You are on page 1of 26

Quantifying the Cumulative

Impact of Change Orders

Rich Camlic
Chair
Cumulative Change Order Impacts
Research Team

2000 CII Annual Conference


Nashville, Tennessee
Quantifying the Cumulative
Impact of Change Orders

Cumulative Change Order Impacts Research Team

RT 158
Cumulative Change Order Impacts
Research Team
Tripp Ahern J. F. Ahern Company
George Armenio General Motors Corporation
Rich Camlic U.S. Steel, Chair
Edward Gibble McClure Company
Brian Griffiths Electrical Corp. of America
Hanford Gross Gross Mechanical Contractors
Awad Hanna University of Wisconsin-Madison
Kevin Hughes FPL Energy
Kam Kamath Black & Veatch
Chris Lloyd-Jones Bechtel
Joe Loftus Sr. Terminal-Andrae Inc.
Wayne Montgomery Kvaerner Process
Greg Thomas Fisk Electric Company
Problem Statement
• Administration boards and courts
recognize that effects of cumulative
impact can go beyond the initial change
itself.

• It is difficult for owners and


contractors to agree that cumulative
impact exists, let alone come to an
equitable adjustment for it.
Research Objectives
1. Investigate how change orders impact
productivity over entire project.
2. Isolate specific, measurable
characteristics of impacted projects.
3. Develop a model capable of identifying
projects impacted by cumulative change.
4. Develop a model to predict the magnitude
of cumulative impact with a reasonable
level confidence.
Results of Research

• Two models (tools) developed


– Determine the probability of impact within a
range of possible outcomes.
– Predict the probable magnitude of impact
within a range of possible outcomes.

• Strong correlation found between the


number of change items and some loss of
labor productivity.
Recommendations to Owners
• The most common reasons for change
orders are Additions, Design Changes
and Design Errors, therefore you
should do more up-front engineering.

• Reduce change order processing time


to decrease the likelihood of impact.

• Require contractors to submit a


manpower loading curve with proposal.
Recommendations to Contractors
• Integrate any changes into the work flow
as efficiently as possible.

• Use project software to track


productivity:
– % complete by earned value

– % complete by actual earned work-hours

– % complete by actual installed quantities


Recommendations to Contractors

• Resource loading relationships


(ratios):
– Actual peak over actual average manpower

– Estimated peak over actual peak manpower

– Actual manpower loading curve versus


estimated manpower loading curve
Methodology
• Developed a comprehensive questionnaire
based on “influencing factors” that we felt
could cause change on a project.

• Used a pilot study to gather data, to


determine how easily the questionnaire
could be answered, and if it would be
useful in achieving our objectives.

• The study was based on work-hours.


Contractor Data
• 57 projects were solicited from
33 mechanical contractors.

• 59 projects were solicited from


35 electrical contractors.

• 116 projects in database.

• Industrial and institutional projects


make up majority of database.
Evolution of the Impact Model
Need to develop a definition for
“DELTA” (productivity loss/gain)
associated with change orders.

Total Actual (Estimated Hours +


Labor Hours Change Order Hours)
X 100
Total Actual Labor Hours
Hypothesis Development

• 75 variables were investigated using


hypothesis testing and analysis of variance
techniques to determine if they had an
impact on projects. All 116 projects were
tested.

• Logistic regression techniques then


identified the eight most significant
variables that impact a project.
Significant Impact Variables
• Mechanical or electrical project
• Percent change
• Estimated/actual peak labor
• Change order processing time
• Overmanning
• Overtime
• Peak/average work-hours
• Percent change orders related to design
issues
The Impact Model
(Simplified logistic regression)

ex
Probability Y =
1 + ex

where X is the sum of the eight significant


“influencing factors” (variables) times their
coefficients plus a constant
Confidence of Impacted Project
.5 does not indicate 50% chance of impact

0.0 0.25 0.5 0.75 1.0

No Some Good Strong


Evidence Evidence Evidence Evidence
Significant Variables for
Magnitude of Impact
• Percent change order work-hours
• Project Manager percent time on project
• Percent owner-initiated change items
• Productivity (tracked or not tracked)
• Overmanning
• Change order processing time
The Quantification Model
% Delta = 0.36866 + 0.11957 percent change
- 0.08065 PM % time on project
- 0.16723 % owner-initiated CO
- 0.09147 productivity
- 0.05213 overmanning
+ 0.022345 CO processing time
This equation predicts the most likely % Delta
(loss/gain of productivity) within a range of
possible outcomes.
Additional Validation of Model
Seven new projects were solicited after close of
research for additional validation of linear regression
model:
• All 7 within ± 15 percent of actual % Delta

• 5 of 7 within ± 10 percent of actual % Delta

• 4 of 7 within ± 5 percent of actual % Delta

This is an indication that our model is a good predictor


of the magnitude of impact.
What Does All This Mean?

• Is this an exact science?

• Can you use these models


with confidence?

• What evidence do I have to


back this up?
No Productivity Tracking & Poor CO
Process Time at 95% Confidence Level
PM %Time on Proj & %OwnerInitCO = Ave
Productivity=0, Overman=0, Processing=5
70%

60%

50%
% Delta

40%

30%
Data
20% Lower CI
10% Upper CI

0%
0% 50% 100% 150%

% Change
Productivity Tracking & Poor CO
Process Time at 95% Confidence Level
PM %Time on Proj & %OwnerInitCO = Ave
Productivity=1, Overman=0, Processing=5
60%

50%

40%
% Delta

30%

20% Data
Lower CI (95%)
10%
Upper CI (95%)
0%
0% 50% 100% 150%

% Change
Productivity Tracking & Good CO
Process Time at 95% Confidence Level
PM %Time on Proj & %OwnerInitCO = Ave
Productivity=1, Overman=0, Processing=1
45%
40%
35%
30%
% Delta

25%
20%
15%
Data
10%
Lower CI (95%)
5%
Upper CI (95%)
0%
0% 50% 100% 150%

% Change
Final Comments
• We do not claim, nor should you expect,
the “absolute” correct answer, but rather
a most likely answer that fits within a
range of possible outcomes both above and
below our predicted value.

• Each project is unique and requires that


project-specific data be used when
applying these models.
Final Comments
• We suggest the owner and contractor agree,
before a contract is signed, to use these
models as a conflict resolution tool, should
the need arise, at the end of a project.

• Owner and contractor should track actual


work-hours against estimated work-hours to
detect negative trends early so steps can be
taken to correct them before they become
a major problem.
Implementation Session

• Find out how this project works out.

• See a demonstration of the model.

• Research team members will answer


questions.

You might also like