Software Cost Estimation

1
Software Cost Estimation
2
Cost Estimation

• Estimating the cost of software development,
  usually in terms of programmer-months.
• The development time, staffing levels, and
  productivity may also be estimated

3
Purpose and Use of Estimations

• feasibility
• pricing
• planning
• staffing
• managing

4
Estimate
5
Estimate versus actual
6
Problems with software development estimations

• Programmer variability
• Product complexity
• Variability of goal

7
1) variability of programmer's ability

• 161 in programming time

8
2) product complexity

• application programs
• (separate, organic)
• data-processing, scientific
• utility programs
• (semi-detached)
• compilers, linkers, analyzers
• system programs
• (embedded)
• o.s., dbms, real-time

9
3) variability of goals

• goal size man-hours
• min size 33 30
• min memory 52 74
• code clarity 90 40
• execute fast 100 50
• min effort 126 28
• clear output 166 30

10
Traditional approaches

• Putnam, Boehm, et al.
• based on size (lines of code)
• formulas for estimating cost

11
Size Estimation

• a) estimate modules
• min minimum size
• best best guess of size
• max maximum size
• b) estimate expected size
• expected size
• E(i) (min 4best max) / 6
• total size S E1 E2 ... En

12
Size Estimation continued

• c) estimate standard deviation
• s.d. of E(i) (max - min) / 6
• s.d. of S (sd2 sd2 ... sd2)1/2

13
An Example of Size Estimation

• Mod Min Best Max E sd
• 1 20 30 50 31.6 5
• 2 30 80 200 91.6 28.3
• 3 100 150 200 150 16.6
• 4 80 110 180 116.6 16.6
• 5 40 70 90 68.3
  8.3
• 6 30 90 110 83.3 13.3
• 7 60 80 100 80
  6.6
• S 31.691.6 150116.668.383.380 621.4
• sd(25800.9275.6275.668.9176.943.6).5
• (1666.5).5 40.8

14
Programmer Productivity

• Application Programs
• PM 2.4 (KDSI)1.05
• Utility Programs
• PM 3.0 (KDSI)1.12
• Systems Programs
• PM 3.6 (KDSI)1.20

15
Example for productivity

• Size Appl Util Sys
• 5K 13.0 18.2 24.8
• 10K 26.9 39.5 57.1
• 15K 41.2 62.2 92.8
• 20K 55.8 86.0 131.1
• 25K 70.5 110.4 171.3
• 30K 85.3 135.3 213.2
• 35K 100.3 160.8 256.6
• 40K 115.4 186.8 301.1
• 45K 130.6 213.2 346.9
• 50K 145.9 239.9 393.6

16
Development Time (Months)

• Application Programs
• TDEV 2.5 (PM) 0.38
• Utility Programs
• TDEV 2.5 (PM) 0.35
• Systems Programs
• TDEV 2.5 (PM) 0.32

17
Example for development time

• size appl util sys
• 5K 6.63 6.90 6.99
• 10K 8.74 9.06 9.12
• 15K 10.27 10.62 10.66
• 20K 11.52 11.88 11.90
• 25K 12.60 12.97 12.96
• 30K 13.55 13.93 13.91
• 35K 14.40 14.80 14.75
• 40K 15.19 15.59 15.53
• 45K 15.92 16.33 16.25
• 50K 16.61 17.02 16.92

18
Average Staffing Levels

• Calculate by dividing PM by TDEV

19
Example for staffing levels

• size appl util sys
• 5K 1.96 2.63 3.55
• 10K 3.08 4.37 6.26
• 15K 4.01 5.87 8.71
• 20K 4.84 7.23 11.02
• 25K 5.60 8.51 13.21
• 30K 6.30 9.72 15.33
• 35K 6.97 10.87 17.39
• 40K 7.60 11.98 19.39
• 45K 8.20 13.05 21.35
• 50K 8.79 14.09 23.27

20
COCOMO Effort Multipliers

• product attributes
• required reliability 0.75 - 1.40
• data-base size 0.94 -
  1.16
• product complexity 0.70 - 1.65
• computer attributes
• execution time constraint 1.00 - 1.66
• main storage constraint 1.00 - 1.56
• virtual machine volatility 0.87 - 1.30
• computer turnaround time 0.87 - 1.15

21
Mythical Man-Month

• Phrase defined by Frederick Brooks
• People and time are not always interchangeable

22
Programmers vs Time
23
Putnam's Cost Estimation Model

• Macro-estimation model
• Relationship between cost and the amount of time
  available for the development effort.
• The model supports the 'mythical man-month' idea

24
Putnam's Model

• Based on Lines of Code
• Rayleigh Curve
• Shows Instantaneous Programmer-power vs Time
• Empirically verified on other projects

25
Rayleigh Curve

• The rayleigh curve is a product of at times e
  raised to the -bt squared
• The at dominates at first giving a steep rise and
  then the negative power of e lowers the value
  slowly

26
Putnams Model

• y instantaneous programmer-power
• K total lifecycle cost ( in programmer-years)
• t time from beginning of project
• td delivery time
• e 2.71828...

27
Software Equation

• There must be a relationship between the
  lifecycle development effort, the development
  time and the size of the project.
• This relationship is given in the Software
  Equation

28
Software Equation

• Ss CK1/3 td4/3
• where
• Ss is the estimated size of the software system
• K is the total lifecycle effort in
  programmer-years
• C is the technology constant
• td is the development time (in years)

29
Example 1 (part 1)

• Ss 100 K and ymax 40 people
• Using the programmerpower as a constraint, the
  shortest development time can be estimated.
• The maximum programmerpower occurs at the
  delivery time (i.e. t td).
• This implies that ymax (K/td)e-1/2
• This gives K/td 65.95.

30
Example 1 (part 2)

• Assuming that the technology constant is
  approximately 10000 , we can substitute into the
  Software Equation.
• By solving for K and substituting, we get td
  1.722 years.
• By substituting the value for td, we get K
  113.57 programmer-years.

31
Example 1 (part 3)

• The development cost is 40 of K or 45.26
  programmer-years.
• The daily productivity of the programmers can be
  calculated. Assuming that there are 250 workdays
  per year, the programmer productivity per day is
  equal to S/(.4K 250) lines, or 8.837 lines/day.

32
Distribution of effort

• effort
  schedule
• size (kdsi) 32 128 32
  128
• plan/req 6 6 12
  13
• pre design 16 16 19
  19
• detail dsgn 24 23
• code/unit 38 36 55
  51
• sys test 22 25 26
  30

33
OTHER ESTIMATES

• Walston-Felix
• E 5.2 L0.91
• E man-months
• L thousand lines of code
• Bailey-Basili (1981)
• E 3.4 .72 DL1.17 /- 25
• DL thousand lines of developed code plus
  comments

34
Function Points

• Attempting to measure functionality
• Widely used in business applications

35
Program Features

• outputs
• inquiries
• inputs
• files
• interfaces

36
Weights for features

• simple average complex
• outputs 4 5 7
• inquiries 4 5 7
• inputs 3 4 6
• files 7 10
  15
• interfaces 5 7 10

37
Questions

• Does the system require reliable backup and
  recovery?
• Are data communications required?
• Are there distributed processing functions?
• Is performance critical?
• Will the system run in an existing, heavily
  utilized operational environment?

38
Questions

• Does the system require on-line data entry?
• Does the on-line data entry require the input
  transaction to be built over multiple screens or
  operations?
• Are the master files updated on-line?
• Are the inputs, outputs, files, or inquiries
  complex?
• Is the internal processing complex?

39
Questions

• Is the code designed to be reusable?
• Are conversion and installation included in the
  design?
• Is the system designed for multiple installations
  in different organizations?
• Is the application designed to facilitate change
  and ease of use by the user?

40
Adjusted Function Points

• The following equation is used to produce the
  final function point value
• FP FP_total K1 K2 SUM(Fi)
• Where
• FP_total is the sum of all function point entries
  from the above table.
• K1, K2, and weighting factors are determined
  empirically.
• Fi are the sum of complexity adjustment values
  based on responses to the following questions.

41
Adjustment Weights