example, project size has a negative relationship with productivity, while the effect of quality assurance/quality control on
productivity is not clear. Based on documented data and derived information, this model can help companies gain operational insight and
thus improve productivity and profitability.
DOI: 10.1061/共ASCE兲0742-597X共2006兲22:1共27兲
CE Database subject headings: Information management; Productivity; Performance characteristics; Engineering firms; Project
management.
quantitative nature, such as duration and actual hours 共Eldin 3 and 4 will help management study and act appropriately.
1991兲. 5. Analyze I-P relationships. This analysis helps identify how a
There are also implicit variables that influence productivity. company deploys the work team or adopts management
They are qualitative in nature and cannot be easily recorded in a methods for a given work. The purpose is to examine a com-
database, such as I2 and P2, expressed in the shaded portions of pany’s decision-making relevance at the early stage of work.
Fig. 1. Work nature 共I2兲 includes variables such as task character- 6. Link I-P-O relationships. Synthesize the preceding relation-
istics, task interdependence, and possessed information. The five ships and examine the combined effect of various variables
variables listed in Fig. 1 are the sources of work uncertainty and to identify dominant productivity-contributing variables.
equivocality 共U&E兲 that describe work characteristics to a great 7. Propose the company’s productivity knowledge. Disseminate
extent 共Chang 2001兲. They are used as proxies to measure work this knowledge so that all managers in the company can
nature. utilize this information profitably.
The management category 共P2兲 includes coordination, stake-
holder needs, schedule effectiveness, etc. These process variables
have been proven to improve project performance 共Chang and
Case Study and Exploring I-P-O Relationships
Ibbs 1998; Kuprenas 2003兲. They are adopted here to remind
management of their importance.
The preceding model and procedure was tested to prove its valid-
ity. This research mainly analyzed quantitative data of an engi-
Implementation Procedure neering consulting company’s projects and interviewed relevant
project managers to supplement the quantitative analysis. One
A concept has to be implemented and tested in order to verify its hundred ninety projects completed between 1996 and 2001 in the
value; thus, a procedure was developed to test the proposed company’s database contained data of eight I1 and 11 P1 vari-
model. It involves two steps to prepare the data and five steps to ables, and parts of them are listed in Fig. 1. This company spe-
analyze and interpret it, which is similar to a data mining process cializes in planning, design, and construction supervision of tra-
共Feelders et al. 2000兲: ditional construction projects such as transportation, hydraulics,
1. Define data needs at the O, I, and P stages. Define produc- and area development.
tivity or other pertinent performance measures 共O兲 such as Productivity is defined as hours per drawing in this research.
profit. The measurable I1 variables usually include project This is a commonly adopted measure and easy to obtain, although
type, contract amount, the client, etc. Other variables can be some other measures may be needed if computer 3D models are
included, such as the design impacting variables 共Chalabi et replacing drawings. The hours/drawing numbers were collected at
al. 1986兲. Because the qualitative I2 variables are not usually the project level and by different disciplines. For example, engi-
readily available, a project nature questionnaire was designed neers of architecture, structure, and mechanical disciplines may
to collect them 共Chang and Chiu 2005兲. Similarly, the pro- charge hours if they work on a drawing. Hours and drawings of
cess variables P1 共and P2兲 are defined. individual disciplines were not analyzed separately, because the
2. Establish database and collect data. A sufficient amount of data retrieval would have been too cumbersome. The projects had
data must be collected and categorized according to the I, P, cost codes for engineers to charge on drawings, specifications,
and O stages. and other documents 共Chang and Tsai 2003兲.
3. Analyze I-O relationships. Extract data from the database If output performance is satisfactory, managers will not care
and start the first step of input-output analysis. The purpose much about input or process variables. But when productivity is
is to know whether productivity can be predicted upon re- not satisfactory, the managers need to explore I1 共and I2兲, as well
ceiving a project. as P1 共and P2兲 to relate the causing variables. The I-P-O relation-
4. Analyze P-O relationships. Determine which process vari- ship is linked through either I-O, or I-P and P-O. They are
ables affect productivity. The relationships identified in steps described individually in the following sections.
to chance.
However, this does not necessarily mean design work is more
productive than the other two phases. Rather, it is because design
I-O Relationships
projects’ main products are drawings, while planning and con-
I-O includes I1-O and I2-O. Fig. 1 shows that work data I1 and struction supervision projects focus on other deliverables such as
work nature I2 are the two categories of input variables. Produc- reports, which are less uniform and tangible. Further research can
tivity in terms of hours/drawing stands for output performance O. be pursued to find other productivity or performance measures for
The I1 variables analyzed hereafter are project type, phase, and planning and construction supervision projects.
duration. Project type and phase are qualitative information, while
duration is quantitative data. Project Duration 共I1兲-O
The project duration-hours/drawing relationship of the 190
Project Type 共I1兲-O projects is charted in Fig. 2. The solid line represents the statisti-
Project type is a variable that can be analyzed to see whether cal parabolic regression function. The coefficient of determination
certain project types are more productive than others. The project R2 = 0.0937 looks low. Prior to this parabolic regression, a linear
type and productivity relationship is shown in Table 1. Productiv- regression resulted in a smaller value R2 = 0.0521, in which the P
ity is ranked by average hours/drawing rates for individual project value= 0.0015 in the ANOVA.
types. It appears that transportation projects 共rail, highway, and Another way of grouping duration is by year. Table 5 displays
air兲 have better productivity, i.e., smaller hours/drawing values. In a more obvious trend. It shows that project productivity has a
contrast, area development and hydraulic projects are not as negative relationship with duration; that is, projects with shorter
productive. durations are more productive, especially those shorter than
To make sure these differences are not attributed to chance, an 2 years. This relationship is confirmed by ANOVA results as
analysis of variance 共ANOVA兲 using a = 0.05 was conducted. The shown in Table 6, in which F = 2.9 is larger than the critical value
results are shown in Table 2. Because F = 0.98 is less than the of 2.26.
critical value of 2.15 for 6 and 183 degrees of freedom 共DOF兲, the One explanation is that, for long lasting projects, people who
null hypothesis of the same productivity cannot be rejected. That are not released still charge their hours on the projects while the
is, the productivity differences among different types of projects number of drawings does not increase in proportion to the project
are not significant. duration. This should garner the company’s attention when
Transportation knowledge is the company’s core competence. managing longer duration contracts.
The general trend in Table 1 indicates that the productivity of The preceding three analysis examples are based on the single
transportation projects is ranked among the top, but this is not variable-productivity relationship. Since productivity is contrib-
concluded by ANOVA. This result deserves the company’s further uted by more than one variable, multiple regressions are more
investigation: are transportation projects really more productive? appropriate to see how the combined effect would explain more
Or is it because transportation projects produce more drawings of the variance. A linear multiple regression was conducted by
and the other types of projects produce fewer? When the reason is including the three variables, but 11 variables were generated
found, knowledge can be generated and transmitted to employees, in this equation: seven project types, three project phases, and
and the company can steer toward higher productivity for one duration. The R2 = 0.1967 increased and the P value was
transportation projects. 8.07E-6.
However, it will become too complicated if all variables are
Project Phase 共I1兲-O included in one regression, and the analysis would be formidable.
The project phase-hours/drawing relationship is shown in Table 3. There are eight input variables 共I1兲, and some of them are quali-
Design has the best productivity of 42 hours/drawing, almost half tative data that need to be split into more variables to be included
the rates for planning and construction supervision. The ANOVA in one multiple regression equation. Complication will further
results 共a = 0.05兲 are shown in Table 4. Because F = 12.0 is larger increase if the other 11 process variables 共P1兲 are included, with-
than the critical value of 3.04 for 2 and 187 DOF, the null out considering the implicit I2 and P2 variables yet. Furthermore,
different companies will collect their own variable data. It is Budget 共P1兲-O
very likely that the results from multiple regressions would be The relationship between budget and hours/drawing is shown in
irrelevant to other companies. Fig. 4. R2 = 0.1375 indicates a higher degree of correlation.
High R2 values are hard to get from single variable-
Downloaded from ascelibrary.org by The City College of New York - CUNY on 12/14/23. Copyright ASCE. For personal use only; all rights reserved.
The QA/QC implementation-productivity relationship is mance. The simple comparisons in Table 11 did not intend to
shown in Table 9. The 190 projects are divided into two groups: prove the relationship between PM and productivity. Instead, they
77 projects with QA/QC implemented and 113 projects without are explained to remind managers of these influential but implicit
Downloaded from ascelibrary.org by The City College of New York - CUNY on 12/14/23. Copyright ASCE. For personal use only; all rights reserved.
Conclusions
charge as much of their time as possible to billable projects
共Goldratt 1997兲. This inference is consistent with the prior long Engineering management needs to analyze cause-effect relation-
duration-low productivity relationship. ships more rigorously in order to get insight about project perfor-
The higher R2 value implies that I and P-O are linked. This mance. This research proposes a model to systematically analyze
linkage provides guidance in performance analysis and interpre- productivity causes in the project operation. Three categories:
tation. For example, a negative P1-O relationship, say, many work data 共I1兲 and work nature 共I2兲, work division 共P1兲 and man-
person-hours with low productivity, needs to be further analyzed agement 共P2兲, and performance 共O兲 are established at a project’s
and interpreted from the project duration variable in I1. input, process, and output stages, respectively. An implementation
procedure is also provided to guide performance analysis and
Work Nature 共I2兲-QA/QC 共P1兲 interpretation. Before using this model, a company has to define
To check the work nature-QA/QC 共I2-P1兲 relationship, the num- their own variables under these categories and collect sufficient
bers of QA/QC and non-QA/QC projects and their work nature in data.
terms of U&E scores are shown in Table 12. 共Only 101 of the The model was tested on a 190-project data set from an engi-
original 190 projects had available data for this U&E survey.兲 As neering design company. The variables of work data, work divi-
shown in Table 12, the 40 projects implementing QA/QC have sion, and management are found to influence design productivity.
lower U&E scores 共28 and 22, respectively兲 than the 61 non- For example:
QA/QC projects 共30 and 23, respectively兲. This means that the • Design has better productivity 共hours/drawing兲 than planning
QA/QC projects were regarded as less uncertain and ambiguous and construction supervision projects. The reason for this is
than the non-QA/QC projects. Further insights about the I-P-O probably that drawings are design’s main products, so its
relationships are described hereafter. drawing number is larger and its productivity value appears
lower.
• Project size has a negative relationship with productivity.
I-P-O Relationships
More productive projects have a shorter duration, smaller con-
When establishing the preceding I-O, P-O, and I-P relationships, tract amount, and smaller budget. The implications are: 共1兲
only one input or process variable with productivity is analyzed at people who are not released from long projects still charge
a time. To check the combined effect of various variables on their hours while the drawing number does not increase; 共2兲
productivity, the I-P-O relationships can be searched. This managers let engineers charge hours on larger projects; and 共3兲
additional analysis provides more insights about work operation mechanisms are needed in the work process, especially from
from input to output. One I-P-O linkage example is explained management, to prevent that negative relationship.
subsequently. • Project nature in terms of U&E would not affect productivity.
This implies that work nature is not a key performance factor;
Project Nature 共I2兲-QA/QC 共P1兲-O some mechanisms and variables in the process transform the
The I2-P1 relationship derived in the work nature-QA/QC section work nature and play a productivity change role.
helps explain the former P1-O relationship. Prior analysis from • The effect of QA/QC on productivity is not clear. Perhaps
Table 10 points out that projects with QA/QC requirements are QA/QC implementation clarifies work uncertainty and ambi-
not necessarily associated with better productivity. A possible in- guity, which saves time, but it also consumes time without
terpretation is that project U&E 共I2兲 is reduced through imple- tangible output, so its influence on productivity is mixed.
menting QA/QC 共P1兲, in which the quality procedures clarify The identified I-O, P-O, I-P, and I-P-O relationships provide
work U&E and this, in turn, saves time. But the effect is not large insights about work operation for the studied company. Other
Hours/drawing 33 32 7 25
Coordination No regular discussions No regular discussions Discussions with Discussions with
owner frequently owner frequently
Stakeholder needs Not much, no record Not much, no record No No
Teamwork Smaller project, Smaller project, Good chemistry Good chemistry
no teamwork no teamwork and common goals and common goals
among team members among team members
Knowledge and experience Fair Fair Good Good
Monthly progress report Only regular time sheets Only regular time sheets Yes Yes
Monthly cost report No No Yes Yes
References