Estimation - Software Metrics

1
Estimation - Software Metrics

• Managers frequently have to measure the
  productivity of software engineers.

2
How is productivity measured?

• What is Productivity?
• Why are we trying to measure Productivity?
• What do these measures really mean?

3
Estimation - Software Metrics

• Estimates are needed for
• project estimates
• time to completion
• cost
• manpower
• measure whether technological improvements are
  effective
• measure individual productivity for salary review

4
Estimation - Software Metrics

• Project estimation effects project scheduling!

5
Estimation is on-going --

• Is progress being made?
• What happens if progress falls behind?
• What if "expenditures" are greater than estimates?

6
Estimation - Software Metrics

• TOTAL COST
• (1) hardware software maintenance
• (2) travel and training
• (3) effort costs??

7
Effort costs???

• SE salaries
• Overhead
• support staff
• physical plant
• networking and communications
• benefits

8
Software Metrics

• Effort costs are approximately twice (2x) SE
  salaries
• So... Overhead is approximately equal to SE
  salaries!

9
Software Metrics

• Productivity measuring some attribute of the
  software
• and
• dividing this by the total effort required for
  development.

10
Software Metrics

• Two types of measurements are available
• Size-related and
• Function-related

11
Software Metrics

• SIZE-related Direct Measures
• What are they? What about an SE project can we
  measure directly?
•  cost
•  effort
•  LOC (lines of code)
•  speed
•  memory size
•  errors

12
Software Metrics

• Function-related Indirect Measures
• Examines overall/useful functionality over a
  period of time
•  functionality
•  quality
•  complexity
•  efficiency
•  reliability
•  maintainability

13
Software Metrics

• Software cost and effort estimation will never be
  an exact science

14
Software Metrics

• Too many variables
• human
• technical
• environmental
• political

15
Software Metrics

• No estimation technique will be entirely
  accurate.
• Estimates fed back into the project, alter the
  behavior being modeled

16
Size-oriented metrics

• KLOC (thous. lines of code) were developed with
  24-person-months of effort at a cost of 168,000
• values represent all software activities
  (analysis, design, code and test) not just coding

17
Size-oriented metrics

• pages of documentation developed
• 29 errors were encountered after software was
  released to the customer within the first year of
  operation
• 3 people worked on the development of the software

18
Size-oriented productivity and quality metrics

• Productivity KLOC/person-month
• Quality error/KLOC
• Cost /KLOC
• Documentation Pages doc./KLOC

19
Size-oriented productivity and quality metrics

• Controversy Why LOC as a key measure?

20
Size-oriented productivity and quality metrics

• Pro LOC are "artifact" of all software
  development project that can be easily counted.
• Con LOC measures are programming language
  dependent. They penalize well-designed but
  shorter programs. They cannot easily accommodate
  non-procedural languages .
• Con When used in estimation, LOC requires may be
  difficult to determine planner must estimate the
  LOC long before analysis and design have been
  completed.

21
Function-related metrics

• Examples
• number of user inputs
• number of user outputs
• number of user inquiries/interactions
• number of files
• number of external interfaces

22
Function-related metrics

• Relate to overall functionality of delivered
  software.
• Function points try to give a weight to these
  functionality measures.
• Each Function point (FP) is individually assessed
  for complexity and given a weight.

23
Function points

• EX.
• Weight of 3 for simple external inputs,
• Weight of 15 for complex internal files.

24
Function points

• FPs (function points) can create similar
  estimation formulas
• productivityFP/person-mth
• quality error/FP
• cost /FP
• documentationpages doc./FP

25
Alternative techniques

• Unadjusted Function Point Count UPC
• UPC of elements of a given type
• X weight
• This is further adjusted by a measure (a value)
  for overall project complexity.

26
But --

• What counts as complexity?

27
Function points

• Final Function Count (FFC)
• FFC UFC x project complexity factors
• These values can compare individuals on a single
  project or across projects.

28

• Function Points are Controversial

29
Function Points Controversy

• Pro FP is programming language independent
  making it ideal for applications using
  conventional and non-procedural languages
• based on data more likely to be known early in
  the evolution of the project
• attractive as an estimation approach

30
Function Points Controversy

• Con method requires some "slight of hand" in
  that computation is based on subjective rather
  than objective data
• difficult to collect after-the-fact data to
  verify
• no direct physical meaning -- its just a number

31
But, any measuring tool

• Can easily be misused- With enough adjustment,
  you can get the numbers to say almost anything.

32
Estimation Accuracy

• Regardless of model
• Care must be exercised when using the results of
  cost estimation
• Often the case that external factors really
  drive the schedule

33
Estimation Accuracy

• Note
• The OS/360 operating system was developed in 2
  yrs because the hardware was going to be ready in
  2 yrs