Estimation â€“ Tools

<< Decomposition Techniques, Estimation â€“ Tools

Work Breakdown Structure >>

Software Project Management (CS615)

LECTURE # 31

6. ESTIMATION

6.5

Estimation Tools

iii.

Measuring effort for a project

c) Constructive Cost Model (COCOMO)

The COCOMO technique is another popular estimation technique. Dr: Barry

Boehm propounded this technique in 1981. COCOMO uses cost driver attributes

to calculate the effort and duration of a project. The COCOMO technique has

three levels of implementation. With each level, the complexity of the model

increases. The levels of the COCOMO technique are:

Basic

ii.

Intermediate

iii.

Advanced

Basic COCOMO

The basic COCOMO technique estimates the effort and cost of a software project

by using only the lines of code. You, use basic COCOMO when you need a rough

estimate of effort, such as during maintenance projects. This is because in such

projects, a majority of the work is already completed. Estimating the effort in the

basic COCOMO technique involves three steps.

1. Estimating the total delivered lines of code

2. Determining the effort constants based on the type of the project

3. Substituting values for lines of code and effort constant in a formula

You have already seen how the total delivered lines of code are estimated. The

next step in the COCOMO model is to determine the type of the project being

developed. The basic COCOMO technique considers three types of projects to

calculate effort.

Organic

Embedded

Semidetached

Organic projects have sufficient and defined objectives. The organizations that

undertake organic projects have ample experience in development and use small

development teams. These are simple business and financial applications; such as

a banking system and inventory system.

233

Software Project Management (CS615)

Embedded projects have stringent and specialized hardware, software, and human

resources requirements. Organizations usually have less experience in developing

such projects. Examples of such projects include real-time operating systems

(RTOS), industrial automation systems, and sophisticated space and aviation

systems.

Semidetached projects are a combination of the preceding two types of software

projects. A new operating system and a database management system (DBMS)

are examples of such projects.

The last step in calculating effort by using the COCOMO technique is to

substitute the values of lines of code and effort constants in the following

formula:

Ej = a1 * (KLOC)a2

In the formula, Ei is the effort for a project. The effort constants, al and a2

depend on the type of project being developed.

ii.

Intermediate COCOMO

Calculation of effort by using the intermediate COCOMO technique involves an

additional step of calculating the effort adjustment factor (EAF). The effort

adjustment factor is calculated by assigning ratings to 15 cost driver attributes.

These cost driver attributes relate to the various aspects of a software project, such

as project, product, personnel, and computer attributes. Using the intermediate

COCOMO technique, you can accurately estimate effort and cost required for a

project. Accurate estimates are very helpful to start new development projects.

Calculating the effort by using the intermediate COCOMO technique is a three-

step process:

1. Estimate the initial development effort by using SLOC. To do this, you use

the following formula:

Ei = a1 * (KLOC)a2

In the formula the initial development effort, KLOC refers to 1,000 lines of code.

The constant values a1 and a2 differ with every project.

2. The second step is to determine the relevant cost driver attributes that affect

your project intensively. This provides you with the value for EAF.

234

Software Project Management (CS615)

Table 1 summarizes 15 commonly used projects, personnel, and product-related

cost driver attributes. The values for each cost driver under each rating are filled

in by an organization based on past experience.

Table 1: Cost Driver Attributes

Rating

Cost Drivers

Very

Extremely

Negligible Low Average High

high

critical

Analyst Capability (ACAP)

Programmer Capability (PCAP)

Programming Language

Experience (LEXP)

Virtual Machine

Experience(VEXP)

Required Software Reliability

(RELY)

Database Size (DATA)

Software Product Complexity

(CPLX)

Execution Time Constraint

(TIME)

Main Storage Constraint

(STOR)

Computer Turnaround Time

(TURN)

Virtual Machine Volatility

(VIRT)

Use of Software Tools (TOOL)

Modern Programming Practices

(MODP)

Required Development

Schedule (SCED)

3. Finally, you calculate the actual effort by multiplying the weighted cost driver

attributes with the initial effort estimate. Typically, the val1,les that rate each

cost driver attribute range from 0.9 through 1.4. For example, if software

reliability (RELY) is of prime importance according to the requirements

specifications, it is provided a rating of high or a value of 1.4. Similarly, if the

time to execute a software program is of negligible importance, you assign a

rating of low or a value of 0.9. For software attributes that are of mediocre

importance, you can assign a value between 0.9 and 1.4.

235

Software Project Management (CS615)

Usually, in organizations, the average rating is assigned a static value of 1.0. To

calculate the estimated effort using the intermediate COCOMO technique, you

use the formula:

E = EAF * Ei

Consider an example for using the intermediate COCOMO technique to calculate

the estimated total effort of a project life cycle. In a customized insurance project,

there are four modules. The total effort estimate of the modules is 3.0 KLOC. The

management has identified four cost driver attributes with the respective

multiplying factors that might affect the project most. In this situation, the values

of al and a2 are 3.2 and 1.05, respectively, because the insurance project is an

organic project. Therefore, you apply the following formula to calculate the initial

effort estimate.

Ei = a1 (KLOC)a2

Ei =3.2 * 31.05

Ei = 3.2 * 3.16

Ei = 10.11

The values assigned to the cost driver attributes that are applicable to a particular

software application are displayed in Table 2. According to the table, the time to

execute a software program is of high importance. Therefore, the attribute TIME

is assigned a value of 1.35. In contrast, the software application does not require a

very high analyst involvement. Therefore, the value assigned to ACAP is very

low or 0.95. Using the same logic, the values for other cost driver attributes are

assigned.

Table 2: Applicable Cost Driver Attributes

Applicable cost driver attributes

Rating

Multiplying factors

CPLX

High

1.2

TIME

Very high

1.35

ACAP

Low

.95

MODP

Average

1.00

Referring to Table 7.9, EAP can be calculated as,

EAF = 1.2 * 1.35 * 0.95 * 1.0 EAF = 1.53

After obtaining the values for the variables Ei and EAF, you can substitute these

values in the formula to calculate the total effort.

E = EAF * Ei

E = 1.53 * 10.11

236

Software Project Management (CS615)

E = 15.5 person months

iii.

Advanced COCOMO

The advanced COCOl\l10 technique uses the steps of the intermediate COCOMO

technique. In addition, it uses costs driver attributes assigned to each phase of the

SDLC such as analysis and design.

Applicability of COCOMO

COCOMO is flexible and capable of using SLOC, FP, and even object points.

Object points are measurable code sections in an object-oriented programming

language, such as C++, Ada, and Java.

You can use COCOMO when the size of a project is extensive and the

requirements of the project are vague. In contrast, SLOC and FP can be used for

projects where either the requirements are more or less known or developers

possess the relevant experience in developing projects.

COCOMO is suitable for complex and sophisticated projects that are expected to

operate within intensive hardware, software, and personnel constraints.

Generally, you can use COCOMO when the software development environment

is new to an organization. In addition, you can use COCOMO when you do not

have baseline data about past projects. However, you need complete data about

your current project to assign weight age to each cost driver attribute. You can use

FP or SLOC techniques when you have enough past project data to assign

accurate weight age to the 14 GSC s and the various information domain value

elements.

d) Delphi Technique

The Delphi technique is a Human-based estimation technique. Human-based

estimation techniques use human experience and analytical skills to estimate the

size, productivity, and effort required for a project. This is a trusted technique and

is widely used in many established organizations to facilitate practical and

reasonable estimation.

The rationale of using the Delphi technique is that when many experts

independently arrive at the same estimate on the basis of similar assumptions, the

estimate is likely to be correct.

The Delphi technique has eight basic steps:

1. Identify the terms that need to perform the estimation activity. In an estimation

activity meeting, three distinct groups of people need to be present.

237

Software Project Management (CS615)

Estimation experts: They usually consist of groups of five or six experienced

project managers. The estimation values provided by the project managers are

based on past project history and their knowledge. However, only those

project managers should b~ invited for estimation whose experience of a past

project matches that of the current project. Otherwise, estimation values may

turn out to be far from realistic.

Estimation coordinator: An estimation coordinator is very similar to a

moderator in a usual meeting. The coordinator facilitates the meeting and

ensures that the goals of the meeting are fully achieved.

Author: An author is similar to a recorder of minutes in a meeting.

2. The author presents the project details including clients' needs and system

requirements to the group of experts. The author also describes the expectations

from the group. The author and experts jointly identify the tasks that need to be

estimated. They also identify the valid assumptions that they need to consider

while estimating. For example, while estimating the effort needed to create a

high-level design, they can assume that the SRS document is approved by the

client.

3. The author and experts arrive at a consensus that any estimation with a specific

variance value will not be accepted. For example, they may decide that any

variance above 25 percent will not be accepted as an estimation value for

computing the project effort or the productivity.

4. The coordinator prepares a list of tasks jointly decided by the team and

distributes the list to all experts. These tasks comprise a project plan.

5. The experts independently make their estimates for each task. After recording

their estimates, they hand over their estimates to the coordinator. This is a critical

step. While making estimates, no discussions or consultations are permitted

because a mutual discussion may influence the estimation logic of the fellow

experts. The coordinator and the author jointly ensure this.

6. The coordinator prepares a summary of estimates for each task in a table as

represented in Table 7.10. After calculating the percentage of variance, the

coordinator marks each task as accepted or not accepted based on the agreed

accepted value.

7. The coordinator hands over the summary to the group of experts and the author.

The group of experts and the author discuss tasks and assumptions where the

percentage of variance is more than the acceptable level. The maximum and

minimum estimates of tasks are not disclosed or discussed. For example, in Table

7.10, the group and the coordinator do not accept the high-level design task

because it exceeds the agreed variance value of 25%. Therefore, the team would

discuss this task to estimate its maximum and minimum effort afresh. To resolve

238

Software Project Management (CS615)

the high percentage of the variance value, some tasks may be broken down

further or combined. This activity of breaking down tasks into smaller levels

involves fresh estimates for those tasks at the smaller levels.

8. Revert to step 5 and repeat the steps. You do this until all tasks are assigned

estimates that have an acceptable percentage of variance value. Figure 7.4.

summarizes the steps of the Delphi technique in the form of a flowchart.

Identify the teams that will

estimate.

Present project details to the expert

group.

Finalize the acceptable variance

value.

Prepare a list of tasks.

Estimate done by the expert group

Prepare a summary of estimates

for each task.

Discuss tasks and assumptions for

not acceptable estimates.

239

Software Project Management (CS615)

Repeat steps until all estimates are

finalized.

Figure 1: Steps of the Delphi Technique

The Delphi technique is a simple and subjective method of estimation. However,

it is a very effective method because most of the estimates are tried and tested.

You can use this method if the project is small or if you have the data and

expertise that can enable unambiguous estimates.

240

Table of Contents: