Chapter 5 Software effort estimation
Further exercises and pointers
1. The size (that is, the effort needed to complete it) of any task will depend on its
characteristics. The units into which the work is divided will also differ. Identify the
factors affecting the size of the task and the work units for the following activities:
• installing computer workstations in a new office
• transporting assembled personal computers from the factory where they were
assembled to warehouses distributed in different parts of the country
• typing in and checking the correctness of data that is populating a new data
base
• system testing a newly written software application
As will be seen the exact dividing line between what influences task size and what is
a productivity driver can be hazy.
Installing workstations.
Task size drivers might include:
Number of workstations to be installed
Number of locations at which workstations are installed
Geographical spread
Productivity drivers might include:
Novelty of the technology/experience of installers
Type of workstation
Availability of access to premises
Transporting assembled personal computers
Task drivers might include
Number of units for estimating loading/unloading
Number of loads (units going to one warehouse) – group of units in a
delivery
Distance of warehouse from factory – traveling time
Productivity drivers might include
Time per unit needed for loading
Size of vehicles
Average speed of delivery vehicles
Time per unit needed for unloading
Typing in and checking input
Tasks size drivers might include:
Number of records to be input
Number of items per record to be input
Productivity drivers might include:
Clarity/availability of input documents
Experience/capability of inputters – this could affect the number of records
that need correction
Ease of making corrections
Systems testing
Task size drivers might include:
Number of test cases
Number of requirements – could be used to estimate number of test cases
Items to be input, items output per test case
Productivity drivers might include:
Ease of test set up
Errors found – this will require retesting
Availability of test automation tools
2. If you were asked as an expert to provide an estimate of the effort needed to make
certain changes to an existing piece of software, what information would you like to
have to hand to assist you in making that estimate?
The estimation of the effort to change a software component requires a different
approach to the development of completely new software as a major factor is the
structure of the existing software.
The estimator might need to know:
The nature of the changes required
Ideally they should be told about the reasons for the change, as they may be
able to suggest alternatives and more economic ways of meeting the
requirement
The structure of the software to be changed
Some idea of which parts of the software need changing
Who will be carrying out the changes – will it be someone who is already
familiar with the software to be changed, or will additional time have to be
allocated for familiarization?
3. A small application maintains a telephone directory. The database for the
application contains the following data types:
Staff reference Room number
Surname Telephone extension
Forenames E-mail address
Title Fax number
Department code
Transactions are needed which:
i) set up new entries;
ii) amend existing entries
iii) delete entries
iv) allow enquirers to list on line the details for a particular member of staff
v) produce a complete listing of the telephone directory entries in alphabetical
order
a) Use this scenario to produce an estimated Mark II FP count. List all the
assumptions you will need to make.
b) Another requirement could be to produce the listing in (v) in departmental order. In
your view should this increase FP count and if so by how much?
a)
transaction inputs outputs entities
accessed
set up new staff reference 9 error 1 directory entry 1
entry to fax number message
in data type list
amend (display) staff reference 1 surname to 9 directory entry 1
fax number or
error
message
amend (update) surname to fax 8 error 1 directory entry 1
number message
delete entry staff reference 1 surname, 3 directory entry 1
forenames
(as check)
error
message
enquiry staff reference 3 full details + 10 directory entry 1
or surname, error
forenames message
listing trigger 1 full details 9 directory entry 1
totals 23 x 0.58 = 13.34 33 x 0.26=8.58 7x 1.66 =11.62
grand total 33.54
b) It can argued that the data presented in the two reports are the same, so that they
are logically the same and so should be counted only once. Most development
environments have easy to use features for sorting data and so the actual amount of
work needed to produce the second report would probably be negligible. However, if
you have every tried to look up details in an unsorted list, the fact that it is sorted can
provide considerable value.
There is therefore a debate among FP practitioners about whether FPs should reflect
the value of the system to the user or the amount of work that is needed to develop
it. This is particularly an issue in relation to the question of reusable components.
4. The following details are held about previously developed software modules.
module inputs entity outputs days
types
accesse
d
a 1 2 10 2.60
b 10 2 1 3.90
c 5 1 1 1.83
d 2 3 11 3.50
e 1 3 20 4.30
A new module has 7 inputs, one entity type access and 7 outputs. Which of the
modules a to e is the closest analogy in terms of Euclidean distance?
module inputs entity outputs days euclidean
types distance
accessed from new
a 1 2 10 2.6 6.78
b 10 2 1 3.9 6.78
c 5 1 1 1.83 6.32
d 2 3 11 3.5 6.71
e 1 3 20 4.3 14.46
new 7 1 7
Module c would appear to provide the best analogy as it is at the least Euclidean
distance from the new module. This provides a base estimate of 1.83 days
5. Using the data in further exercise 4 above, calculate the Simons Mark II FPs for
each module. Using the results, calculate the effort needed for the new module
described in additional exercise 4. How does this estimate compare to the one based
on analogy?
module inputs entity outputs days FPs
types
accesse
d
a 1 2 10 2.6 6.50
b 10 2 1 3.9 9.38
c 5 1 1 1.83 4.82
d 2 3 11 3.5 9.00
e 1 3 20 4.3 10.76
totals (a) days (b) fps
16.13 40.46
productivity FPs 2.51
/per day (b/a)
new 7 1 7 new FPs 7.54
(c)
estimate for new 3.01
(c/(b/a))
Note that this is higher than that estimate given in 4. Partly this is because the new
project may be closest to module c but it is bigger in terms of the number of inputs
and outputs that it has. It also has an estimate bigger than that for module a because
Mark II FPs assume that input procedures are more difficult to implement than
outputs and therefore gives them a higher weighting.
6. Given the project data below:
projec inputs output entity system program developer
t s accesses users -ming days
languag
e
1 210 420 40 10 x 30
2 469 1406 125 20 x 85
3 513 1283 76 18 y 108
4 660 2310 88 200 y 161
5 183 367 35 10 z 22
6 244 975 65 25 z 42
7 1600 3200 237 25 y 308
8 582 874 111 5 z 62
X 180 350 40 20 y
Y 484 1190 69 35 y
Note X and Y are new projects for which estimates of effort are needed.
a) What items are size drivers?
inputs, outputs, entity accesses (system users for certain aspects) – the more of
these there are the larger the development job
b) What items are productivity drivers?
Programming language – the number of lines of code that can be produced in a day
will depend, in part, on the programming language
c) What are the productivity rates for programming languages x, y and z?
x 10 FPs a day
y 7 FPs a day
z 12 FPs a day
d) What would be the estimated effort for projects X and Y using a Mark II function
point count?
project inputs outputs entity system users program- developer FPs
accesses ming days
language
1 210 420 40 10 x 30 297.4
2 469 1406 125 20 x 85 845
sub-totals 115 1142.4
productivity 9.9
(FPs/day)
3 513 1283 76 18 y 108 757.15
4 660 2310 88 200 y 161 1129.48
7 1600 3200 237 25 y 308 2153.42
sub-totals 577 4040.05
productivity 7.0
(FPs/day)
5 183 367 35 10 z 22 259.66
6 244 975 65 25 z 42 502.92
8 582 874 111 5 z 62 749.06
sub-totals 126 1511.64
productivity 12.0
(FPs/day)
X 180 350 40 20 y 261.8
Y 484 1190 69 35 y 704.66
The above table gives the FP counts. Using the productivity rate for programming
language y, the estimate for Project X would be 262/7 i.e. 37 days, and for Project Y
705/7 i.e. 101 days
e) What would be the estimated effort for X and Y using an approximate analogy
approach?
Project X seems closest to Project 5 which provides an estimate of 22 days, and
Project Y seems to be closest to Project 3 which gives an estimate of 108 days
.
f) What would have been the best estimating method if the actual effort for X turns
out to be 30 days and for Y turns out to be 120 days? Can you suggest why the
results are as they are and how they might be improved.
If we use a measurement based on the percentage error, calculated as absolute
(actual-estimate)/actual we get the following:
Project X Project Y
FP method 23% 16%
analogy 27% 10%
This illustrates that no one estimating method can be assumed to be the most
accurate in all cases.
The analogy estimate for Project X might have been improved by adjusting for the
fact that Project X is to be written in programming language y while Project 5 was
written in programming language z
7. A report in a college time-tabling system produces a report showing the students
who should be attending each time-tabled teaching activity. Four files are accessed:
the STAFF file, the STUDENT file, the STUDENT-OPTION file and the TEACHING-
ACTIVITY file. The report contains the following information:
Teaching activity reference
Topic name
Staff forenames
Staff surname
Title
Semester (1 or 2)
Day of week
Time
Duration
Location
For each student:
student forename
student surnames
Calculate the Mark II FPs that this transaction would generate. Can you identify the
factors that would tend to make the two methods generate divergent counts?
Mark II
Inputs 1 ( a trigger) x 0.58 = 0.58
Entity accesses 4 x 1.66 = 6.64
Outputs 12 x 0.26 = 3.12
Total 10.34
