Function Point Estimation and Demonstration on Website Projects

Compared with software projects, the estimation of website projects is not mature enough without any methods for size or workload estimation. This paper researched Function Point Analysis (FPA), the most important size estimation method in software projects and then constructed Commerce Website Function Point Analysis (CWFPA) by introducing principles of FPA into commerce website projects. The paper reserved the basic definitions of software FPA for CWFPA and researched some key factors combined with characteristics of commerce website projects including the functional complexity, function point weight and value adjustment factor. In the end, a commerce website project PET was given as a demonstration of CWFPA.


Introduction
According to IT project work breakdown structure, IT projects can be divided into four types: website project, communication engineering project, system integrated project and software project (Feng et al., 2009).Software projects and website projects have something in common in project management (Mustafa & Bingunath, 2003;Alshawi, 2000;Leonard-Barton & Sinha, 1993).However website project management is less regular than software projects (Jiudian, 2002).IT project size estimation is an important part of IT project valuation.At present, software projects have already had systematic size estimation methods, the most typical of which is Function Point Analysis (FPA).Web pages are usually adopted as the measure unit of website project for it is the web pages linked by several hyperlinks that are directly presented in front of users in website projects.So far, website projects don't have systematic estimation methods.In the early phases, website projects and software projects can only obtain the function requirements of owners thus they can be estimated from the perspective of owners.The ultimate functions of website projects and software projects have a lot in common (Mustafa & Bingunath, 2003;Leonard-Barton & Sinha, 1993), so if we introduce FPA into website projects estimation and make research combined with the characteristics of website projects, maybe an effective size estimation method of website projects can be constructed.This paper introduces the principles of FPA into commerce website projects and constructs the Website Function Point Analysis (CWFPA) with detailed research on some key factors in function point estimation in commerce website projects.In the end, a pet website is given as a demonstration of CWFPA.

Principles of CWFPA
The workload of commerce website projects and software projects is intangible, which can be divided into function points.Therefore the function point estimation of website projects can take example by the principles of FPA.FPA is a function-oriented estimation model for software firstly brought forward by Allan Albrecht (1979).Meanwhile website projects can use principles of improved FPA for reference including Feature Point (Jones, 1987), NESMA (Netherlands Function Point User Group, 1997), 3D Function Points (Whitmire, 1995) and so on.According to the principles of software project FPA, Website Function Point Analysis (CWPFA) can be constructed as Figure 1.

Definitions of CWFPA
Commerce website projects and software projects have a lot of similarities, so function point estimation of website projects can borrow the basic definitions of software FPA.
(1) Function type and functional complexity.Commerce website project function points have two types including data functions and transactional functions.Data functions are either internal logical files (ILFs) or external interface files (EIFs), while transactional functions are either external inputs (EIs), external outputs (EOs) or external inquiries (EQs).These five function types have a certain complexity.The ILF and EIF complexity depends on the count of corresponding DET and RET, while the EI, EO and EQ complexity lies on the count of corresponding DET and FTR.
(2)Function point weight.Every complexity of every function type is given one corresponding weight.The purpose of setting function type weight is to translate the function points of different complexity into ones of unified scaling, thus obtaining UFP count.
(3)Value Adjustment Factor (VAF).Every characteristic of commerce website project are given a Degree of Influence (DI) ranged on a scale of zero to five, which represents the contribution of each characteristic to the whole project.The Total Degree of Influence (TDI) of all the characteristics can be calculated by DI.TDI can directly decide the VAF.The product of UFP and VAF is AFP, which is shown as Formula (1).

Key Factors of CWFPA
Although CWFPA reserves parts of the basic definitions of FPA in software projects, the treatment or calculation of these basic definitions depends on the characteristics of commerce website projects.Compared with software FPA and other improved FPA, CWFPA has a largely different treatment or calculation on some key factors including the functional complexity, function point weight and VAF.

Functional Complexity of Commerce Website Projects
In software FPA and other improved FPA, functional complexity mostly depends on the range of DET and RET (or FTR) (Garmus & Herron, 2003).The range of DET and RET (or FTR) has only three levels including low, average and high.For each function type, when one of DET and RET (or FTR) remains unchanged and the other changes a lot, its functional complexity remains unchanged; while when one of DET and RET (or FTR) remains unchanged and the other changes between two adjacent numbers, its functional complexity changes.The above analysis shows that the setting of functional complexity in software FPA and improve FPA has obvious shortages.
As the number of DER and RET (or FTR) determines the functional complexity while functional complexity determines the function point weight, it is considered that the function point weight can be directly determined by the number of DER and RET (or FTR) in CWFPA.That is to say, there is no need to determine the functional complexity.The detailed setting of function point weight is discussed in the following section.

Function Point Weight of Commerce Website Projects
In software FPA and other improved FPA, function point weight depends on its functional complexity.Therefore the setting of function point weight has the same problems as functional complexity.It is necessary to discuss the setting of function point weight in CWFPA.Take ILF as an example.According to the complexity matrix and unadjusted function point table of ILF, the corresponding relationship between the number of DET and RET and function point weight can be shown in Table 1.
Table 1.The corresponding relationship between the number of DET and RET and function point weight for ILF (before data processing)

Function point weight
The number of DET 1~19 20~50 ≥51 The number of RET This table is based on International Function Point Users Group (1999).
In order to obtain the relationship between the function point weight and the number of DET and RET, we adjust the values in Table 1 into Table 2 according to the principles as follows: (1) for the number of DET and RET, take the median of each range as its value; (2) for ranges without upper limit, take the lower limit as its value; (3) for function type weight, transfer it into the median value.x y , ( 0,1, 2; 0,1, 2)   i j . As can be seen in Table 3, ILF has nine known numbers.Then the problem is to seek for an approximate function ( , ) P x y to make it satisfied with the condition ( , ) , ( 0,1, 2; 0,1, 2) . The paper introduces interpolation method in numerical analysis to this problem.Fixing x , considering ( , ) f x y to be the function of y, make parabolic interpolation of y for ( , ) P x y Thereinto,

y y y y u y y y y y y y y y u y y y y y y y y y u y y y y y
Based on Formula (2), make parabolic interpolation of x for ( , )   y P x y with interpolation nodes o x , 1 x , 2 x .Recording ( ) i w x as the interpolation cardinal function, use Lagrange Interpolation, and the interpolation function obtained is ( , ) x y P P f x y as Formula (4).
x y P P f x y is ( , ) P x y the approximate function that we seek for.It approximately expresses the function relation that ( , ) x y is mapped to z .
For ILF, combined with Formula (3)~(5) using the data in Table 3, the relationship between z and ( , ) x y of ILF can be obtained as Formula (6).
Similarly, the relationship between z and ( , ) x y of EIF, EI, EO and EQ can be shown in Formula ( 7)~(10).(10) By Formula (6)~(10), the function point weight can be determined by the number of DET and RET (or FTR) of each function type in commerce website project, thus obtaining the UFP of the commerce website project.

VAF of Commerce Website Projects
In software FPA and improved FPA, VAF depends on the DI of General System Characteristic (GSC).But some GSCs have lost the functions borne in the initial definition thus losing its contribution as adjustment factors.In the mean time, some new characteristics have revealed in the function point estimation of commerce website projects as the development of technology and the movement of network environment.Therefore, the existing FPA can not effectively calculate the VAF in commerce website projects.

General Module Characteristics of Commerce Website Projects
(1) The concept of General Module Characteristics of commerce website projects.The paper puts forward the concept of General Module Characteristics (GMCs) of commerce website projects by subdividing the concept of GSCs in software projects according to the existing problems in software FPA.The commerce website project GMCs refer to the characters that can reflect the different attributes of commerce website project modules.There into, commerce website modules refer to the subsystems with complete code and independent functions in one commerce website project.
(2) The setup principles of commerce website project GMCs.Commerce website projects accord with the common characteristics of IT project.For example, it is a one-time activity with a certain task range, quality requirement, schedule requirement and resource limit.Besides, commerce website project has its own characteristics including the universality of involved areas, the inaccuracy of target, the intensity of intelligence, the uncertainty of web style, and the frequency of requirements changes and so on.Therefore, the setup of website project GMCs should obey the following principles: a.If the meaning of GSC is suitable for commerce website project module and its name is suitable for modules, take it as GMC; b.If the meaning of GSC is suitable for commerce website project module while its name is not fit for modules, take it as GMC after modifying its name suitable for commerce website modules; c.If the GSC has lost its meaning as an adjustment factor to different commerce website modules in the present system environment, delete it; d.If there is a characteristic that does not exist in software FPA but has great influence on commerce website projects in the present network or system environment, add it to GMC.
(3) The determination of commerce website project GMCs.According to the setup principles of commerce website project GMCs, nine GMCs can be obtained by the adjustment of the fourteen GSCs in software FPA combined with the characteristics of commerce website projects.These nine GMCs include Data Communications, Performance, Transaction Rate, Online Update, Complex Processing, Maintenance Ease, New Technology Use, Security and Exploitation Level.Thereinto, New Technology Use, Security and Exploitation Level are newly-added GMCs with their descriptions to determine DI shown as Table 4~ Table 6.Other six characteristics still adopt the descriptions in software FPA.There are security restrictions for 20%~40% of the input, update and searches of specific fields.For example, the commerce website project considered confidential by the government 3 There are security restrictions for 40%~60% of the input, update and searches of fields.

4
There are security restrictions for 60%~80% of the input, update and searches of specific fields.

5
There are strict security restrictions and it must maintain the audit data produced by visitors and specific data they access.

A Case Study
This chapter will give a demonstration of CWFPA combined with an electronic commerce website project.The author has participated in the requirement analysis and detailed design as well as coding work of this commerce website project thus familiar with the overall project.That is why this commerce website project is chosen.This commerce website project is called PET.All the statistics here are reported at most two decimal places.

Background of PET
core function of PET is the online sales of pets.By inquiring can scanning the pet information, the client can buy pets on line.Meanwhile, PET also has other functions such as providing various kinds of information.PET has four modules, as is shown in Figure 2.There are eight datasheets of PET, the main information of which is shown in Table 7. Fields in different datasheets vary.(2) Calculation of VAF.PET has four modules and each module has nine GMCs.In order to obtain the VAF in WFPA, the paper has made a questionnaire to gain the weight and DIs of GMCs of PET modules according to Formula ( 12).This survey put out one hundred copies of questionnaire and took back ninety-two copies with experienced project developers as its object.All the data used in this paper is from the statistic results of this survey.The paper use square law method to deal with the result of paired comparison to obtain the weight of the GMCs.
a. VAF of basic information module.According to the result of the survey, the weight of the nine GMCs of PET's basic information module ( 0 1q W ) and their DIs is as Table 13.
The VAF of basic information module can be calculated according to Formula (12):  (2) Calculation of VAF.Grading the present fourteen GSCs according to the actuality of PET, we can obtain the following result shown as Table 15.

A Comparative Analysis of the Results
The function points and function points count of PET can be calculated as Table 16 based on CWFPA and FPA combined with the actual exploitation of PET.As can be seen in Table 16, compared with actual function points count, the results obtained according to the CWFPA is more exact.Compare with the results of present FPA, the veracity of basic information module using CWFPA is 12.4% higher, user information management module 4.2% higher, order information management module 6.45% higher and online vote module 15.75% higher.It is obvious that pure information browse have more differences in function point estimation.Take basic information module for example, the difference lies in that the number of DET and RET (or FTR) of function types are less thus the function type weight is lower in CWFPA than in present FPA; while the function types of user information management module, order information management module and online vote module are more complex.CWFPA is more suitable for actual exploitation.

Conclusions
(1) The unique of commerce website projects has unique estimation requirement.Commerce website project has its own characteristics including the universality of involved areas, the inaccuracy of target, the intensity of intelligence, the uncertainty of web style, and the frequency of requirements changes and so on.These characteristics reveal that the present FPA and improve FPA can not satisfy the demand of commerce website projects estimation.Meanwhile, CWFPA makes up the present FPA in theory and has a larger application scope.
(2) In allusion to the problems that existed in the setup of functional complexity and function point weight in software FPA, the paper wiped off the determination of functional complexity and obtained an approximate function between function point weight and the number of DET and RET (or FTR) for one specific function type, thus eliminating the sudden change of weights when the number of DET and RET (or FTR) changes.
(3) Compared with the concept of GSC, website project is more suitable to adapt the concept of GMC.By reviewing the applicability of the former characteristics, commerce website project is given nine GMCs including Data Communications, Performance, Transaction Rate, Online Update, Complex Processing, Maintenance Ease, New Technology Use, Security and Exploitation Level combined with the characteristics of commerce website projects.The weights of these nine GMCs need to be determined according to the reality.
(4) The demonstration part made a compared analysis of CWFPA and the present FPA.CWFPA theory analysis fill the defect of the existing function point analysis method, and expand the range of application.The empirical part of this paper proves the feasibility and accuracy of CWAFP.In CWFPA, the paper obtained five approximate functions through binary function interpolation method in numeric analysis.These approximate functions are not perfect and need a further research.

Table 4 .
Descriptions to determine DI of new technology the above.*These new technologies include cookie, SSI, Flash, XML, XHTML.

Figure 2 .
Figure 2. Structure of PET

Table 2 .
The corresponding relationship between the number of DET and RET and function point weight for ILF (after data processing) Suppose the number of DET is x, the number of RET (or FRT) is y, the function point weight is z.
ij z in ( , ) i j

Table 3 .
Known numbers of ILF

Table 5 .
Descriptions to determine DI of security

Table 6 .
Descriptions to determine DI of exploitation level When calculating VAF, software FPA and improved FPA simply sum up all the DIs and gain TDI in which way all the GSCs are of the same importance that is the same weight.In Fact, different GMCs in commerce website project have different importance.It is not true to give the same weight.Therefore, the weight of different GMCs needs to be worked out.There are differences in different commerce website project modules, so different modules should have different weights.The paper just puts forward the idea to solve the problem with detailed calculation of GMCs weights in the demonstration part.

Table 7 .
Main information of datasheets of PETThrough analyzing the information in PET datasheets, the data functions and transactional functions of different modules of PET can be obtained as Table8~Table11.

Table 9 .
Function type analysis of user information module

Table 10 .
Function type analysis of order information management module

Table 11 .
Function type analysis of online vote module Calculation of UFP count.According to the function type of PET modules, the UFP count of PET modules is shown in Table 12 combined with CWFPA.

Table 13 .
The weight of the nine GMCs of PET's basic information module ( That is to say, the VAF of basic information module is 0.88.By the Consistency Check of paired comparison matrix of this module, its consistency rate ..0.04 0.1 C R  , so it passes the Consistency Check.b. VAF of user information management module.Its calculation is the same as that of basic information module.By calculation, the VAF of user information management module is 0.94.By the Consistency Check of paired comparison matrix of this module, its consistency rate ..0.04 0.1 C R  , so it passes the Consistency Check.c.VAF of order information management module.Its calculation is the same as that of basic information module.By calculation, the VAF of order information management module is 1.02.By the Consistency Check of paired Calculation of UFP count.According to the function type analysis of PET modules, the UFP count of PET modules is shown in Table14combined with FPA.

Table 15 .
DIs of fourteen GSCs in FPA according to the actuality of PET

Table 16 .
A comparative analysis of the results of function points and Man-hour cost