View
142
Download
1
Category
Tags:
Preview:
DESCRIPTION
IWSM Presentation
Citation preview
ANALYSIS OF THE PER-UNIT WORK EFFORT AND PER-UNIT WORK COST OF THE
WEB INFORMATION PORTAL ENHANCEMENT – A CASE STUDY FROM POLAND
Beata Czarnacka-Chrobot, Ph.D.Professor, Warsaw School of Economics, Poland,
Institute of Information Systems and Digital Economy,e-mail: bczarn@sgh.waw.pl
Problem Statement (1/3)
This analysis presents a case study on tender competitition …concerning the enhancement of
Web Information Portal (WIP) of …one of the largest public institutions
in Poland: ZUS - Social Insurance
The COSMIC method is used as a software size measurement method in ZUS
Work cost/effort per unit of WBI enhancement should be measured with regard to 1 CFP
In competition one of the 3 potential developers (D3) offered possibility to enhance WIP system at the cost of 300 US dollars per 1 CFP
Problem Statement (2/3)
Other 2 attempted to prove that this is not possible at such unit cost
According to them, the unit cost was underestimated even several times
D1D2
D3
Problem Statement (3/3)
Analysis served as a main and sufficient basis for settling legal dispute between a company D3 and 2 competing companies
It was also necessary to prove that: the average per-unit work effort on the level of 16 work-
hours per 1 CFP offered by D3 is not underestimatedand to relate these two attributes: per-unit work effort and per-
unit work cost
D1D2
D3
Aim of the Paper
To demonstrate that:
It is possible to carry out enhancement of the WIP at such per-unit cost in Polish circumstances
The offer of D3 does not feature the so-called abnormally low tender price
(as defined in the Polish act “Public Procurement Law”)
This is possible to choose given tenderer (D3)
due to the legal reasons
Per-unit cost of software development/enhancement depends directly on
per-unit work effort/its inverse: productivity
difficult to determine - it depends on about 50 factors, e.g.:
Assumptions for the Case Study (1/2)
size and type of software
type of project
field of application
technological environment (e.g. language generation,
hardware platform and specific languages)
many other factors affecting
productivity
Documentation delivered by D3: 1) the subject of considerations is the work cost/effort per
unit concerned only size resulting from FUR (NFR were not considerated in this analysis)
2) Boundary functional size of system: 6500 CFP3) Type of software: dedicated business application (data-
driven), 4) Type of project: enhancement of the existing software
(specific actions not mentioned)5) PL: 3GL should be employed, in
particular Java/J2EE 6) Hardware platform: PCs.
Assumptions for the Case Study (2/2)
Appropriate resources of own benchmarking data, …which would allow to determine specific to a given
software organization work effort/cost per unit, …are not owned by Polish Social Insurance Institution
This situation is normal in Poland where COSMIC method has been employed for a relatively short time now
Usefulness of Benchmarking Data (1/3)
Usefulness of repositories with general and commonly available benchmarking data reveals
The largest repository with such data for dedicated software development/ enhancement projects (D/EP) with products measured using the FSM methods is managed by the ISBSG
Current version of the ISBSG repository contains data: concerning about 6000 projects from about 30 countries for software development and enhancement normalized according to the ISO/IEC 15939 standard [„Systems
and software engineering – Measurement process”] verified and representative of the current technology
The most important ISBSG report: „The performance of business application, real-time and component software projects. An analysis of COSMIC - measured projects in the ISBSG database”, made with the COSMIC cooperation
Usefulness of Benchmarking Data (2/3)
It analyses for various types of software systems the work effort per unit with regard to 1 CFP: • project delivery rate (PDR) = inverse of productivity
ISBSG report: large business application enhancement projects (above 1000 CFP) should be treated as development projects
ISBSG: three key factors determining work effort per unit and therefore cost per unit :
Usefulness of Benchmarking Data (3/3)
1) application size 2) generation of programming
language 3) hardware platform
4) particular programming
language
The PDR varies with regard to the software size:
The productivity median for BA DP increases (PDR median decreases) along with the growth of size
Economies of scales may be noted for BA DP
Work effort per unit decreases which should entail decrease of the cost per 1 CFP
Dedicated Business Applications Enhancement Per-unit Work Effort with regard to 1 CFP (1/4)
The analysis of PDR based on data from 162 new business application development projects [BA DP] (1999-2011):
Dedicated Business Applications Enhancement Per-unit Work Effort with regard to 1 CFP (2/4)
PDR(work-hours/1 CFP)
3GL Java/J2EE programming
language
PC
PDR median 24.5 23 23Minimum PDR 2.7 3 3Maximum PDR 330.6 139 13925% of the projects have PDR not higher than
14.3 15.8 8
PDR values dependent on programming language and hardware platform (PDR median normalized for 100-200 CFP size):
If this is possible to consider all 3 key factors, the data should be treated as initial for which correction coefficients should be applied:
Dedicated Business Applications Enhancement Per-unit Work Effort with regard to 1 CFP (3/4)
PC
Java (and similar) 20 ± 10
Application size in CFP
10-50 50-100
100-200 200-300 300-500 500-1000
1000+
PDR correction coefficient
1.2 1.1 1.0 0.9 0.8 0.65 0.5
Dedicated Business Applications Enhancement Per-unit Work Effort with regard to 1 CFP (4/4)
Conclusion no. 1: Taking into account the adopted assumptions and the
above presented data it should be stated that the work effort per unit (PDR) of WIP enhancement on
the level of 16 work-hours per 1 CFP cannot be regarded as underestimated work effort
The ISBSG report - limited to the analysis of the unit effort – not unit cost - with regard to 1 CFP
Data for unit cost may be found for the IFPUG/NESMA FSM and used in order to estimate work cost per unit with regard to 1 CFP indirectly: by taking into account dependencies between the
IFPUG/NESMA UFP and COSMIC FP (only examples to show differences):
Analysis of the dedicated BA EP unit cost with regard to 1 IFPUG UFP – indirect calculation (1/6)
Is there any report, which analyses unit cost with regard to 1 CFP...?
No.Author (year) Sample
sizeFormula (regression
analysis) R2
Number of CFP for UFP =
1001. Fetcke
(1999)4 CFP = 1.1 × UFP – 7.6
(UFP – number of IFPUG unadjusted function points)
0.97 102.4
2. Abran, Azziz, Desharnais (2005)
6
CFP = 0.84 × UFP + 18(UFP – number of IFPUG
unadjusted function points)
0.91 102
3. Desharnais, Abran (2006)
14 CFP = 1.0 × UFP – 3(UFP – number of IFPUG
unadjusted function points)
0.93 97
4. Van Heeringen (2007) 26 CFP = 1.22 × UFP – 64(UFP – number of NESMA
Unadjusted Function Points)
0.97 58
5. C. Jones (2007) - 1 UFP = 1.15 CFP(UFP – number of IFPUG
unadjusted function points)
- 86.9
6. Cuadrado-Gallego, Buglione, Domínguez-
Alda, de Sevilla, Gutierrez de Mesa,
Demirors (2010)
- 1 UFP 1 CFP(UFP – number of IFPUG
unadjusted function points)
- 100
The results of conversion differ greatly - it is not possible to determine precisely what is
dependency between IFPUG/NESMA UFP and
COSMIC FP
The researchers and practitioners suggest for general estimation purposes:
1 IFPUG/NESMA UFP ≈ 1 CFP Based on the ISBSG report „Software Project Costs” - the cost
per unit with regard to 1 IFPUG UFP: for majority of cases: from 300 USD to 1000 USD median: 716 USD
Analysis of the dedicated BA EP unit cost with regard to 1 IFPUG UFP – indirect calculation (3/6)
The work cost in particular country
Poland has ranked among countries having the lowest cost of work per 1 IFPUG UFP (estimated to be app. 155 USD on average)
Analysis of the dedicated BA EP unit cost with regard to 1 IFPUG UFP – indirect calculation (4/6)
Country Cost per unit (1 IFPUG UFP)Japan 1600Sweden 1500Switzerland 1450France 1425Great Britain 1400Germany 1300USA 1000
India 125Poland 155Hungary 175Thailand 180Venezuela 190Mexico 200Argentina 250
These differences result from macroeconomic factors and although they get smaller over time, they do remain significant
The ISBSG data - in over 80% concern projects executed in countries with several times higher cost per unit (even 10)
Therefore the cost per unit of delivering 1 UFP in Poland should oscillate around lower boundary of the cost indicated by the ISBSG data (300 USD per 1 IFPUG UFP)
Analysis of the dedicated BA EP unit cost with regard to 1 IFPUG UFP – indirect calculation (5/6)
The relation between the offered work effort per unit (16 work-hours per 1 CFP) and cost per unit (300 USD per 1 CFP): 1 work-day (8 work-hours) would cost about 150 USD, average
monthly cost of work = 3375 USD It is fairly high as for Polish conditions - it may comprise
average gross pay of a developer and company’s profit margin
Conclusion no. 2: The cost per unit (1 CFP) of WIP enhancement on the
level of approx. 300 USD cannot be regarded as abnormally low cost
Analysis of the dedicated BA EP unit cost with regard to 1 IFPUG UFP – indirect calculation (6/6)
There is no basis to consider the price of 1 CFP offered by a D3 company as abnormally low tender price as defined in the “Public Procurement Law”
The offer of D3 developer can be choosen due to the legal reasons
D3 company offered attributes close to the lower boundary value - it is now facing a difficult task to keep to them
The analysis served as a main and sufficient basis for settling legal dispute between a company D3 and 2 competing companies – in favour of D3 company
Conclusions
Thank you for your attention.
Any questions?
Beata Czarnacka-Chrobot, Ph.D.Professor, Warsaw School of Economics, Poland,
Institute of Information Systems and Digital Economy,e-mail: bczarn@sgh.waw.pl
Recommended