Effort Estimation - Faculty & Staff Web Pages

advertisement
Effort Estimation
Software Effort Estimation
Effort Estimation

Estimating



The process of forecasting or approximating the
time and cost of completing project deliverables.
The task of balancing the expectations of
stakeholders and the need for control while the
project is implemented
Types of Estimates


Top-down (macro) estimates: analogy, group
consensus, or mathematical relationships
Bottom-up (micro) estimates: estimates of
elements of the work breakdown structure
Which view is correct?







Rough order of magnitude is good enough. Spending
time on detailed estimating wastes money
Time is everything; our survival depends on getting
there first! Time and cost accuracy is not an issue.
The project is internal. We don’t need to worry about
cost.
The uncertainty is so great, spending time and money
on estimates is a waste.
The project is so small, we don’t need to bother with
estimates. Just do it.
They used an internal estimate “for strategic decisions”
and then we had to live with it.
We were burned once. I want a detailed estimate of
every task by the people responsible.
Macro versus Micro Estimating
Conditions for Preferring Top-Down or Bottom-up
Time and Cost Estimates
Condition
Strategic decision making
Cost and time important
High uncertainty
Internal, small project
Fixed-price contract
Customer wants details
Unstable scope
Macro Estimates
X
Micro Estimates
X
X
X
X
X
X
Estimating Projects: Preferred Approach

Make rough top-down estimates.

Develop the WBS/OBS.

Make bottom-up estimates.

Develop schedules and budgets.

Reconcile differences between top-down and
bottom-up estimates
Estimating Guidelines for Times,
Costs, and Resources
1.
Have people familiar with the tasks make the
estimate.
2.
Use several people to make estimates.
3.
Base estimates on normal conditions, efficient
methods, and a normal level of resources.
4.
Use consistent time units in estimating task times.
5.
Treat each task as independent, don’t aggregate.
6.
Don’t make allowances for contingencies.
7.
Adding a risk assessment helps avoid surprises to
stakeholders.
Refining Estimates

Reasons for Adjusting Estimates





Interaction costs are hidden in estimates.
Normal conditions do not apply.
Things go wrong on projects.
Changes in project scope and plans.
Adjusting Estimates

Time and cost estimates of specific activities are
adjusted as the risks, resources, and situation
particulars become more clearly defined.
Refining Estimates (cont’d)

Contingency Funds and Time Buffers





Are created independently to offset uncertainty.
Reduce the likelihood of cost and completion time
overruns for a project.
Can be added to the overall project or to specific
activities or work packages.
Can be determined from previous similar projects.
Changing Baseline Schedule and Budget

Unforeseen events may dictate a reformulation of
the budget and schedule.
Why Refine an Estimate?
Methods for Estimating Project
Times and Costs

Macro (Top-down)
Approaches

Consensus methods

Ratio methods

Apportion method

Function point methods for
software and system projects

Learning curves
Project Estimate
Times
Costs
Apportion Method of Allocating Project
Costs Using the Work Breakdown Structure
Methods for Estimating Project
Times and Costs (cont’d)

Micro (Bottom-up)
Approaches

Template method

Parametric Procedures
Applied to Specific Tasks

Detailed Estimates for the
WBS Work Packages

Phase Estimating: A Hybrid
Duration vs. Effort vs. Productive Time



Duration is the elapsed time in business
working days
Work effort is the labor required to complete
an activity. Work effort is typically the amount
of focused and uninterrupted labor time
required to complete an activity.
Productive time considers the percentage of
the work day that can be devoted to project
activity work. Estimates in IT range from 6675%, recent estimates of about 50-65%
(same client base). This doesn’t include
unexpected interruptions!
Elapsed time vs. work time
Software Cost Estimation
•What is the Problem?
•100 - 200% cost overruns are not uncommon
•15%of large projects never deliver anything
•31% of new IS projects cancelled before completion ($81 billion)
•What are the consequences?
•Economic
•Technical
•Managerial
•What is gained through effective software cost-estimation?
•schedule/staffing estimates
•better understanding of a particular project
Why are we bad at software
estimation?
•Complexity
•Infrequency
•Uniqueness
•Underestimation bias
•Goals not estimates
Basic Steps in Software Estimation







Identify project objectives and requirements
Plan the activities
Estimate product size and complexity
Estimate effort, cost and resources
Develop projected schedule
Compare and iterate estimates
Follow up
Software Cost-Estimation Methods







algorithmic
expert judgement
similar, completed projects
equate to available resources
Price-to-win
Top-down (global estimate)
Bottom-up (each component separately estimated)
Algorithmic Models
COCOMO
ESTIMACS
ESTIPLAN
FAST
FUNCTION
POINTS
MAINSTAY
PRICE
SLIM
SOFTCOST-R
SPQR
TRW (Boehm)
Computer Associates (Rubin)
AGS Management Systems
Freiman Parametric Systems (Freiman)
IBM (Albrecht)
Mainstay Software Corporation
RCA
QSM (Putnam)
Reifer Consultants (Tausworthe)
Software Productivity Research (Jones)
Basic Algorithmic Form
Effort = constant + coefficient*(size metric) +
coefficient*(cost driver 1) +
coefficient*(cost driver 2) +
coefficient*(cost driver 3) +
…..
size metric
lines of code
‘new’ versus ‘old’ lines of code
function points
SLOC as an Estimation Tool

Why used?
 early systems emphasis on coding

Criticisms
 cross-language inconsistencies
 within language counting variations
 change in program structure can affect count
 stimulates programmers to write lots of code
 system-oriented, not user-oriented
How many Lines of Code in this program?
#define LOWER 0 /* lower limit of table */
#define UPPER 300 /* upper limit */
#define STEP 20
/* step size */
main ()
/* print a Fahrenheit-Celsius conversion
table */
{
int fahr;
for (fahr=LOWER; fahr <= UPPER; fahr=fahr+STEP)
printf(“%4d %6.1f\n”, fahr, (5.0/9.0)*(fahr-32));
}
COCOMO Cost Drivers














Required software reliability
data base size
product complexity
computer execution time constraint
computer storage constraint
computer turnaround time
analyst capability
programmer capability
application experience
hardware/software experience
programming language experience
use of modern programming practices
use of software tools
required development schedule
Algorithmic Model Conclusions

Algorithmic Models can do a good job in estimating
required effort
•
Good project data must be collected and analyzed in
order to derive useful algorithms
•
Calibration is essential as the specific environment
is critically important

Effort estimates do have other uses
 Productivity evaluation of project teams or software
development technologies
 Objective negotiating tool with users in changes in
scope and impact on budget/schedule
Function Count Systems View:
Functionality Types
Interface
Files
Inputs
Internal
Files
Queries
Outputs
Function Points
History
 Non-code oriented size measure
 Developed by IBM (A. Albrecht) in 1979, 1983
 Now in use by more than 500 organizations
world-wide
What are they?
 5 weighted functionality types
 14 complexity factors
Functionality Types
EXTERNAL USER
input
type
output
type
inquiry
type
Internal
Logical File
External
Interface File
input type
output type
inquiry type
Application Boundary
Other Applications
Processing Complexity Adjustment
1)
2)
3)
4)
5)
6)
7)
8)
9)
10)
11)
12)
13)
14)
data communications
distributed functions
performance
heavily used configuration
transaction rate
on-line data entry
end user efficiency
on-line update
complex processing
reusability
installation ease
operational ease
multiple sites
facilitates change
Each rated on scales equivalent
to the following:
Not present
Incidental Influence
Moderate Influence
Average Influence
Significant Influence
Strong Influence
=0
=1
=2
=3
=4
=5
Function Point Calculation
Function Counts =
Function Points =
where
xi
wj
ck
5
3
FC   x i w j
i  1 j 1

 14 
FP  FC.65  .01  ck 
 k  1 

= function i
= weight j
= complexity factor k

Need to track employees and their work
- Add, change, delete, queries, and reports
- Two types of employees, salaried and hourly

Employees can have more than one job assignment

Standard job descriptions are retained by system
Employees can have more than one location and
locations can have more than one employee
- Another system stores the location data

Detailed Function Point Counting Rules
(1) Internal Logical Files (ILFs) Rules:
 Each major logical group of user data or control
information
 Data is generated, used and maintained by the
application
In Practice:
 Count at logical (external design) level
 In DB environment generally a relational table = a logical
file (before extensive normalization)
 Ignore multiple views
Detailed Function Point Counting Rules
(2) External Interface Files (EIFs) Rules:
 Files passed or shared between
applications
 Reference data only (not transactions)
In Practice:
 Look for “read only” usage
 Count special database extracts
Example - ILFs and EIFs
Employee - entity type
- Employee name
- SSN
- Number of dependents
- Type (salary or hourly)
- Location name (foreign key)
 Salaried employee - entity subtype
- Supervisory level
 Hourly employee - entity subtype
- Standard Hourly rate
- Collective Bargaining Unit Number

Example - ILFs and EIFs

Job - entity type
-

Job Assignment - entity type
-

Job name
Job number
Pay grade
Effective Date
Salary
Performance Rating
Job Number (foreign key)
Employee SSN (foreign key)
Job Description
-
Job Number (foreign key)
Line number (not known to users)
Description line
Example - ILFs and EIFs

Location - entity --maintained in another system
- Location Name
-

Address
Employee SSN (foreign key)
COUNTING STEPS:
- Count number of ILFs and EIFs
-
Assign them a complexity weighting
Counting ILFs and EIFs

Three ILFs:
- Employee
- Job
- Job Assignment
- not Job Description (logically part of Job)
- not Location (an EIF)
- not Salaried Employee (a Record Element Type)
- not Hourly Employee (a Record Element Type)

One EIF:
- Location
Counting ILFs/EIFs - Complexity
Record
Element
RecordTypes
Element
Types(RETs)
Data Element
Data Types (DETs)
Element
Types
1-19 (DETs)
20-50
51+
(RETs)
51+
<2 1-19 Low 20-50
Low
Average
<2
Low
Low
Average
Average High
High
2-5 2-5
LowLow Average
>5 Average
Average High
High
High
>5
High
Three ILFs:
•Employee - 8 DETs and 2 RETs
•Job - 4 DETs and 1 RET
•Job Assignment - 5 DETs and 1 RET
One EIF: Location - 3 DETs and 1 RET
ILF and EIF Unadjusted FPs
Low
External
Input
External
Output
Logical
Internal
File
External
Interface
File
External
Inquiry
Average
High
x3
x4
x6
x4
x5
x7
3 x7
x10
x15
1 x5
x7
x10
x3
x4
x6
Detailed Function Point Counting Rules
(3) External Inputs (EIs) Rules:
 Each unique user data/control type that enters
application
 Adds/Changes/Deletes data in Internal logical
file
 Each transaction type is an external input
In Practice:
 Not necessarily equal to screens
 Don’t confuse with inquiries (no change to
data)
Counting EIs - Raw Data

Employee Maintenance
Add, change, delete Employee
 Employee Inquiry; Employee Report
Job Maintenance
 Add, change, delete Job
 Job Inquiry; Job Report
Job Assignment Maintenance
 Assign Employee to Job
 Job Assignment Inquiry; Job Assignment Report
 Transfer Employee
 Evaluate Employee
 Delete Assignment


Location Reporting

Location Inquiry; Location Report
Counting EIs - Complexity
File Types
File Types
Referenced
Referenced
(FTRs)
(FTRs)
<2
2
>2
<2
2
>2
DataTypes (DETs)
Data Element
Element
Types
1-4
5-15
+15
(DETs)
Low
Average
1-4Low
5-15
15+
Low
Average
High
Low
Low
Average
Average Average
High
High
Low
High
Average
High
High
Example EIs (3 of 10):
• Create Employee- 10 DETs, 2FTRs (Employee
and Location) => Average
• Delete Employee- 3 DETs and 1 FTR=> Low
• Assign Employee to Job - 6 DETs and 3 FTRs
(Employee, Job and Job Assignment)=> High
External Input (EI) Unadjusted FPs
External
Input
External
Output
Logical
Internal
File
External
Interface
File
External
Inquiry
Low
Average
High
6 x3
2 x4
2 x6
x4
x5
x7
x7
x10
x15
x5
x7
x10
x3
x4
x6
Detailed Function Point Counting Rules
(4) External Outputs (EOs) Rules:
 Each unique user data/control type that exits
application
 Unique means different format or processing logic
 Can be sent directly to users as reports/messages, or
to other applications as a file
In Practice:
 Processing must be involved (don’t count output
response to an inquiry)
 Detail and summary outputs count separately
Counting EOs - Raw Data

Employee Maintenance
Add, change, delete Employee
 Employee Inquiry; Employee Report - 6-19 DETs
Job Maintenance
 Add, change, delete Job
 Job Inquiry; Job Report- 5 DETs
Job Assignment Maintenance
 Assign Employee to Job
 Job Assignment Inquiry; Job Assignment Report
 Transfer Employee
 Evaluate Employee
 Delete Assignment


Location Reporting

Location Inquiry; Location Report- 6-19DETs
Counting EOs - Complexity
File Types
Data
File Types
Referenced
Element
Referenced
Delta Element
Types (DETs)
File
Types
(FTRs)
Types Data
(FTRs)
Referenced
Element 20+
1-5 (DETs)
6-19
(FTRs)
Types Average
1-5 Low
6-19
20+
<2
Low
(DETs) High
<2 2-3
LowLow
Low
Average Average
1-4
5-15 High 15+
2-3 >3
Low
AverageAverage
High
High
<2
Low
Low
>3
Average
High
HighAverage
2
Low
Average
High
>2
Average
High
High
Example EOs :
• Employee Report- 6-19 DETs, 2FTRs (Employee
and Location) => Average
• Job Report-5 DETs and 1 FTR=> Low
• Job Assignment Report - 6-19 DETs, 3 FTRs
(Employee, Job and Job Assignment)=> Average
External Output Unadjusted FPs
Low
External
Input
External
Output
Logical
Internal
File
External
Interface
File
External
Inquiry
Average
High
x3
x4
x6
1 x4
3 x5
x7
x7
x10
x15
x5
x7
x10
x3
x4
x6
Detailed Function Point Counting Rules
(5) External Inquiry (EQ) Rules:
 Each unique input/output combination where an input
causes and generates an immediate output
 Unique means different format or processing logic
In Practice:
 No processing involved. If result is calculated or derived
field, then it is an input and an output
 Help systems typically counted as external inquiry
 Rate complexity as the higher of the input/output value
Counting EQs - “Medium Cooked” Data
Employee Maintenance
 Employee Inquiry- 2 FTRs and 9 DETs (output)
Job Maintenance
 Job Inquiry - 1 FTR and 4 DETs (output)
Job Assignment Maintenance
 Job Assignment Inquiry- 1 FTR and 5 DETs
(output)
 Location Reporting
 Location Inquiry - 2 FTRs and 5 DETs (output)

RESULT - Use EI and EO matrices => 3 low
complexity and 1 average (employee)
EQ Unadjusted FPs
Low
External
Input
External
Output
Logical
Internal
File
External
Interface
File
External
Inquiry
Average
High
x3
x4
x6
x4
x5
x7
x7
x10
x15
x5
x7
x10
3 x3
1 x4
x6
Total Unadjusted Function Points
External
Input
External
Output
Logical
Internal
File
External
Interface
File
External
Inquiry
Low
Average High
6 x3
2 x4
2 x6
1 x4
3 x5
x7
3 x7
x10
x15
1 x5
x7
x10
3 x3
1 x4
x6
Total = 96 Unadjusted FPs
Are Function Points a “Silver Bullet”?
“The function-point metric, like LOC, is relatively
controversial...Opponents claim that the method requires
some ‘sleight of hand’ in that computation is based on
subjective, rather than objective, data...”
R. Pressman Software Engineering p. 94
"Variants in FP counting methodologies can result in
variances of up to +/- 50%."
Capers Jones Selecting a FP Counting Method
“Within organizations the variation in function point
counts about the mean appears to be within 30%...”
G. Low and D.R. Jeffery IEEE TSE Jan. 1990
Software Estimating Rules of
Thumb

Rule 1: One function point = 100 logical source code
statements (procedural languages)



300 for assembly languages, < 20 for some OO languages
Rule 2: Raising the number of function points to the
1.15 power predicts the approximate page counts
for paper documents associated with software
projects
Rule 3: Creeping user requirements will grow at an
average rate of 1% per months over the
development schedule

For a 2 year project, functionality at delivery will be 24%
larger then when requirements were collected.
Software Estimating Rules of
Thumb (continued)

Rule 4: Raising the number of function points to 1.2
power predicts the approximate number of test
cases created.


Assume each test case will be executed about 4 times
Rule 5: Raising the number of function points to the
1.25 power predicts the approximate defect potential
for new software projects


Defect potential is sum of bugs (errors) in requirements,
design, coding, user-documentation + bad fixes or
secondary errors introduced fixes prior errors.
For enhancements: raise to 1.27 power
Software Estimating Rules of
Thumb (continued)

Rule 6: Each software review, inspection, or
test step will find and remove 30% of the
bugs that are present


Implies 6-12 consecutive defect-removal
operations to achieve high-quality software
Rule 7: Raising the number of function points
to the .4 power predicts the approximate
development schedule in calendar months.

Longer for military projects; for enhancements
applies to size of enhancement (not base product)
Software Estimating Rules of
Thumb (continued)

Rule 8: Dividing the number of function points
by 150 predicts the approximate number of
personnel for the application


Includes software developers, QA, testers,
technical writers, DBAs, project managers
Rule 9: Dividing the number of function points
by 500 predicts the approximate number of
maintenance personnel

Raising function point to .25 power predicts
approximate number of years the application will
stay in use
Software Estimating Rules of
Thumb (continued)

Rule 10: Multiply software development
schedules by number of personnel to predict
the approximate number of staff months of
effort.



1000 function points raised to .4 = 16 calendar
months
1000 function points / 150 = 6.6 full time staff
16 * 6.6 = 106 staff months to build project
Software Estimating Rules of
Thumb (continued)



Staff month: 22 working days with 6
productive work hours each day
132 work hours per month
Capers-Jones IEEE Computer March 1996:
notes limitations of these types of heuristics
Download