1. Cost estimation & Documentation
Unit - IV
Cost estimation &
Documentation
- Prof. Prasad Mahale

2. GOOD ESTIMATES
Prediction is one that guide us for decision making through out the life cycle.
Prediction is useful only if it is reasonably (practically) accurate.
We do not expect our prediction to be exact, but we expect them to be close to actuals
Record the actual completion times of large number of projects, trying to implement same set of requirements.

4. GOOD ESTIMATES What is an estimate?
We think prediction as a range or window, rather than a single number
Estimate/prediction is not a target rather an estimate is a probabilistic assessment, so that the value produced by estimate is really the center of a range.
According to Demarco Estimation is a prediction that is equally likely to be above/below actual result
Estimate define as median of unknown distribution where it divides the arc under the curve into two equal parts A & B. In other words Actual value is just as likely to be less/greater than estimate

5. GOOD ESTIMATES
What is an estimate?

8. Evaluating estimate accuracy:-
GOOD ESTIMATES
Evaluating estimate accuracy:-
Once our project is ongoing, we compare actual value with estimate
E- estimate value, A – actual value then relative error in estimate is
Computed as
RE must be positive or negative depending upon values of A & E
Mean Relative Error for n projects is computed to be:
We need to define magnitude of relative error as a absolute value of RE & denoted as MRE.
Then for set of n projects, mean magnitude of relative error is:

10. 7. COST ESTIMATION : PROBLEM & APPROACHES
Prediction of likely amount of effort ,time & staffing level required to build a s/w system
Cost estimation/effort estimation use interchangeably
Cost estimation needed throughout life cycle-determine whether a project is feasible in terms of cost –benefit analysis
Preliminary estimation are difficult to obtain & inaccurate
Prediction support planning & resource planning

12. 7. COST ESTIMATION : PROBLEM & APPROACHES
Problem with cost estimation:-
Profession
Novelty of Application
Political Problems
Technical problem
4-approaches of cost effort & schedule estimation
Expert opinion
Analogy
Decomposition
models

13. 7. COST ESTIMATION : PROBLEM & APPROACHES
Problems with Cost Estimation
Profession: it takes advantages of repeatability of their task
Novelty of Application introduces more uncertainty in prediction
Political Problems: Manager converts estimation into targets or manipulates estimation parameter to fit into already defined outcomes
Technical problem: Since we are not encourage to collect data about process and product so very few historical records are available for prediction.
Organization spending more effort & time for estimation in early phases of project and giving less effort after project completion

14. COST ESTIMATION : PROBLEM & APPROACHES
Current Approaches to Cost Estimation
Expert opinion
Analogy
Decomposition
models

15. 1.Expert opinion:- 2. Analogy:-
- by mature developer’s personal experience
Manager describes parameter of project, expert make prediction based on experience
Expert may use tools, models or other method for estimation, but not visible to requester
Strength relies on quality of expert & experience
2. Analogy:-
- Estimate compare proposed project with one/more past projects
Identify similarity & difference
Describe a project in terms of key characteristics
Analysis is documented

16. 3. Decomposition:- 4. Models:- - More thorough analysis
Focus on product deliver or task require to build s/w
s/w describe in terms of smallest component
Activities are decomposed into low-level tasks & estimate are made
low-level estimate are combine to produce a complete project estimate
4. Models:-
- Identify key attributes to effort, generate mathematical formulae
- take i/p as experience of developer, implement Language, degree of reuse
- I/p value may require decomposition

17. Bottom up & top down estimation
Bottom up estimation begin with lowest level parts of product or task & provide estimate for each
Combine low-level estimate into higher level one, it is applies to decomposition technique only
Top –down estimate begin with overall process or product
Full estimate is made & then estimate for component parts are calculated

20. MODEL OF COST & EFFORT Cost model:- Constraint model:-
Provides direct estimate of effort or duration
Based on empirical data reflecting factors that contribute to overall project cost
It has one primary i/p & no. of secondary adjustment factors known as cost drivers
Cost drivers are characteristic of project, process, product or resource that affects the effort or duration
Constraint model:-
Demonstrate relationship over time between 2 or more parameter of effort, duration or staffing level
Rayleigh curve is used as constraint model in several commercial product

21. Regression based Model:-
It is used to build cost model by collecting data from past projects & examining relationship among attributes
Once basic equation is define, estimate can be adjusted by other secondary cost factors
In fig. it indicate, how much a regression equation can be derived

22. Each data point represent a project
Regression has been performed using logarithm of physical size on X-axis & log of project effort on Y-axis
Transforming linear equation:-
log E = log a + b log S
Log-log domain to real domain yield an exponential relationship of form
E = a Sb

23. If size were a perfect predictor of effort then every point of graph would lies on line of equation with residual error zero
Next step:- identify factor that cause variation between predicted & actual effort
Factor analysis help to identify additional parameter
Add this factor to model as cost driver & assign weighting factors to model their effort
Weight are applied to R.H.S. of effort equation
E=(aSb).F

24. ORIGINAL COCOMO MODEL:-EFFORT
Collection of 3 models
Basic model:- applied when little about project is known
Intermediate model:- applied after requirement are specified
Advanced model:- applied when design is completed
All 3 take same form
E =aSb.F
E= effort in person month, S= size in KDSI
F = adjustment factor
Value of a & b depend on development model determined by type of s/w under construction

25. Fig:- effort parameter for three modes of COCOMO
Value of a & b depend on development model determined by type of s/w under construction
Fig:- effort parameter for three modes of COCOMO
Organic system:- involve data processing
It tend to use d/b & focus on transaction & data retrieval
Ex:- banking or accounting system
Embedded system:-
-contain real time s/w that is integral part of large h/w based system
Ex:- missile guidance system
Semi-detached:-
Between organic & embedded
Development mode a b
organic
2.4
1.05
Semi-detached
3.0
1.12
embedded
3.6
1.20

30. Developers experiences
Organic
Semi-Detach
Embedded
Team Size
Small Size
Medium Size
Large Size
Size
2-50 KLOC
KLOC
300 & Above KLOC
Developers experiences
Experienced Developers needed
Average Experienced peoples
Very Little previous experience
Environment
Familiar Environment
Less Familiar
Significant Environment changes (Almost new env.)
Innovation
Little
Medium
Major
Deadline
Not Tight
Tight
E.g.
Payroll System
Utility System
Air Traffic Control System

31. B. ORIGINAL COCOMO: DURATION
Predict elapsed time from effort
Same form as effort
D=a Eb.F
D= duration in month
Co-efficient & exponent depend on development mode, it give optimal estimate of project duration for a given effort
Development mode a b
organic
2.5
0.38
Semi-detached
0.35
embedded
0.32

32. In COCOMO 2.0
Stage 1:-
Project builds prototype to resolve high risk issues involving user interface, s/w & system interaction, performance or technological maturity
Little is known about size of final product
Estimates the size in terms of object –point
Stage 2:-
Designer explore alternative architecture & concept of operation
- Again not enough info. To support effort & duration, but more as compare to stage 1

34. In COCOMO 2.0
Estimate the size in terms of function point, estimate functionality of product
Stage 3:-
Development begun & more information is known
This stage correspond to original COCOMO model
Estimates the size in term of LOC

35. 3 . Putnam’s SLIM model:-
-Putnam develop constraint model called SLIM
Applied to project exceeding 70,000 LOC
It assume that effort for s/w development. Project is distributed similarly to a collection of Rayleigh curves, one for each major development activity
as shown in fig:-

37. Putnam use empirical observations about productivity level to derive s/w equation from Rayleigh curve formula
This equation relates size to several variables: technology factor C, total project effort measured in person year K, & elapsed time to delivery td in year
Relationship expressed as –
td is a point at which Rayleigh curve reaches a maximum
C takes up to 20 different values
SLIM eq. include 4th power-strong implication for resource allocation on large project

38. Putnam's SLIM Model To estimate effort/duration, Putnam introduce:
D0 = K / td 3
D0 - manpower acceleration constant (12.3 for new software with many interface and interactions with other system), 15 for stand-alone system, 27 for re-implementation of existing systems)

39. To allow effort or duration estimation, Putnam derived the equation –
Ex: for manpower acceleration
15- for stand alone system
27- for reimplementation of existing system
These values obtain from studies of existing system
By combining equation 1 & 2, Putnam derived equation for Effort as
SLIM uses separate Rayleigh curve for-
Design & code
Test & validation
Maintenance
management

40. Multi-project models Effort estimates are affected by other projects
Cost can be amortized over several upcoming projects.
Reuse normally involves multiple projects

41. S/W Documentation:
Describe how document is checked for accuracy & also describe criteria for document distribution
Need to get clear idea about our product & to make it user friendly
Minimum components of documentation set include:-
s/w Requirement Specification(SRS)
s/w Design Description (SDD)
s/w Interface Documentation (SID)
s/w Test Documentation (STD)
s/w Development plan(SDP)
User Documentation
Document Distribution

42. 1. Software Requirement Specification(SRS)
Describe capabilities ,states & functionalities of all aspect of system
Include major component & subcomponents and internal interfaces of s/w & d/b
Include items required by user
This type of documentation need to be prepare for each deliverable & non-deliverable item of s/w ,firmware & h/w
2. Software Design Description (SDD)
Describe major component & subcomponents of s/w including d/b , internal interfaces
Include use of computerized design tool
Design & coding standards are used in development of deliverable & non-deliverable item of s/w

43. 3. Software Interface Documentation (SID)
Describe capabilities & functionalities of all interfaces between any two component of system.
Include major component & subcomponents and both internal & external interfaces of s/w.
4. Software Test Documentation
Include
Test design procedure & specification
Test log
Test Summary Report

44. 5. Software Development Plan
No good standard for it. Even ISO 9000 doesn’t require it
Whatever format choose to use ,come up with format ,get everyone agree on it & finalize to it
Consistency is very important in SDP
6. User Documentation
It describes the req. i/p ,options & other user activities necessary for successful use & application of system
2- objective
To train user to get help & accurate service from system
Provide effective on-going reference for usage of system

45. 7. Document Distribution
Process Interface to Configuration Management:
Interface between document distribution function & s/w configuration mgmt. function need to be define for purpose of document distribution
All procedure are needed to defined & implemented for all document before their distribution
Distribution to Internal Project Personnel
Procedure for distribution of system document to internal project personnel must be clearly defined
Distribution to External Project Personnel
Procedure for distribution of system document to external project personnel must be clearly defined

46. PROBLEM WITH EXISTING MODELING METHODS
Model accuracy evaluation take one of two form
Study applied model to past project data from no. of projects & compare actual effort with predicted effort
Study use a selection of models to estimate effort for particular project
-Reason of existing modeling method fail in their goals:-
Model structure:-
Key determinant of effort is size
But relation between effort & size is unclear
Value of a & b differ from data set to data set
most model include adjustment for diseconomy of scale

47. 2. Overlay complex models
Putnam model implies that decreasing duration increase effort, but increasing duration decrease effort
Most model perform poorly because they developed from post-hoc analysis of particular data set
2. Overlay complex models
-many model include adjustment factors such as COCOMO cost factor & SLIM’s technology factor to provide flexibility
Cost driver doesn’t always improve accuracy of estimate
3. Product size estimation
-size is not measurable early in life cycle
COCOMO & SLIM req. size in LOC but LOC can’t be measured from req. document
Estimate of LOC may be very inaccurate

48. DEALING WITH PROBLEMS OF CURRENT ESTIMATION METHODS
-step to improve accuracy of prediction
1. Local data definitions:-
Use size & effort measure that are defined consistently across the environment
Measure must be understandable by all who use & 2 people measuring should produce same no. of rating
2. Calibration:-
It improve accuracy of all models.
Ensuring that values supplied to model are consistent with model requirement & expectations
Calibration can’t regarded as one time activity
Model must be re-evaluated & re-calibrated periodically

49. DEALING WITH PROBLEMS OF CURRENT ESTIMATION METHODS
3. Independent estimation group:-
-Assign estimation responsibility to a specific group of people
Group provide estimate to all project, storing result of all data capture & analysis in historical d/b
Advantages:-
Group member are familiar with estimation & calibration technique
so consistent from one project to another
2. Estimation experience gives better sense of when project deviating from average or standard so adjust factor are realistic
3. By monitoring d/b, group can identify trends & perform analysis that are impossible for single project

50. Group estimate:- Group can re-estimate periodically
4. Reduce i/p subjectivity:- gives accurate estimation model based on homogeneous data set
5. Preliminary estimate & re-estimation:-
Estimate improve by 2 –steps
Basing preliminary estimate on measure of available product & re-estimating as more info. Available & more product produced
Group estimate:-
-derive preliminary estimate from group
-individual get experienced with wide variety of project
- It is done by wideband Delphi technique

51. Basic procedure:-
Co-ordinator present each expert with a specification of proposed project & info.
Co-ordinator call a group meeting, expert ask a question & discuss estimate issue
Each expert provide a initial estimate of total project & develop effort, estimate provide a range between a upper & lower bound
Co-ordinator prepare & circulate a summary report, indicating individual & group estimate
Hold meeting to discuss current estimate
Finally expert provides anonymous estimate, repeat step 2 to 6 until satisfactory

52. Variance = upper bound-lower bound
estimate= lower bound + 4(most likely)+upper bound 6
i.e. Group estimate is average of individual estimate
Variance = upper bound-lower bound
B. Estimation by analogy:-
Effort & duration are compare with same completed project
C. Re-estimation:-

53. Decompose cost model:-
6. Alternative size measure for cost estimation
Function bang:- no. of primitive in DFD
Data bang :- no. of entities in ER diagram
Also measure:-
No. of diagram no. of processes amount of data
No. of terminal processes – no. of processes in issued document
7. Locally develop cost model:-
5-steps in defining local cost model
Decompose cost model:-
Identify phrases & activities then,
Suggest no. of cost model require & form basic sources of data for them

54. b) Formulate cost theory:-
Describe relation between i/p & o/p
c) Collect data:-
- To check validity, collect data on project that is completed in particular environment represented by local model
- Theoretically 3 project to test cost theory of i/p
Practically 7 to 10 project
d) Analyze data & evaluate model:-
- Regression analysis used to analyze
e) Check model:-
- Mean magnitude of relative error ,PRED(0.25), & EQF used to determine accurate estimate

55. IMPLICATION FOR PROCESS PREDICTION
Want to predict how many requirement will change during development or no. of req. fault we can expect to find during inspection
Want to predict size or complexity of design from req. , amount of design that can be reuse from past project
No. of design fault
Needed to predict when coding will be complete ,how much code will be reusable on future project

56. IMPLICATION FOR PROCESS PREDICTION
When integration & system testing can begin
When testing will complete & when product turnover to customer
Each is collection of activities with a clears starting & end point & involve uncertainty
Making accurate predicting is challenging but necessary part of software engineering