COCOMO - COnstructive COst MOdel
PA017 SW Engineering II → Aspects of SW Development
Management
Jaroslav Ráček Josef Spurný
Faculty of Informatics, Masaryk University
October 18, 2022
Core ideas
Cost for developing an application s directly related to SW size
Precision of SW size estimation depends on development phase
Estimation is more precise in later phases
Precision of estimation may differ four times (4:1) both ways
25 KLOC ← 100 KLOC → 400 KLOC
For some programmers, estimating KLOC is easier than
estimating man-days
For COCOMO, we assume knowledge of KLOC for estimate
software cost
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 2 / 33
Core ideas
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 3 / 33
COCOMO
COCOMO - Constructive Cost Model
B. Boehm (IBM) - 1981, 1984, 1995
#SLOC as a main SW size and complexity indicator
Empirical relaionships E = E(KSCLOC1) and T = T(KSLOC)
where
E – effort (work expenditure, man-days, man-months)
T – time for development (calendar months)
Estimation are usually done for larger projects, hence KLOC
1
Kilo (1000) SourceCode Lines of Code
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 4 / 33
COCOMO
Sources of empirical data:
larger number of previous complex projects
different kind of projects with different objectives
different languages, technologies, development environments
Data from multiple senior team members (managers, tech.
leaders)
This data will be used to set-up parameters for COCOMO.
The better input data, the more precise estimation (output, KLOC)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 5 / 33
Original COCOMO
How precise will the estimation be? → 3 levels of detail:
Basic model: unrefined estimation of E(KSLOC) and T(KSLOC)
based on KSLOC estimation
Intermediate model: including other factors on E(KSLOC) and
T(KSLOC) estimation (e.g., what tools are used, how experienced
is the team, how mature is PM, etc.)
Advanced model: similar to Intermediate model but involves
development phases
More precise estimation only makes sense when good inputs are
available
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 6 / 33
Original COCOMO
How challenging is the project? → 3 development models:
Organic model: simple, easy-to-deliver projects, usually smaller
scope
Semi-detached model: intermediate level of project difficulty,
in-between Organic and Embedded level
Embedded model: large-scope, challenging projects, require
high level of experience and creativity from team
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 7 / 33
Organic model
small projects (SW < 50 KSLOC)
very good understanding of requirements
little restrictions, high level of freedom when designing interface
team experienced with delivery of similar projects
low dependency on specific HW or technology
minimal need for new algorithms and architectures
little risk of deadline-shortening
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 8 / 33
Organic model - examples
scientific or experimental applications
simple business models and applications
simple application for storage management
simple application for manufacturing process management
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 9 / 33
Semi-detached model
middle-sized projects (SW < 300 KSLOC)
reasonably good understanding of requirements
significant restrictions for interface design
team has some experience with delivery of similar projects
medium level of dependency on specific HW or technology
medium level of need for new algorithms and architectures
indispensable risk of deadline-shortening (project buffer is
usually included)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 10 / 33
Semi-detached model - examples
project of medium size and complexity (e.g. workflows, user
account management included...)
semi-complex application for storage management
semi-complex manufacturing process management application
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 11 / 33
Embedded model
SW with various size (not only large projects)
only general outline of requirements is known
high amount of restrictions, strict requirements for interface
design
team has some experience with delivery of similar projects
high level of dependency on specific HW or technology
very high requirements for new algorithms and architectures
high risk of deadline-shortening (project buffer is usually
included)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 12 / 33
Embedded model - examples
ambitious and complex systems
complex signal and control processing systems
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 13 / 33
Effort and time
Calculating E(KSLOC) and T(KSLOC):
E = a.(KSLOC)b
T = c.Ed
where a, b, c, d are pre-defined model parameters:
basic-intermediate-advanced model
organic-semidetached-embedded model
→ 9 possible combinations
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 14 / 33
Parameter values
Empirical values (obtained via statistics from existing projects) for
parameters used to calculate E(KSLOC) and T(KSLOC):
there are values of a, b, c, d defined for each of 9 combinations of
models
examples:
basic model + organic model:
a = 3.0, b = 1.12, c = 2.5, d = 0.35
intermediate model + semi-detached model:
a = 2.8Fc, b = 1.2, c = 2.5, d = 0.32
where Fc is corrective factor (usually slightly above 1.0)
Over time, the formula remained the same but the calibration of
parameters was refined.
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 15 / 33
Parameter values
Empirical values for parameters used to calculate E(KSLOC) and
T(KSLOC) Intervals for parameter values:
a ∈ [2.4, 3.6] for basic model
a ∈ [2.8Fc, 3.2Fc] for intermediate and advanced model
b ∈ [1.05, 1.20]
c = 2.5
d ∈ [0.32, 0.38]
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 16 / 33
Parameter values
Empirical values for parameters used to calculate E(KSLOC) and
T(KSLOC)
In the basic model, all parameters are constant
In intermediate and advanced model (for all development
models), the value of a depends on Fc, all other parameters are
constant
The corrective factor Fc is a product of 15 attributes (cost drivers)
specific for development process
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 17 / 33
Corrective factor
4 Areas of attributes having impact on corrective factor Fc
attributes of SW product
HW attributes
attributes of development team
project attributes
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 18 / 33
Corrective factor
Attributes having impact on corrective factor Fc may have values on
the following scale:
very low
low
normal
high
very high
extremely high
These values have appropriate discrete numeric values.
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 19 / 33
Corrective factor - attributes of SW product
RELY - requested reliability (0.75 - 1.40)
DATA - database size (0.94 - 1.16)
CPLX - product complexity (0.70 - 1.65)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 20 / 33
Corrective factor - HW attributes
TIME - time restricted calculation (1.00 - 1.66)
STOR - memory/disc usage (1.00 - 1.56)
VIRT - reliability (vulnerability) of virtual machines (0.87 - 1.30)
TURN - turnaround time (0.87 - 1.15)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 21 / 33
Corrective factor - development team
ACAP - analyst capability (1.46 - 0.71)
PCAP - programming capability (1.42 - 0.70)
AEXP - experience with similar projects (1.29 - 0.82)
VEXP - experience with particular virtual machine (1.29 - 0.90)
LEXP - experience with particular programming language (1.14 -
0.95)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 22 / 33
Corrective factor - project attributes
MODP - usage of modern programming techniques (1.24 - 0.82)
TOOL - usage of SW tools (1.24 - 0.83)
SCED - precise planning (1.23 - 1.10)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 23 / 33
Cost drivers overview
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 24 / 33
Steps for COCOMO application
The following steps are performed when COCOMO is applied:
Calculation of nominal effort En
Definition corrective factor Fc
Definition of actual (refined) effort E
Calculation of development time T and other factors relevant for
project
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 25 / 33
Original COCOMO
Values a, b, c, d are identical for intermediate and advanced level of
model
for intermediate level, the calculation is applied for whole
project
for advanced level, the calculation is applied for individual
phases of development lifecycle
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 26 / 33
COCOMO for modification of existing application
ESLOC = ASLOC(0.4DM + 0.3CM + 0.3IM)/100, where:
ESLOC = equivalent SLOC
ASLOC = estimated number of modified SLOC
DM = percentage of modifications in design
CM = percentage of modifications in code
IM = integration effort (percentage of original work)
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 27 / 33
Evolution of COCOMO2
Motivation:
new SW processes
new methods for SW size measurement
new phenomenon of SW re-usability
need to make decisions on incomplete information
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 28 / 33
Extended and modified versions
COCOMO2 (1995) - 3 different models, vary by precision / amount of
parameters:
ACM - Application Composition Model
for projects with modern tools and GUI
EDM - Early Design Model
for initial estimations in early project phases
PAM - Post Architecture Model
for estimations once architecture is already specified
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 29 / 33
Extended and modified versions
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 30 / 33
New factors included in estimation
RUSE - requested re-usability
DOCU - documentation created during implementation
RCPX - complexity and reliability of product
VMHV - virtual machine variability - host
VMHP - virtual machine variability - periphery
PVOL - HW platform variability
PDIF - HW platform complexity
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 31 / 33
New factors included in estimation
PERS - personal capabilities
PREX - personal experience
PCON - personal continuity in project
PEXP - experience with given platform
LTEX - experience with language and tools
SECU - security
SITE - development in multiple sites
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 32 / 33
Extended and modified versions
COCOMO2 - estimating work expenditure and SW size when
modifying existing application:
ESLOC = ASLOC(AA + SU + 0.4DM + 0.3CM + 0.3IM)/100, where:
ESLOC = equivalent SLOC
ASLOC = estimated number of modified SLOC
DM = percentage of modifications in design
CM = percentage of modifications in code
IM = integration effort (percentage of original work)
AA (assessment and assimilation) = work investment needed to
determine the extent to which the existing module can be used
without modification
SU (software understanding) = readability and understanding of code
Jaroslav Ráček, Josef Spurný ·COCOMO - COnstructive COst MOdel ·October 18, 2022 33 / 33