Function Point Analysis
PA017 SW Engineering II → Aspects of SW Development
Management
Jaroslav Ráček Josef Spurný
Faculty of Informatics, Masaryk University
October 25, 2022
Motivation
For COCOMO, we need KLOC estimation which might be
challenging to estimate correctly (remember 4:1 variance in
initial estimation)
Can we avoid that...?
Can we focus on the user perspective instead...?
Yes, we can – Function point analysis
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 2 / 33
Core ideas
In the analytical phase of project, we collect detailed functional
requirements on a system
Hence, we can focus on functionality to estimate SW cost
Since we do not count KLOC, we are language-independent
(high-level vs. low-level language)
In its essence, producing software is manufacturing process
requiring human work
Therefore, we simply need to determine:
Manufacturing unit
Work cost for producing this unit
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 3 / 33
Estimations using function points
Function point = normalized software project measure
Measurement focused on application aspect, not on technical
aspect
Size measure, not complexity measure
Measurement of functions and data, not on code
International Function Point Users Group - www.ifpug.org
Capers Jones:
Applied Software Measurement (1997)
Estimating Software Costs (1998)
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 4 / 33
Function point - principle
Preliminary estimation with limited information
At the end of analytical phase, when we have screen mockups
and detailed description of functionality, we can calculate FP
count with relatively good precision (±10%)
Measures used: inputs, outputs, enquiries, internal memories,
external memories
General estimation concept
Est = project size * complexity * risk factors
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 5 / 33
Function points overview
Link to source
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 6 / 33
Types of function points
FP related to transaction functions
EI – external inputs
EO – external outputs
EQ – external enquiry
FP related to data functions
ILF – internal logical files
EIF – external interface files
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 7 / 33
ILF – internal logical files
1 ILF = each large logical set of user data or information used to
manage application
We include every logical file or each logical data set from
end-user perspective that are created, used or maintained by the
application
Rather than physical files, we count logical data set as perceived
by the end-user, or as it is defined during analysis or design
phase of the project
We exclude files that are not accessible to end-user through UI
and that are not maintaned independently
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 8 / 33
ILF – internal logical files
Logical entity or entity group from end-user perspective = 1 ILF
Logical internal file generated or maintained by an application =
1 ILF
Matrix or file maintained by end-user = 1 ILF
Data file or configuration file used by an application to process
data = 1 ILF
Attribute entity maintained only by parent entity = 0 ILF
Associative entities containing only key attribute = 0 ILF
Internal temporary or classifier entity = 0 ILF
A file created as a result of using certain technology (e.g., index
file) = 0 ILF
External sample file, read only = 0 ILF
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 9 / 33
EIF – external interface files
We include every large logical set of user data or control
information used by an application
These data/information are maintained by an external
application. We include every logical file or logical data set from
end-user perspective
We include every large logical set of user data or control
information that are collected from other application via an
interface
The extraction shall not trigger a change to any internal logical
files. If this happens, then we count this as External Input (EI)
instead
Example: stationary radar for speeding
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 10 / 33
EIF – external interface files
Files or records extracted from other application (used as
references, links) = 1 EIF
Database read by other application = 1 EIF
Internal logical file of other application used as a transaction = 0
EIF, 1 EI
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 11 / 33
EI – external inputs
We include every unique user data or enquiries that enter an
application via interface and perform CRUD operation over
internal logical data file
We include control information which enters via interface and
ensures consistency with user-defined function
External input shall be considered as unique if design requires
processing logic that is different from other external inputs
Typically, each user screen (data entry set) is one FP, it doesn’t
matter whether it is done via one screen (one big page) or
multiple consecutive pages
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 12 / 33
EI – external inputs
GUI screen allowing to add, edit or cancel = 3 EI
Set of related screens that are processed as one transaction = 1
EI
Two different screens with different layout of data entry, but
with same processing logic = 1 EI
Two different screens with same format, but with different
processing logic = 2 EI
Single screen with multiple unique functions = 1 EI per each
function
Automated data input or transaction from other application = 1
EI per each transaction
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 13 / 33
EI – external inputs
Entry of user inputs/commands into application = 1 EI
Data entry (OCR) with one transaction = 1 EI
Data modification function following a enquiry = 1 EI and 1 EQ
Individual choices from menu on a screen = 0 EI
Modification of a matrix or file maintained by an end-user = 1 EI
Duplicate screen that has been already included = 0 EI
External inputs used only because certain technology is used = 0
EI
Make a selection from list of values (dropdown menu) = 0 EI
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 14 / 33
EO – external outputs
We include every unique data or control data leaving application
via an interface
External output is considered as unique if it has unique data, or
if design of an external system requires different processing
method
External outputs are often reports, output files sent to other
application or messages for user
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 15 / 33
EO – external outputs
Data export on a screen = 1 EO
Summarizing report – batch processing = 1 EO
Automated data export or transaction towards other applications
= 1 EO
Error messages returned as a result of input transaction = 0 EO
Backup files = 0 EO
Output shown on screen and sent to printer = 2 EO
Output files created for technical purposes = 0 EO
Output shown as bar chart as well as pie chart = 2 EO
Output showing result of calculation = 1 EO
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 16 / 33
EQ – external enquiry
We include every unique input/output tuple where input is cause
and output is effect
EQ is considered unique when it cause different data element
types to be produced, or when it requires different data
processing logic to be used (when compared to other EQ)
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 17 / 33
EQ – external enquiry
Online input followed by online output without change to data
files = 1 EQ
Enquiry followed by a change input = 1 EQ and 1 EI
Input and output on a help screen = 1 EQ
Online input leading to printing without data change = 1 EQ
Selection from list of values with dynamic data = 1 EQ
Selection from list of values with static data = 0 EQ
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 18 / 33
Function point calculation
Before the calculation, we have to classify EI, EO, EQ, ILF, EIF by
measures
Measure low average high total
EI x 3 x 4 x 6
E0 x 4 x 5 x 7
EQ x 3 x 4 x 6
ILF x 7 x 10 x 15
EIF x 5 x 7 x 10
Total unmodified function points:
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 19 / 33
Input complexity measure (EI, EQ)
FTR = File Types (User Data Groups) Referenced
DET = Data Element Type (Attribute)
RET = Record Element Type (User View)
FTRs 1-4 DETs 5-15 DETs 16+DETs
0-1 low low average
2-3 low average high
4+ average high high
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 20 / 33
Output complexity measure (EO, EQ)
FTR = File Types (User Data Groups) Referenced
DET = Data Element Type (Attribute)
RET = Record Element Type (User View)
FTRs 1-4 DETs 5-15 DETs 16+DETs
0-1 low low average
2-3 low average high
4+ average high high
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 21 / 33
File complexity measure (ILF, EIF)
FTR = File Types (User Data Groups) Referenced
DET = Data Element Type (Attribute)
RET = Record Element Type (User View)
RETs 1-19 DETs 20-50 DETs 51+DETs
0-1 low low average
2-4 low average high
5+ average high high
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 22 / 33
General system characteristics – evaluation scale
So far, we have focused only on the data perspective
However, we need to consider other parameters which would
impact work expenditure to deliver given system
These parameters include performance, criticality, security and
other aspects
14 characteristics evaluated on a scale according to an impact
on application, where
0 = no impact
1 = random impact
2 = low impact
3 = average impact
4 = significant impact
5 = substantial impact
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 23 / 33
General system characteristics – factors
1. Does system require reliable backup and recovery?
2. Are data communications required?
3. Is there a distributed processing?
4. Is performance critical?
5. Will the system be operational in current intensely used
operation environment?
6. Does system require online data entry?
7. Does online data entry require input transactions over multiple
screens or operations?
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 24 / 33
General system characteristics – factors
8. Are main files maintained online?
9. Are input, outputs, files and enquiries complex?
10. Is internal processing complex?
11. Is the code designed such that it is reusable?
12. Are conversion and installations included in the design?
13. Is system designed to be installed multiple times for different
organizations?
14. Is system designed to support changes and to be user-friendly?
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 25 / 33
Function points count
Function points count
= (0.65 + GSP
100 ) ∗ UFP, where
GSP = sum of general system characteristics evaluation
UFP = unmodified function points
Note: FP can be used to estimate any type of SW: desktop or mobile
app, SW in cars, home appliances, etc.
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 26 / 33
New and modification projects
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 27 / 33
How to calculate FP
1. Identify and calculate ILF, EIF, EI, EO, EQ.
2. For each ILF and EIF, identify RET and DET count. For each EI,
EO, EQ, identify FTR and DET.
3. Use complexity matrix to determine count of simple, average
and complex items of EI, EO, EQ, ILF and EIF.
4. Calculate unmodified FP count.
5. Evaluate 14 system characeristics.
6. Sum up the characteristics to determine Technical complexity
factor.
7. Calculate Modified FP count = SW size from functional
perspective
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 28 / 33
Size estimation (Caspers Jones)
1 FP = X commands (LOC)
Basic assembler: 320
Macro assembler: 213
C: 128
FORTRAN: 107
C++: 64
SQL: 13
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 29 / 33
Further estimations
FP1.15 ≈ SW project paper documentation pages count
FP1.2 ≈ test scenarios count
FP1.25 ≈ error potential for new SW
FP0.4 ≈ calendar months needed for SW delivery
FP
150 ≈ employees needed for SW delivery
FP
750 ≈ employees needed to maintain SW in desired state
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 30 / 33
Measurement examples
7 student projects (Bachelor/Masters theses, seminar projects)
written in C, C++, Java and Delphi @ FI MU were measured
FP calculation was done, LOC estimation based on this
calculation, then compared to reality
Time required to deliver project was estimated using COCOMO2
and FP, results were compared
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 31 / 33
Comparison of real and estimated code size
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 32 / 33
Comparison of real and estimated project time
Jaroslav Ráček, Josef Spurný ·Function Point Analysis ·October 25, 2022 33 / 33