business intelligence for small enterprises: an open source approach

46
Department of Computer and Systems Sciences Rustam Aliyev Business Intelligence for Small Enterprises An Open Source Approach This thesis corresponds to 20 weeks of full-time work.

Upload: rstml

Post on 15-Nov-2014

111 views

Category:

Documents


1 download

DESCRIPTION

Business Intelligence for Small Enterprises: An Open Source ApproachMaster's Thesis WorkKTH, Department of Computer and Systems Sciences

TRANSCRIPT

Page 1: Business Intelligence for Small Enterprises: An Open Source Approach

Department of Computer and Systems Sciences

Rustam Aliyev

Business Intelligence for Small Enterprises

An Open Source Approach

This thesis corresponds to 20 weeks of full-time work.

Page 2: Business Intelligence for Small Enterprises: An Open Source Approach

ii

Abstract. During the last decade, Business Intelligence (BI) became inevitable technological advantage of the large enterprises which could afford to buy, implement and maintain BI solutions. These days, small size enterprises which form 98% of all enterprises in the EU have realised competitive and financial benefits of the BI. However, limited IT budgets of small companies and BI’s high TCO (total cost of ownership) may cause power gap between large and small enterprises persists to enlarge where latter will find it increasingly difficult to compete. This paper explores open source (OS) approach to BI, whether OS could be an alternative to the commercial solutions, and more important whether OS BI provides cost saving. It defines and evaluates grounds that used for comparison of OS and commercial BI solutions.

Keywords: business intelligence, open source, small enterprise, SME, cost-benefit analysis

Page 3: Business Intelligence for Small Enterprises: An Open Source Approach

iii

Acknowledgements. I would like to thank Ms. Maria Berghotlz for her tremendous support and assistance in the preparation of this thesis. In addition, special thanks are due to Mr. George Hodosi, whose comments were very important. I also thank Mr. Johan Dahlin and Mr. Fatih Poyraz for their invaluable input.

Page 4: Business Intelligence for Small Enterprises: An Open Source Approach

iv

Table of Contents

1 INTRODUCTION ..............................................................................................................................1

1.1 BACKGROUND .............................................................................................................................1

1.2 BUSINESS INTELLIGENCE ...........................................................................................................1

1.3 NATURE OF SMALL ENTERPRISES ............................................................................................3

1.4 RESEARCH PROBLEM .................................................................................................................3

1.5 RESEARCH GOAL .........................................................................................................................4

1.6 RESEARCH METHOD ..................................................................................................................4

1.7 LIMITATIONS ...............................................................................................................................5

2 EXTENDED BACKGROUND .........................................................................................................6

2.1 OPEN SOURCE SOFTWARE .........................................................................................................6

2.2 INFORMATION SYSTEM EVALUATION .......................................................................................7

3 VALUING BI SYSTEMS ..................................................................................................................9

3.1 IDENTIFYING BI COSTS ..............................................................................................................9

3.2 IDENTIFYING BI BENEFITS THROUGH DECOMPOSITION ..................................................... 10

3.3 IDENTIFYING BI BENEFITS THROUGH ITS CHARACTERISTICS ............................................ 12

3.3.1 BI Functionalities .......................................................................................................... 13

3.3.2 BI Solution Maturity ..................................................................................................... 15

4 RESULTS AND FINDINGS ......................................................................................................... 16

4.1 COMMERCIAL BI SUITES ......................................................................................................... 16

4.2 OPEN SOURCE BI SUITES ........................................................................................................ 17

4.3 INTERVIEW AND SURVEY OF BI CONSULTING COMPANIES ................................................. 19

4.4 EXPERIMENTAL SETUP ........................................................................................................... 21

4.4.1 System Overview ........................................................................................................... 22

4.4.2 Setup Results .................................................................................................................. 24

Page 5: Business Intelligence for Small Enterprises: An Open Source Approach

v

5 ANALYSIS AND DISCUSSION ................................................................................................... 25

5.1 COMPARISON OF BENEFITS .................................................................................................... 25

5.2 COMPARISON OF COSTS .......................................................................................................... 26

6 CONCLUSION ................................................................................................................................. 30

6.1 FUTURE RESEARCH ................................................................................................................. 32

7 BIBLIOGRAPHY ............................................................................................................................ 33

APPENDICES ......................................................................................................................................... 7-1

APPENDIX A. RAWAT’S CRITERIA FOR ASSESSING OS BI ........................................................... A-1

APPENDIX B. SURVEY QUESTIONS ................................................................................................ B-2

APPENDIX C. INTERVIEW AND SURVEY RESULTS ....................................................................... C-4

APPENDIX D. BUSINESS CASE DESCRIPTION ............................................................................... D-5

Page 6: Business Intelligence for Small Enterprises: An Open Source Approach

vi

List of Figures

FIGURE 1.1. BUSINESS INTELLIGENCE DATA LIFECYCLE ....................................................................................... 2

FIGURE 3.1. CALCULATING NET BENEFITS, ADOPTED FROM PENG MODEL ..................................................... 9

FIGURE 3.2. TYPICAL COSTS FOR BI IMPLEMENTATION PROJECT ..................................................................... 10

FIGURE 3.3. QUADRANT GRAPH FOR GENERIC BI BENEFITS CLASSIFICATION ................................................ 11

FIGURE 3.4. CRITERIA CATEGORIES FOR BI SUITE COMPARISON ...................................................................... 13

FIGURE 4.1. ENTERPRISE TYPES MAINLY TARGETED BY THE BI CONSULTING COMPANIES ......................... 19

FIGURE 4.2. CUSTOMERS OF THE BI CONSULTING COMPANIES ......................................................................... 19

FIGURE 4.3. COMPLETE BI SOLUTION FUNCTIONALITIES. ................................................................................. 20

FIGURE 4.4. AVERAGE COSTS OF THE COMMERCIAL BI SUITE IMPLEMENTATION ......................................... 21

FIGURE 4.5. DATA FLOW AND COMPONENTS OF THE EXPERIMENTAL SYSTEM .............................................. 22

FIGURE 4.6. SAMPLE TRANSFORMATION AND JOB IN PENTAHO DATA INTEGRATION ................................. 23

FIGURE 5.1. TCO OF THE COMMERCIAL AND OS BI OVER A FIVE-YEAR PERIOD ........................................... 28

FIGURE D.1. DOMAIN MODEL OF OPERATIONAL DATABASE .......................................................................... D-5

List of Tables TABLE 2.1. COMPARING OPEN SOURCE SOFTWARE LICENSE TYPES.................................................................. 7

TABLE 2.2. COMPARISON BETWEEN EXISTING BENEFIT MEASUREMENT MODELS ........................................... 8

TABLE 3.1. BI FUNCTIONALITIES CRITERIA GROUP FOR BI EVALUATION ....................................................... 14

TABLE 3.2. SOLUTION MATURITY CRITERIA GROUP FOR BI EVALUATION ...................................................... 15

TABLE 4.1. COMPARISON RESULTS OF THE COMMERCIAL BI SUITES ............................................................... 17

TABLE 4.2. COMPARISON RESULTS OF OS BI SUITES.......................................................................................... 18

TABLE 4.3. ACQUISITION COSTS FOR EXPERIMENTAL SETUP ............................................................................ 24

TABLE 5.1. COMPARISON OF BENEFIT COMPONENTS FOR THE COMMERCIAL AND OS BI ............................ 25

TABLE 5.2. COMPARISON OF COST COMPONENTS FOR THE COMMERCIAL AND OS BI .................................. 27

Page 7: Business Intelligence for Small Enterprises: An Open Source Approach

1

1 Introduction

1.1 Background

Information is considered the most valuable asset of any organization regardless of

the size of that organization. Every operation that organizations perform generates

lots of raw data. For instance, a simple sale of any product could generate huge

amounts of data, like date of sale, price, discount, customer name, address, other

demographic details like age, gender, which sales representative sold the product,

when the product was manufactured, raw materials, supplier information, and so

on. This raw data has to be converted into useful information for the decision

makers in order to improve performance of the organization. Considering the fact

that there are numbers of different business processes within any organization,

there is a definite need of a sophisticated information system. Secondly, the

availability of the right information on the right time to the right person is another

most challenging goal for any organization.

Information systems researchers and technologists have built and investigated

different decision support systems (DSS) for approximately 40 years. History of

developments in this area began with data collection techniques in the late 1960s.

Later, in the 1970s, there were theory developments and the implementation of

financial planning systems followed by spreadsheet DSS in the early and mid

1980s. At that point what the system provided was simple data collection that

involved historical data referred to as information discovery. Ever-increasing

amounts of data resulted in the creation of Data Warehouses, Executive

Information Systems (EIS), OLAP (Online Analytical Processing) and Business

Intelligence (BI) in the late 1980s and early 1990s. Finally, the chronicle ends with

knowledge discovery techniques such as data mining, knowledge-driven DSS and

the implementation of Web-based DSS in the mid-1990s (Power 2007).

1.2 Business Intelligence

There exist many professional definitions of BI; however none of them is a

standard. Business Intelligence is rather an umbrella term for a broad category of

applications and technologies for gathering, storing, analyzing, and providing

access to the data (Turban, et al. 2006, 423). Definitions usually encompass

personal and group DSS, EIS, data warehousing, and knowledge management

systems.

Page 8: Business Intelligence for Small Enterprises: An Open Source Approach

2

Internal

Data

External

Data

Personal

Data

Data

Warehouse

Metadata

Data

Marts

Data

Marts

OLAP,

Queries,

EIS, DSS

Data Mining

Data

Visualisation

Decision

Support

Knowledge

and its

Management

ResultsData AnalysisData StorageData Sources

Figure 1.1. Business Intelligence data lifecycle, adopted from (Turban, et al. 2006, 410)

In Business Intelligence, transformation of data into information and knowledge is

generally accomplished through the process depicted on figure 1.1. It starts with

raw data extraction from sources. These data are transformed into new formats

(usually defined according to metadata) and loaded into data warehouses or data

marts. These first two steps are usually served by ETL (extraction, transformation,

load) software. Analysis tools access the warehouse and data marts to get pieces of

data they need. The analysis is done with data analysis and mining tools which

look for patterns, and with intelligent systems, which support data integration.

Finally, the result of these activities are presented to decision makers using

different visualisation techniques or stored in organization’s knowledge base

(Turban, et al. 2006, 410).

Throughout this thesis work we refer to the components mentioned above as a BI

components or BI functionalities. Software product which contains a number of the

BI components is referred as BI suite, BI system or BI solution. Finally, by BI

implementation project (sometimes simply BI project or BI implementation) we

consider complete BI implementation life-cycle from business case assessment to

production release and maintenance.

According to Gartner’s (2007) annual survey of 1400 CIOs worldwide business

intelligence became no.1 technology priority in 2007. It is important to mention

that BI is a multi billion dollar market dominated by giant vendors such as Oracle,

SAP, SAS and Microsoft.

Another recent trend in the BI area has been observed in the adoption of open

source software. In the past few years a number of open source players have

entered the BI market. With the success of open source business models, many

commercial organisations have elaborated strategies to capitalize on them (Lerner

and Tirole 2000). A very recent research conducted by Ventana (Ventana 2006)

shows that 83% of organizations are considering, are in the process of deploying,

or have already implemented an open source BI solution.

Page 9: Business Intelligence for Small Enterprises: An Open Source Approach

3

1.3 Nature of Small Enterprises

According to the European Commission (EC) definition (EC 2003) small business

are defined as an enterprises employing less than 50 people and having less than

€10M annual turnover or annual balance sheet total. At the same time, small

enterprises represent 98% of all enterprises in the EU and employ 40-65% of the

workers in the private sector of the Member States (EC 2007). Consequently, small

enterprises can be considered as economically and socially important players in

the EU countries.

Another EC report (EC 2006) states that only about 10% of small firms used

specific ICT (Information and Communications Technology) solutions for

marketing, sales and procurement compared with 20% of medium-sized and

almost 30% of large firms. The same report says: “SMEs (Small and Medium

Enterprises) still suffer from limited understanding of ICTs and their potential,

limited budget for ICT investments and difficulty in recruiting ICT professionals”.

Even though the majority of small businesses use ICT solutions daily, this usage

usually engage mainly internet and email access. Very few small enterprises use

computers in decision support roles (Gibson and Arnott 2003).

1.4 Research Problem

There are many considerations and risks that organizations have to evaluate

before the adoption of BI solutions and the most important factor is the cost. This

is due to the fact that BI solutions are very expensive, both to purchase and

maintain (Raden 2007). Although, there is no exact figure because the

implementation of BI depends upon number of factors like number of end users,

functionalities, and so on. In fact, mostly large organizations can afford BI

implementation where the investment cost could be millions of dollars and still fit

in budget. This huge amount of investment is certainly beyond the reach of

medium and especially small size organizations. According to Gibson and Arnott

(2003), small businesses are often faced with limited access of finances to support

the purchase of business intelligence.

Now, on one hand, we have a large number of small businesses who cannot afford

implementation of expensive BI solutions from industry leaders. On the other

hand, these BI industry giants cannot offer entry level solutions without changing

their business models.

The exploitation of BI technology is important in the development of the small

enterprise sector. In their research, Gibson and Arnott (2003) concluded that: “If

large organizations are going to continue to exploit the latest decision-supporting

technologies, and small businesses continue to tread wearily in terms of adopting

Page 10: Business Intelligence for Small Enterprises: An Open Source Approach

4

modern business intelligence, the power gap will only continue to widen. If the

power differential between large and small businesses persists to enlarge, small

businesses will find it increasingly difficult to compete in a modern economy with

resulting significant social and economic destabilization”.

This is to say that there is a strong need to develop an understanding of

opportunities which open source Business Intelligence can bring to small

businesses.

1.5 Research Goal

The purpose of this research is to find out whether an open source (OS) BI suites

can facilitate for small enterprises to remain competitive. We will consider that OS

BI can be an alternative for small enterprises if the questions below answered

positively:

1. Does an OS BI provide enough business value for small enterprises to be

considered as an alternative to the commercial BI suites?

2. Does an OS BI solution provide cost savings in comparison with commercial

BI suites?

3. Does an OS BI actually fit into small businesses’ ICT budget?

1.6 Research Method

A comparative analysis of open source and commercial BI suites has been chosen

as a primary method in order to reach the goal and answer the questions stated in

section 1.5. To answer first and second questions we used divisive ("top-down")

approach to find grounds for the comparison of a BI suites’ value. Thus, the top

business criterion – the business value of the BI solution is divided into smaller

clusters unless the level of measurable and comparable criteria is reached (for

instance, we cannot estimate total cost of the system as an atomic criterion while

one of its sub-criterion “license fee” can be easily evaluated). Based on a number of

investigated “bottom” level criteria the most comprehensive OS BI suite is chosen

for further business value evaluations and comparisons with the value of

commercial BI suites.

To gather background information for the comparison, extensive literature review

of the latest on-going trends related to BI and business value evaluation methods

were conducted; published information about the BI products and formal BI

system documentation were also consulted.

This research project also used the most common qualitative research method

employed in information systems research, the case study. An experimental setup

Page 11: Business Intelligence for Small Enterprises: An Open Source Approach

5

of a chosen OS BI within a real small enterprise was used to provide

complementary information to answer research questions 1, 2 and 3 of section 1.5.

Finally, a number of BI integrators and solution providers were interviewed in

order to capture an average business value of a commercial and an open source BI

suites. Interviewees were asked to answer generic questions regarding BI market

as well as provide cost estimation for the case study mentioned above. In order to

increase accuracy of the results, more companies have been approached through

the web-based survey form where case study cost estimations were not included.

Interview and survey results mainly address questions 1 and 2, as well as helping

to answer question 3.

1.7 Limitations

We already mentioned that a Business Intelligence implementation is a very time

consuming initiative and includes many steps of data analysis and design. Thus,

due to lack of time and resources the experimental setup provided in this paper

has been limited to only one business process of one small enterprise.

Medium and even large enterprises could also benefit from OS BI. However, we did

not include these types of enterprises in the scope of our research.

Yet another limitation is BI components. As it has been mentioned above, BI is an

umbrella term for a broad category of applications. In our case we identified

broadly accepted components and defined subjective BI scope.

Page 12: Business Intelligence for Small Enterprises: An Open Source Approach

6

2 Extended background

2.1 Open Source Software

Open Source Software (OSS) is primarily defined as software which is freely

redistributable and includes the source code (Varner 1999). This is vastly different

from the mainstream software industry where source code is highly guarded and

programs are only distributed in their binary, non-modifiable format. The

development process of OSS also differs as it involves large number of software

developers at many different locations and organizations sharing code to develop

and refine software programs.

While the attention of businesses to the phenomenon of OSS has been recent, the

basic behaviours are much older in their origins. According to Lerner and Tirole

(2000), the tradition of sharing and cooperation in software development began in

early 1960s mainly with cooperative development efforts of the UNIX operating

system where programmers in different organizations shared source code. The

General Public License (GPL) was a noticeable innovation introduced by the Free

Software Foundation (FSF) in response to intellectual property right enforcements

by commercial companies in 80s. Users had to agree to make the source code freely

available and not to impose licensing restrictions on others, in exchange for being able

to use and modify GPL software. The widespread diffusion of Internet access in the

early 1990s led to a dramatic acceleration of open source activity. As we will

mention below, interactions between commercial companies and the open source

community also became commonplace in the 1990s. These interactions created

demand to bundle the cooperatively developed software with proprietary code

and led to adoption of more flexible licenses. The licenses under which OSS is

released today vary greatly, but two points that we mentioned in the beginning

(freely redistributable and available source code) remain consistent. Some of these

licenses and their characteristics are shown in table 2.1.

In the past years growing interest has brought large market diffusion and capital

investment to open source software. A number of open source products, such as

the Apache web server, dominate in their product category, while major

corporations, including Hewlett Packard, IBM, Sun and Microsoft have lunched

own open source projects (Lerner and Tirole 2000).

Today, major enterprises are running mission-critical functions on open source,

big vendors have lined up to support it, and reliable applications have emerged. In

a survey of 375 information executives, 54 percent said that within five years open

source would be their dominant server platform. 59 percent of respondents said a

lower total-cost-of-ownership (TCO) is open source’s primary strength and those

who have implemented it confirm huge TCO reductions (Koch 2003).

Page 13: Business Intelligence for Small Enterprises: An Open Source Approach

7

License Type

Code protected by copyright?

Can code be used in Closed

Source Project?

Can project that uses code, be

sold?

Must Source Code be

released?

Provides for

patents?

Public Domain No Yes Yes No No

BSD/MIT Yes Yes Yes No No

GPL (v2) Yes No No Yes No

LGPL Yes Yes Yes Yes No

MPL/CDDL Yes Yes Yes Yes Yes

CPL/EPL Yes Yes Yes Yes Yes

Table 2.1. Comparing Open Source Software License Types (Corbett and Ward 2006)

However, OSS has its own disadvantages. The same survey of IT executives shows

that 52 percent mentioned a lack of vendor support as open source’s primary

weakness. To fill this gap, major vendors such as Dell, HP, IBM, Oracle and Sun

recently have announced in various ways that they would begin supporting open

source products (Koch 2003).

2.2 Information System Evaluation

Limited budgets of small enterprises led project sponsors to demand an estimate

of financial benefits before approving funding for any IT project. IT funding

approaches are designed to recover the cost of building and maintaining the

information system in an enterprise (Pearlson and Saunders 2006). The main goal

is to at least cover the cost of the information system.

The variety of IT investment areas from warehouse automation to decision

support systems has created demand for different investment evaluation methods.

IT capital investment decisions can be analyzed by traditional and IT specific

investment evaluation methods.

Organizations often use traditional (non-IT specific) methods like Net Present

Value (NPV), Return on Investment (ROI), and Activity Based Costing (ABC) for IT

investment evaluation (Turban, et al. 2006, 561-564). Nevertheless, some

researchers argue that traditional evaluation techniques are not suitable for

evaluating projects with significant strategic benefits such as BI projects (Irani and

Love 2001).

IT projects in many cases generate intangible benefits such as faster time to

market, employee and customer satisfaction, grater organizational agility, and

improved control (Turban, et al. 2006, 561-564). Ignoring value of intangible

benefits may lead the organization to reject IT investments. Therefore, financial

analyses need to consider tangible and intangible benefits.

Page 14: Business Intelligence for Small Enterprises: An Open Source Approach

8

The most straightforward method for evaluation of intangible benefits is to make

rough estimates of monetary values for all intangible benefits. However, putting

monetary value on the IT investment is not an easy task (Turban, et al. 2006, 561-

564).

Additionally, in the past years the focus of industries has shifted from cost

reduction alone to maximizing both: IT benefits and business benefits (Shields and

Bharucha 2003). Growing demand for benefit measurement tools and complexity

of calculation methods has resulted in a wide range of IT specific investment

evaluation frameworks. Some of these specific frameworks and their summaries

are shown in table 2.2.

Model Source Model Summary

Economic Added Value (EVA) Stern & Stewart

Accurate measure of post-tax return. Uses retrospective analysis which is not suitable for prior estimation of returns in IT.

Total Value of Ownership (TVO) Gartner Comprehensive view of costs and benefits.

Total Economic Impact (TEI) Giga Comprehensive view of costs, benefits and risks.

Rapid Economic Justification Microsoft Cost-benefit analysis, multiple stakeholder view and risk assessment.

Business Value Index Intel Cost-benefit analysis, multiple stakeholder view and risk assessment.

Table 2.2. Comparison between existing benefit measurement models, adopted from (Shields and Bharucha 2003)

Page 15: Business Intelligence for Small Enterprises: An Open Source Approach

9

3 Valuing BI Systems

In order to compare different BI suites, comparison grounds should be defined

first. In this chapter, criteria for comparison have been defined using a divisive

(“top-down”) approach. We start from the most important, top business criterion –

the value of the BI solution brought to the enterprise. Then we continue by

dividing business value into smaller criteria categories until we reach the level

where it is possible to compare BI systems. We will consider this level reached

when it is possible to measure criteria.

Executives are constantly evaluating the cost versus the benefit of different

business decisions. A typical question is: “Which initiative will yield the greatest

benefit to the organization?” Organisations often use cost-benefit analysis

approaches which compare the total value of the benefits with the associated costs

(Turban, et al. 2006, 561).

Figure 3.1. Calculating Net Benefits, adopted from PENG model (Dahlgren 1997)

The effect on profit (net benefit) consists of gross benefits minus costs of

Information System (Dahlgren 1997) (figure 3.1). Thus, in order to calculate net

benefits and prove its profitability we need first to calculate costs and gross

benefits of a BI system. In the next two sections we are choosing comparison

grounds for costs and total benefits.

3.1 Identifying BI Costs

The most basic method for calculating the cost is to add up the costs of all the

components. Many companies calculate initial and ongoing maintenance costs in

this way (Pearlson and Saunders 2006, 256).

In our case, summing up the initial and maintenance costs does not provide an

entirely accurate total cost for BI initiatives since it includes other costs such as

user trainings which span all over a BI lifecycle. Therefore, a method known as

Total Cost of Ownership (TCO) is used. This technique was introduced by the

Gross Benefits

Costs

Net Benefit

Net Benefit = Gross Benefits - Costs

Page 16: Business Intelligence for Small Enterprises: An Open Source Approach

10

Gartner Group in late 1980s and used to calculate more accurate IS costs. Besides

initial and maintenance costs, TCO includes costs associated with technical

support, administration, and training (Pearlson and Saunders 2006, 256).

Cost components of IT projects can be classified into three categories: acquisition

cost, operations cost, and control cost (Turban, et al. 2006, 567). Since we evaluate

integrated BI suites, control costs such as consolidation and standardisation are

not applicable in our case and excluded from further cost calculations. Based on

literature study (Turban, et al. 2006, 567) typical cost components of BI

implementation project are reflected in figure 3.2.

The TCO method described above aimed to calculate overall cost of BI system for a

defined period of exploitation as a part of overall benefit calculation. These costs

are used as grounds for further comparison of BI suites.

Figure 3.2. Typical costs for BI implementation project

3.2 Identifying BI Benefits through Decomposition

While it is easy to quantify all BI costs, benefits are much more complex and

difficult to quantify and calculate. Soft benefits, such as the ability to make future

decisions, are making it difficult to measure the payback of the investment. The BI

initiative is a long term strategic information system investment where it is hard to

evaluate all benefits brought to the enterprise. Below we try to show most obvious,

well known and general benefits of BI. Nevertheless, and important, a detailed

analysis in relation to the specific industry might identify more benefits.

BI Costs

Acquisition Costs

Software licenses

RDBMS or other backend system licenses

Installation and configuration

Hardware and other infrastructure

Operations Costs

Maintenance

Support

Upgrades

User Trainings

Page 17: Business Intelligence for Small Enterprises: An Open Source Approach

11

Better identification and understanding of BI benefits can be reached through

categorisation. Knightsbridge (2005) proposes to simplify the process of

identifying benefits by separately considering the two primary categories of

benefits: revenue enhancements and cost savings. Cost savings are defined as the

difference in the costs associated with the new BI initiative versus the costs

associated with maintaining the existing information environment. Revenue

enhancements are defined as the beneficial activities that result from decisions

individuals make by using information from the BI solution.

Steve and Nancy Williams (2003) identify two another dimensions of BI benefits in

their article: “the business value of BI lies in its use within management processes

that impact operational processes that drive revenue or reduce costs, and/or in its

use within those operational processes themselves.”

Figure 3.3. Quadrant graph for generic BI benefits classification

These four dimensions of BI benefits (i.e. Revenue Enhancements, Cost Savings,

Management Process Improvement, and Operating Process Improvement) are

combined together in the quadrant graph on figure 3.3 in order to classify and find

the tangibility level of the common BI benefits as identified by Moss and Atre

(2003). The top-left (red) square on the picture identifies the BI solution’s impact

on revenue enhancement through management process improvement which is

Identification of new business opportunities and markets Improved opportunity recognition Improved time to market Improved decision making Increased organisational agility Identification of under-performing product lines or products

Enhanced speed of new product development Increased ability to content with competitors Improved customer service/satisfaction

Reduced operating costs Automation of manual processes Improved operating processes

Improved information dissemination Improved analysis (e.g. reduced customised reporting requests) Informed decision making

Tangible Intangible

Inta

ngib

le

Management Process Improvement Operating Process Improvement

Reven

ue

En

han

cem

en

ts

Co

st

Sa

vin

g

Page 18: Business Intelligence for Small Enterprises: An Open Source Approach

12

difficult to evaluate because there is usually no direct attribution. For instance,

monetary value of the benefit “identification of new business opportunities and

markets” cannot be estimated at all since we don’t know in advance whether new

markets exist or not. The bottom-right (green) square contains benefits of

operational process improvement associated with cost savings. In contrast to “red”

benefits, these benefits are easy to identify because of the ability to compare the

new BI solution to the old operational environment. Finally, bottom-left and top-

right (yellow) boxes contain intangible benefits which could be estimated through

complicated evaluation methods (e.g. improved customer satisfaction).

It could be clearly seen from figure 3.3 that 12 of 15 benefits are either completely

intangible (red square) or could only be calculated through complex evaluation

methods (yellow squares). This classification method can be used for tangibility

level identification of more specific benefits by placing them in one of the four

squares in figure 3.3.

As already mentioned, BI is a strategic tool and its primary role is the support of

business strategy and business processes. As can be seen from figure 3.3, the BI

initiative does not contribute as much in terms of being used as a process

automation (operational) tool. The four dimensions of figure 3.3 could also be used

to claim that instead of evaluating BI separately, it should be analyzed together

with the business processes it is supposed to enhance. To additionally support that

claim we would like to refer to authors Luftman and Muller (2005), who also

conclude that the benefits come not from the latest and greatest IT solution, but

from how a business modifies its practices based on its new technology. In spite of

that, BI applications still provide some level of pure operational functionality like

KPI (Key Performance Indicators) monitoring, automatic reporting and invoicing.

A number of benefit measurement models has been discussed in this section as

well as in section 2.2. One observation regarding the various ways of assessing BI

benefits discussed so far is that reaching a precise valuation may take longer than

is reasonable to make an investment decision. This is due to the fact that most of

them are complicated and require assigning monetary value to the intangible

benefits. Therefore, in this thesis we try to avoid these comprehensive BI benefit

calculations based on assigning of monetary values. Instead, in the next section, we

explore an alternative approach where a number of criteria influencing benefits

brought by BI solution are defined, criteria that are not dependent on monetary

value assessment per se.

3.3 Identifying BI Benefits through its Characteristics

Yielded benefits are directly related to the characteristics of the BI suite. In this

section we describe alternative approach of BI benefits identification through the

influencing characteristics. This approach suits our research method (section 1.6)

Page 19: Business Intelligence for Small Enterprises: An Open Source Approach

13

since it allows comparing of different BI suites’ benefits without assigning

monetary values.

Rajeev Rawat’s (2007) criteria for assessing Open Source BI alternative has been

adopted to define the characteristics influencing gross value. In our research some

irrelevant criteria have been eliminated, since the original list contains criteria for

complete evaluation of BI project (including business needs, risk assessment, etc.)

while we only looking for BI suite comparison grounds. Rawat’s original list of

criteria for Open Source BI alternative assessment can be found in Appendix A.

Three key criteria categories shown below have been extracted from Rawat’s list of

criteria for further BI suite comparison:

1. BI Functionalities

2. Solution Maturity

3. Price

It could be clearly seen that first and second criteria categories address benefits of

BI system, while third category – price overlaps with the costs which has been

discussed in section 3.1. Identified criteria categories are hierarchically brought

together in figure 3.4.

In section 3.1 acquisition costs and operation costs are already been decomposed

into the smaller criteria which could be defined and compared. BI functionalities

and solution maturity criteria categories will be decomposed in the next sections.

3.3.1 BI Functionalities As it has been mentioned in chapter 1, BI is an umbrella term covering a set of

enterprise applications. Different commercial vendors have different

understanding of BI and what it should include. Nevertheless, there are some key

components which exist in all commercial distributions.

Value (Net Benefit)

Gross Benefits

BI Functionalities

Solution Maturity

Costs

Acquisition Costs

Operation Costs

Figure 3.4. Criteria categories for BI suite comparison

Page 20: Business Intelligence for Small Enterprises: An Open Source Approach

14

BI Functionalities

Infrastructure Integration All tools in the platform should use the same security, metadata, administration, portal integration, object model and query engine, and should share the same look and feel.

Reporting Reporting provides the ability to create formatted reports in different formats (e.g. PDF, Microsoft Excel, etc.)

Dashboard Dashboard includes the ability to publish formal, web-based reports with intuitive displays of information, including dials, gauges and traffic lights. These displays indicate the state of the performance metric, compared with a goal or target value.

Ad Hoc Query and Reporting

Also known as self-service reporting, this capability enables users to ask their own questions of the data, without relying on IT to create a report. These functions imply semantic layer which operates between business users and data sources allowing users to navigate without understanding underlying data structure.

Spreadsheet Services In some cases, BI platforms are used as a middle tier to manage and execute BI tasks, but office tools (particularly Microsoft Office Excel) acts as the BI client. In these cases, it is vital that the BI vendor provides integration with office suites.

OLAP Server Enables end users to analyze data with extremely fast query and calculation performance, enabling a style of analysis known as "slicing and dicing".

OLAP UI Front-end tools for the analysis known as "slicing and dicing".

Data Mining This capability enables organizations to classify categorical variables and to estimate continuous variables using advanced mathematical techniques.

ETL ETL stands for extract, transform, and load. ETL is a process that enables businesses to consolidate their data from different sources and in different formats.

Alerts Event based alerts sent by email, SMS, etc. Events usually triggered by business rules.

Repository The repository defines the functions and services to store BI structured data and metadata (e.g. report templates, business rules, and semantic layer data).

Security This enables administrators to define different user roles and permissions based on business needs.

Scheduling Ability to schedule report generation and other actions.

Table 3.1. BI Functionalities criteria group for BI evaluation, adopted from (Rawat 2007)

The list of functionalities, summarised in table 3.1, is adopted from Rajeev Rawat’s

(2007) “BI function requirements”. Some functionality such as data mining and

spreadsheet services have been added based on the literature study (Gartner,

Magic Quadrant for BI Platforms, 1Q07 2007). In section 4.1 we will provide an

analysis of major commercial BI vendors to support our claims about general

acceptability of these BI suite functionalities.

It is important to note that most of these functionalities could be decomposed

further. For instance, reporting may include a number of different output formats

and chart types. Since the number of these details is more than one hundred

(Howson 2007) we will remain at this level.

Page 21: Business Intelligence for Small Enterprises: An Open Source Approach

15

3.3.2 BI Solution Maturity Maturity is another important criterion, especially from open source software

perspective. In order to make it definable and measurable it has been broken down

into sub-criteria (e.g. measurable criteria such as duration of the product

availability in the market and number of customers actively using product can

show how mature this product is).

In addition to the standard maturity criteria defined in Rawat’s (2007) list, we

have added open source related criteria such as OS community size and activity.

Community importance as a driving force of an open source projects described in

section 2.1. Another open source related attribute – number of downloads

indirectly shows popularity of the product. The resulting list of solution maturity

criteria is shown in table 3.2.

Solution Maturity

Availability duration Indicates duration of availability in the market.

Customer base Indicates number of customers using complete or partial BI solution

Standards compatibility Defines whether BI suite follows standards or use proprietary solutions.

Ready-to-use out of the box Items, functionalities, or features provided do not require any additional installations, plug-ins, expansion packs, or products.

ISV-ready: embeddable & extensible Can be embedded, adopted or extended for specific needs.

Localised and internationalised Translations and localisations are available.

Training availability Customers can order trainings provided by vendor.

Open Source Community Size Number of registered open source community members.

Open Source Community Activity Number of forum and mailing list posts made by open source community members.

Downloads (Open Source only) Number of product downloads (indirectly shows popularity).

Table 3.2. Solution maturity criteria group for BI evaluation, adopted from (Rawat 2007)

Page 22: Business Intelligence for Small Enterprises: An Open Source Approach

16

4 Results and Findings

In the next two sections (4.1 and 4.2) benefits of commercial and open source BI

suites estimated and compared based on the criteria defined in sections 3.3.

Sections 4.3 and 4.4 mainly explore costs of BI implementation through the two

different research methods (survey and case study) in order to increase results’

accuracy. Further analysis of these findings will help us to answer research

question stated in section 1.5 (section 5).

4.1 Commercial BI Suites

According to the Gartner’s Magic Quadrant (2007), the leading BI software vendors

for 1Q 2007 were Business Objects, SAS, Cognos, Oracle, and Hyperion. With recent

acquisitions of Hyperion, Business Objects, and Cognos by Oracle, SAP and IBM

(respectively) the number of main software vendors has become even smaller.

Thus, in the analyses we have included the following commercial BI products:

1. Business Objects XI from SAP

2. Oracle BI Suite from Oracle (including Hyperion)

3. SAS BI from SAS Institute

4. Cognos BI from IBM

Published information about the leading commercial BI products and their formal

system documentation were consulted in order to support the claims of generality

of the accepted BI suite functionalities defined in table 3.1. We found that all

claimed functionalities are provided by reviewed commercial BI suites. The only

exception was the BusinessObjects XI which relies on 3rd party tools in data mining

and OLAP server. Summary of these analyses based on the criteria defined in

section 3.3 can be found in table 4.1.

It is important to note that the portfolio of these major vendors usually stretches

beyond BI and covers advanced business applications such as Business

Performance Solutions (BPS) which are not included in our list of functionalities.

Page 23: Business Intelligence for Small Enterprises: An Open Source Approach

17

BI Functionalities* Bu

sin

ess

Ob

jects

XI

Ora

cle

BI

SA

S B

I

Co

gn

os

BI

Infrastructure Integration Yes Yes Yes Yes

Reporting Yes Yes Yes Yes

Dashboard Yes Yes Yes Yes

Ad Hoc Query and Reporting Yes Yes Yes Yes

Spreadsheet Services Yes Yes Yes Yes

OLAP Server No Yes Yes Yes

OLAP UI Yes Yes Yes Yes

Data Mining No Yes Yes Yes

ETL Yes Yes Yes Yes

Alerts Yes Yes Yes Yes

Repository Yes Yes Yes Yes

Security Yes Yes Yes Yes

Scheduling Yes Yes Yes Yes

Solution Maturity

Availability duration (years) >10 >10 >10 >10

Customer base * 43.000 n/a 43.000 23.000

Standards compatibility ** 6 5 8 6

Ready-to-use out of the box Yes Yes Yes Yes

ISV-ready: embeddable & extensible Yes Yes Yes Yes

Localised and internationalised Yes Yes Yes Yes

Training availability Yes Yes Yes Yes

* Information is taken from the vendor’s official web site ** Scores based on a scale of 0 (weak) to 10 (strong) and adopted from (Datamonitor 2007)

Table 4.1. Comparison results of the commercial BI suites

4.2 Open Source BI Suites

Our study of the available OS BI suites on the market has identified four major

players based on their popularity (community size and activity):

1. JasperIntelligence from JasperSoft

2. Pentaho BI from Pentaho

3. OpenI from Loyalty Matrix

4. SpagoBI from Engineering Ingegneria Informatica

Comparison results of these four open source products based on the criteria

defined in section 3.3 is represented in table 4.2. In order to be considered as a

comprehensive alternative to a commercial BI, the OS BI suite should provide all

functionalities previously defined. We chose Pentaho for further comparison since

Page 24: Business Intelligence for Small Enterprises: An Open Source Approach

18

only Pentaho provides all defined functionalities as well as the most active

community and a high number of downloads. It should be noted that JasperSoft’s

high number of community members and downloads can be explained respectively

by mandatory registration (as community member) for all downloads and greater

number of downloadable items.

BI Functionalities* Ja

sp

er

Inte

llig

en

ce

Pe

nta

ho

BI

Op

en

I

Sp

ag

oB

I

Infrastructure Integration Yes Yes Yes No

Reporting Yes Yes Yes Yes

Dashboard No Yes No Yes

Ad Hoc Query and Reporting Limited Yes No No

Spreadsheet Services No Yes No Yes

OLAP Server Yes Yes Yes Yes

OLAP UI Yes Yes Yes Yes

Data Mining No Yes Yes Yes

ETL Yes Yes Yes Yes

Alerts Yes Yes No No

Repository Yes Yes No Yes

Security Yes Yes Yes Yes

Scheduling Yes Yes No Yes

Solution Maturity*

Availability duration (months) 30 28 27 28

Customer base 6.000 n/a n/a n/a

Standards compatibility Partial Yes Yes Yes

Ready-to-use out of the box Yes Yes Yes Yes

ISV-ready: embeddable & extensible Yes Yes Yes Yes

Localized & internationalized Yes Yes No Yes

Training availability Yes Yes No Yes

Open Source Community Size 57.700 18.800 n/a n/a

Open Source Community Activity 32.300 53.800 1200 1200

Downloads 2.468.717 >2.000.000 19.926 n/a

* Information is taken from the vendor’s official web site

Table 4.2. Comparison results of OS BI suites

Page 25: Business Intelligence for Small Enterprises: An Open Source Approach

19

4.3 Interview and Survey of BI Consulting Companies

In order to provide an answer to the second research question (whether OS BI

solution provides cost saving) we have conducted a number of interviews and

web-based surveys. Interviews were conducted based on the questionnaire from

Appendix B in order to better understanding of the real situation in the market.

Unfortunately, very few companies accepted request for interview, mainly due to

lack of time of the potential interviewee. In order to increase accuracy of the

results, more companies have been approached through the simplified web-based

survey form where we included only questions 1–4 (questions aimed to

understand attitude of consulting companies regarding small enterprises) from the

same questionnaire (Appendix B). Overall, two consulting companies providing BI

services have been interviewed in person and four other companies responded to

the web survey. Complete results of interviews and surveys are given in Appendix

C.

Figure 4.1. Enterprise types mainly targeted by the BI consulting companies

Figure 4.2. Customers of the BI consulting companies

0%

50%

100%

0%

20%

40%

60%

80%

100%

Small Medium Large

Enterprise Types

Q1: What types of enterprises are your company’s main target group?

Small3,3%

Medium14,7%

Large82,0%

Q2 & Q3: What types of enterprises are your company’s customers?

Page 26: Business Intelligence for Small Enterprises: An Open Source Approach

20

Web survey and interview results for the questions 1-3 (Appendix B) has been

summarised in figures 4.1 and 4.2. It can be clearly seen that none of the

respondents consider small enterprises as a main target group (figure 4.1), though

small enterprises constitute 3.3% of their customers (figure 4.2). According to the

interviewed consultants, the majority of the small and medium enterprises

implementing BI (figure 4.2) expect fast growth in the next 5-10 years and

consider BI investments as a part of their long-term vision.

Questions number five and six (Appendix B) was mainly aimed to support our list

of functionalities explored in section 3.3.1. It can be seen from figure 4.3, where

results for these questions depicted, that the major part of proposed functionalities

were considered as a part of BI while no additional functionalities were proposed

by interviewees. The only exception was alerts which track significant events (e.g.

profit target exceed). According to the one of the consultants, largely used nightly

data updates make alerting functionality ineffective; however alerts would be

much appreciated in the operational (real-time) BI systems.

Figure 4.3. Complete BI solution functionalities.

Finally, interviewees were asked (questions no. 7-8, Appendix B) to estimate the

cost components defined in section 3.1 for the experimental setup case which has

been described in Appendix CAppendix D. Gathered costs have been averaged and

reflected in figure 4.4. This is important to note that all interviewees offered

Microsoft’s BI products (PerformancePoint 2007 and SQL Server 2005) as most

suitable commercial products for this case (question no. 9, Appendix B).

0%

25%

50%

75%

100%Reporting

Dashboard

OLAP Server

OLAP UI (Slicing & Dicing)

ETL

Spreadsheet Services

Ad Hoc Query and Reporting

Data Mining

Alerts

Scheduling

Q5 & Q6: Which BI functionalities exist in your complete BI solutions?

Page 27: Business Intelligence for Small Enterprises: An Open Source Approach

21

Nevertheless, interviewees mentioned another important fact that other vendors

such as BusinessObjects or Oracle would have a notably more expensive solution.

Figure 4.4. Average costs of the commercial BI suite implementation

4.4 Experimental Setup

An experimental OS BI system has been set up in the company with the purpose of

increasing accuracy of the results and providing additional investigation for the

questions stated in section 1.5. In order to get authentic results we were looking

for a company producing a lot of raw operational data and demanding

transformation of these data into the information. Telecom is one of the industries

where every action of the customer produces noticeable amount of raw data. At

the same time chosen company should fall into the category of small enterprises.

All said above was the main reason for choosing a mobile content provider

company for an experimental setup. Another important reason was the telecom

background of the author. Further, we will refer to the selected company by

“MobileComet” pseudo name (according to agreement we cannot disclose

company’s real name).

MobileComet is a small enterprise with annual turnover of €500K and 10

employees. Company provides content (i.e. ringtones, pictures, games, etc.) for

200 000 SEK

550 000 SEK

125 000 SEK

300 000 SEK

10 000 SEK

15 000 SEK

BI software, RDBMS, and other licenses

Consulting, installation, and configuration

Hardware and other infrastructure

Maintenance and Support

Upgrades

User Trainings

Cost SEK

Q7 & Q8: Avarage costs of the commercial BI suite implementation over 1 year period

Page 28: Business Intelligence for Small Enterprises: An Open Source Approach

22

mobile phone owners. In order to better understand customers' behaviour and

identify new markets company wants to be able to process operational data which

located in a number of different sources.

Based on the interview with the company representatives we defined following

key requirements for the experimental setup:

Automatic generation of the recurring financial reports and invoices

Ability to build ad-hoc reports and make ad-hoc analysis of the current sales

data and customer behaviour information.

Provide easier, online and centralized access to all business data.

Archive operational data (operational database should only keep data older

than 3 months).

Complete business case description can be found in Appendix D.

4.4.1 System Overview

Pentaho OS BI platform has been chosen for experimental setup based on the

results from section 4.2. Data flow and components of the implemented BI system

depicted on figure 4.5.

The hardware platform used to develop and run Pentaho BI was Intel Core2Duo

workstation with 2GB of RAM and running Linux operating system. Data

warehouse has been built on MySQL RDBMS. Note that both, Linux and MySQL are

open source products.

Operational DB

(Internal Data)

Mobile Operators

(External Data)

Excel Sheets

(Personal Data)

ETL

(Pentaho Data

Integration)

Data Warehouse

(MySQL backend)

Data AnalysisData StorageData Sources

OLAP

(Pentaho

Analysis)

Metadata

(Pentaho

Metadata Editor) Ad-hoc

Reporting

Ad-hoc Query

Dashboard

Reporting and

Charting

Data Visualisation

Figure 4.5. Data flow and components of the experimental system

As shown on figure 4.5, Pentaho Data Integration (DI) is the very first in the chain

of BI components performing extracting, transforming, and loading (ETL) from

different data sources into single data warehouse nightly. Sample Pentaho DI

transformation and job are shown on figure 4.6. Construction of this component

(ETL) consumed 75% of the total work amount.

Page 29: Business Intelligence for Small Enterprises: An Open Source Approach

23

Next step was description of data warehouse structure through Pentaho Metadata

Editor and publishing this metadata to the Pentaho BI server. As soon as metadata

information published, end users were able to run self-service (ad-hoc) reports.

Finally, Pentaho Report Designer was used to create a number of sample reports

and invoices with charts and tables. Some reports have been scheduled for

periodic generation and distribution by email.

Figure 4.6. Sample Transformation and Job in Pentaho Data Integration

Although we tried to utilise all major functionalities of the Pentaho BI platform

defined in table 4.2, there are still a number of the components (Dashboard, Data

Mining) which we were unable to test due to time constraints. However,

MobileComet found those major components satisfactory for their current needs.

Page 30: Business Intelligence for Small Enterprises: An Open Source Approach

24

4.4.2 Setup Results

Experimental setup results are discussed below in the form of answers to the

research questions stated in section 1.5.

RQ1: Does an OS BI provide enough business value for small enterprises to be

considered as an alternative to the commercial BI suites?

All business needs described in the mini case (Appendix D) have been met.

Customer found results and functionalities of the BI software completely

satisfactory.

RQ2: Does an OS BI solution provide cost savings in comparison with commercial BI

suites?

We will answer this question indirectly by providing costs for the experimental

setup. In the next sections these costs are compared to the commercial BI costs in

order to answer this research question. Based on the setup results, acquisition

costs have been estimated and reflected in table 4.3.

Acquisition Cost Type Estimated Cost

BI, RDBMS, and other software licenses 0 SEK

Consulting, installation, and configuration equivalent of 160 work hours

Hardware and other infrastructure 40.000 SEK

Table 4.3. Acquisition costs for experimental setup

Professional services (consulting, installation, and configuration) have been

provided for free, as a part of this research. Due to this reason monetary value of

these services could not be estimated directly. However, it can be calculated by

considering cost of the resources (internal staff member or outsource consultant).

For instance, monetary equivalent of 160 hours for MobileComet’s employee could

vary between 30.000-60.000 SEK.

RQ3: Does an OS BI actually fit into small businesses’ ICT budget?

According to the estimations above, acquisition cost of the experimental system

was less than 100.000 SEK. CEO of MobileComet found this figure very attractive

and mentioned that he would allocate this amount from the budget for the similar

basic BI solution.

Page 31: Business Intelligence for Small Enterprises: An Open Source Approach

25

5 Analysis and Discussion

5.1 Comparison of Benefits

In this section we compare commercial and OS BI suites based on the criteria

forming BI benefits which have been defined and explored in section 3.3. In case of

a very close match between these criteria we will consider that an enterprise will

gain similar gross benefits from different BI suites.

For comparison simplicity and due to very close criteria match between different

commercial BI suites we generalise them as a “Commercial BI”. Data for

Commercial BI and Pentaho OS BI have been taken from sections 4.1 and 4.2

respectively. Side by side comparison of the criteria has been provided in table 5.1.

BI Functionalities Commercial BI Pentaho OS BI

Infrastructure Integration Yes Yes

Reporting Yes Yes

Dashboard Yes Yes

Ad Hoc Query and Reporting Yes Yes

Spreadsheet Services Yes Yes

OLAP Server Yes Yes

OLAP UI Yes Yes

Data Mining Yes Yes

ETL Yes Yes

Alerts Yes Yes

Repository Yes Yes

Security Yes Yes

Scheduling Yes Yes

Solution Maturity

Availability duration (years) >10 2.3

Customer base >23.000 n/a

Standards compatibility Partial Yes

Ready-to-use out of the box Yes Yes

ISV-ready: embeddable & extensible Yes Yes

Localised and internationalised Yes Yes

Training availability Yes Yes

Table 5.1. Comparison of benefit components for the commercial and OS BI

It can be seen from table 5.1 that both solutions cover the whole range of

functionalities. It should be noted that functionalities of the commercial BI suites

are usually more advanced. Nevertheless, according to the experimental setup

Page 32: Business Intelligence for Small Enterprises: An Open Source Approach

26

results (section 4.4.2) OS BI functionalities are advanced enough to meet basic

needs of small companies.

Analysis of the solution maturity characteristics has shown overweight on the

Commercial BI’s side mainly because of the large number of customers and longer

period of solution availability. Nonetheless, OS BI has a single but very important

advantage over commercial solution – higher standard compatibility. OS BI

solutions, and particularly Pentaho, built upon open standards such as BPEL

(Business Process Execution Language) for process orchestration, and CWM

(Common Warehouse Metamodel) for data warehouse metadata. Thus, customers

can avoid "vendor lock-in" of proprietary BI software vendors.

It is necessary to mention that standards used by OS BI (e.g. CWM, BPEL, etc.) are

very new in comparison with the BI industry. This fact could be one of the major

reasons why commercial BI vendors poorly support them.

To sum up, benefit comparison analysis of the Commercial BI and Pentaho OS BI

has shown that OS suite provides complete set of basic BI functionalities while

commercial solutions are far more mature. From this we conclude that OS BI will

meet basic needs of the small enterprises and yield gross benefits nearly similar to

commercial BI. Thus, in accordance with the equation shown in figure 3.1 (net

benefit = gross benefits – costs), net benefit difference between commercial and OS

BI will depend only on the costs of the BI implementation.

5.2 Comparison of Costs

Costs for the commercial BI solution have been estimated in section 4.3 based on

the interview results. Unfortunately, we were unable to interview any consulting

company offering OS BI since there are very few such companies worldwide.

To estimate OS BI costs, we will consider the case where Pentaho BI

implementation would be done by the same consulting company. In that case,

hardware costs, total work amount, and fees for the professional services (i.e.

consulting, installation, etc.) will remain the same. Apparently, there will be no

license fees for the software since we would use open source RDBMS (e.g. MySQL)

and operating system (e.g. Linux). Costs for the both cases (commercial and OS)

have been put together in table 5.2.

In addition to the costs for outsourced solutions we can calculate costs of

insourced OS BI solution based on the experimental setup results from section

4.4.2. Moreover, in order to better understand costs we will calculate TCO of all BI

solutions for a five-year period.

Page 33: Business Intelligence for Small Enterprises: An Open Source Approach

27

Cost Type Estimated Cost of Outsourced Commercial BI

Estimated Cost of Outsourced Pentaho OS BI

Estimated Cost of Insourced Pentaho OS BI

Acquisition Costs

BI software, RDBMS or other system licenses 200.000 SEK 0 SEK 0 SEK

Consulting, installation, and configuration 550.000 SEK 550.000 SEK 60.000 SEK

Hardware and other infrastructure 125.000 SEK 125.000 SEK 40.000 SEK

Operations Costs

Maintenance and Support 300.000 SEK/year 300.000 SEK/year 270.000 SEK/year

Upgrades 10.000 SEK/year 0 SEK/year 0 SEK/year

User Trainings 15.000 SEK/year 15.000 SEK/year 0 SEK/year

TCO, 1 year period 1.190.000 SEK 990.000 SEK 370.000 SEK

TCO, 5 year period 2.500.000 SEK 2.250.000 SEK 1.450.000 SEK

Table 5.2. Comparison of cost components for the commercial and OS BI

This is important to emphasise the great cost of the professional services (i.e.

consulting, installation, etc.) and support which together form 91.11% of the OS

BI’s TCO over a five-year period. Such a high costs for professional services could

be explained by results of the survey which have shown that none of the

responded BI consulting companies see small enterprises as their main customer

group (figure 4.1) and prices for their services mainly set for medium and large

enterprises.

Clearly, small enterprises will save grate amount of the investment only by

reducing cost of professional services. Experimental setup results suggest that

utilising internal resources can save up to 80% on the professional services

(consulting company estimated professional services for mini case 550.000 SEK

(table 5.2), while experimental setup has shown roughly 30.000–60.000 SEK

(section 4.4.2) cost for the same item). In that case, company’s IT department may

be hired for installation, configuration and support of the BI software while

business analysis and consulting services outsourced.

However, projecting costs based on experimental setup case for the longer time

period will unveil grater TCO of insourced OS BI. Experimental setup case

proposed (section 4.4.2) usage of internal resources in order to minimise initial

setup expenses. Based on that assumption we can calculate ongoing maintenance

and support costs. According to experimental setup results, average monthly

salary of internal staff member estimated at 45.000 SEK. BI solutions require daily

maintenance to ensure that nightly ELT processes successfully completed and

reports are prepared. Moreover, it requires regular updates of data sources, ETL

processes and design of new reports. All these tasks may require up to 50%

utilization of the single internal resource which will be equivalent to 270.000 SEK

annually (45.000 SEK × 0.5 × 12 months). Thus, in comparison to consulting

companies, utilization of internal resources will save roughly 10% on support and

maintenance (300.000 SEK vs. 270.000 SEK).

Page 34: Business Intelligence for Small Enterprises: An Open Source Approach

28

Figure 5.1. TCO of the commercial and OS BI over a five-year period

It can be seen from table 5.2 that cost saving brought by outsourced OS BI over the

first year will be 16.80% of the commercial BI’s TCO. If we project these expenses

over the 5 year period (figure 5.1), we will find that cost saving gets more and

more insignificant in the light of growing maintenance and support costs. Thus, as

it is shown in figure 5.1, cost saving for 5 year period will represent only 10% of

the commercial BI’s TCO.

Such insignificant difference between commercial and OS BI implementation costs

can be explained by the nature of the particular commercial solution. As it was

mentioned in section 4.3, all estimated commercial solutions are based on the BI

software from Microsoft. In turn, Microsoft recently announced (Microsoft 2006)

products which are aimed to bring “BI for the masses” and make it affordable for

small and medium enterprises. Nonetheless, according to the interview results

from section 4.3 figures for the commercial solution would be notably higher if it

was solution from BusinessObjects or Oracle.

Another factor influencing diversity between OS and commercial BI TCO is the

number of the end-users. In accordance with the mini case requirements

(Appendix CAppendix D) license fees estimated in table 5.2 allow up to 5 end-user

connections. As company grows, cost of commercial software licenses may

1 190 000

1 525 000

1 850 000

2 175 000

2 500 000

990 000

1 305 000

1 620 000

1 935 000

2 250 000

370 000

640 000

910 000

1 180 000

1 450 000

0

500000

1000000

1500000

2000000

2500000

3000000

Year 1 Year 2 Year 3 Year 4 Year 5

TCO of BI suites over a five-year period

Commercial BI (outsourced) Open Source BI (outsourced) Open Source BI (insourced)

Page 35: Business Intelligence for Small Enterprises: An Open Source Approach

29

increase many-fold and drastically broaden the gap between OS and commercial BI

TCO.

Nevertheless, an open source solution has important advantages over commercial

BI listed below:

Availability of complete source code

Built upon open standards (reduce proprietary vendor lock-in risk)

Access to latest updates and patches

Active community (forums, mailing lists, etc.)

Finally, we would like to highlight difference of hardware costs between

experimental setup (40.000 SEK) and interviewees’ costs (125.000 SEK). While

only one server was used in experimental setup, all interviewees proposed

solution with two servers – production and development (consulting company

may require provisioning of two separate servers). In contrast to latter solution

which is typical for large enterprises, small enterprises can gain from virtualisation

techniques (running two separate operating systems on a single server).

Page 36: Business Intelligence for Small Enterprises: An Open Source Approach

30

6 Conclusion

In order to find out whether an open source BI suites can facilitate for small

enterprises to remain competitive, three research questions were stated in section

1.5:

1. Does an OS BI provide enough business value for small enterprises to be

considered as an alternative to the commercial BI suites?

2. Does an OS BI solution provide cost savings in comparison with commercial

BI suites?

3. Does an OS BI actually fit into small businesses’ ICT budget?

To answer these questions business value evaluation through cost-benefit analysis

approach were chosen in section 3. Using divisive ("top-down") approach we tried

to estimate monetary value for both – costs and benefits of different BI suites.

While costs have been easily decomposed, decomposition of the benefits wasn’t

successful. Using quadrant graph for BI benefit classification (figure 3.3) we found

that majority of BI benefits are either completely intangible or could be calculated

only through complex evaluation methods because of strategic nature of BI. Thus,

in section 3.3, we explored an alternative approach for benefit estimation where

criteria influencing benefits are defined. Four commercial and four OS BI suites

have been evaluated against defined criteria in sections 4.1 and 4.2 respectively.

Results have shown that Pentaho is the most comprehensive and mature BI among

OS suites. Therefore, Pentaho has been chosen for further comparison with

commercial products.

Two different approaches have been chosen in order to estimate the costs for the

same BI implementation case and gather supportive information:

1. Interviews and web-based surveys of consulting companies helped to

capture costs of outsourced solution and answer additional questions

stated in questionnaire (Appendix B).

2. Experimental integration was set up in order to measure functionality of OS

BI and capture the costs of insourced solution.

Results of the research summarised and grouped below according to the research

questions stated earlier.

Q1: Does an OS BI provide enough business value for small enterprises to be

considered as an alternative to the commercial BI suites?

Costs and benefits of the commercial and OS BI implementations have been

compared based on the interviewees’ (BI consulting companies) estimations and

experimental setup results. Experimental setup of OS BI and benefit comparison

between the commercial BI and OS BI have shown that OS BI meets basic needs of

the small businesses and yields gross benefits nearly similar to commercial BI.

Page 37: Business Intelligence for Small Enterprises: An Open Source Approach

31

Nevertheless, the greatest value of OS BI lays in minimizing risk of proprietary

vendor lock-in and maximizing flexibility of BI solution through the support of

open standards and source code availability.

Q2: Does an OS BI solution provide cost savings in comparison with commercial BI

suites?

Cost saving provided by OS BI solution formed only 10% of the commercial BI’s

TCO over a five-year period (2.500K SEK versus 2.250K SEK). Such insignificant

cost difference between commercial and OS BI solutions can be explained by the

number of factors:

a. Estimated commercial solutions are based on the BI software from the

Microsoft which recently announced products (PerformancePoint 2007)

which are aimed to bring “BI for the masses” and make it affordable for

SMEs. This explains low costs of the commercial solution.

b. Estimated license fees of the commercial solution allow only up to 5 end-

user connections. As company grows, cost of commercial software licenses

may increase many-fold and drastically broaden the gap between OS and

commercial BI TCO.

On the other hand, experimental setup suggests that insourced OS BI

implementation can provide 42% cost saving in comparison to outsourced

commercial BI over a five-year period. Huge variance between these two cases can

be explained by:

a. Survey results which have shown that none of the responded BI consulting

companies see small companies as their main target group (figure 4.1).

Consequently, quality and prices for their professional services mainly set

for medium and large enterprises.

b. Hardware costs. Consulting companies offered two different servers for

production and development systems while for experimental setup we

proposed virtualization (one server running two virtual operating systems).

Nevertheless, insourcing of BI implementation may reduce costs of the commercial

BI implementation as well. Based on the analysis of these results we concluded

that although OS BI does not provide significant cost saving by itself, small

companies may save up to 80% on acquisition costs and 10% on maintenance

costs by using OS BI advantages such as active community (forums, mailing lists,

etc.) and access to the latest updates.

Q3: Does an OS BI actually fit into small businesses’ ICT budget?

Experimental setup results (section 4.4.2) have shown that insourced OS BI

implementation has acquisition costs up to 100.000 SEK which entirely fitted in

the IT budget of the company. Despite the low acquisition costs, operational costs

Page 38: Business Intelligence for Small Enterprises: An Open Source Approach

32

almost as high as for commercial solution. Unfortunately, due to lack of time, we

were able to provide experimental setup only in one small enterprise which

reduces accuracy of this answer.

6.1 Future Research

Although small enterprises form 98% of all enterprises in Europe, not all of them

would benefit from the BI. For instance, small IT budgets, very low BI gross

benefits or simply raw data unavailability (lack of any ICT solution) makes BI

implementation worthless.

Therefore, in depth analysis of small enterprises based on the average figures and

costs from this research may expose small companies’ real demand for BI.

Page 39: Business Intelligence for Small Enterprises: An Open Source Approach

33

7 Bibliography

Corbett, Peter, and Dean Ward. Open Source Reporting Solutions for Institutional Reporting: The BIRT Approach. 2006.

Dahlgren, Lars Erik. Make IT profitable! PENG - A practical tool for financial evaluation of IT benefits. 1997.

Datamonitor. "Decision Matrix: Selecting a Business Intelligence Vendor." SAS.com. April 2007. [On-line] http://www.sas.com/news/analysts/datamonitor_bi_0407.pdf (accessed November 2007).

EC. European Charter for Small Enterprises. European Commission, 2006.

EC. European Conference on Craft and Small Enterprises. European Commission, 2007.

EC. The new SME definition: User guide and model declaration. European Commission, 2003.

Gartner. "Business Intelligence Market Will Grow 10 Percent in EMEA in 2007 According to Gartner." Gartner Corporation Web site. January 30, 2007. [On-line] http://www.gartner.com/it/page.jsp?id=500680 (accessed October 1, 2007).

Gartner. "Magic Quadrant for BI Platforms, 1Q07." Gartner Group Web site. January 26, 2007. [On-line] http://mediaproducts.gartner.com/reprints/oracle/145507.html (accessed October 1, 2007).

Gibson, Marcus, and David Arnott. "BI for Small Business: Assessment, Framework & Agenda." 7th Pacific Asia Conference on Information Systems. 2003. 743-759.

Howson, Cindi. "Blank BIScorecard." BIScorecard.com. October 2007. [On-line] http://www.biscorecard.com/evaluations.asp (accessed October 2007).

Irani, Z., and P. E. D. Love. "The Propagation of Technology Management Taxonomies for Evaluating Investments in Information Systems." Journal of Management Information Systems 17(3), 2001: 161-177.

Knightsbridge. Measuring the value of business intelligence and data warehousing initiatives. Knightsbridge Solutions LLC, 2005.

Koch, Christopher. Open Source - Your Opensource Plan. March 2003. [On-line] http://www.cio.com/article/31768/Open_Source_Your_Opensource_Plan (accessed October 2007).

Lerner, Josh, and Jean Tirole. "The Simple Economics of Open Source." 2000.

Luftman, Jerry, and Hunter Muller. "Total Value Of Ownership: A New Model." Optimize Magazine, Issue 22, 2005.

Moss, Larissa T., and Shaku Atre. Business Intelligence Roadmap: The Complete Project Lifecycle for Decision-Support Applications. 2003.

Page 40: Business Intelligence for Small Enterprises: An Open Source Approach

34

Pearlson, Keri E., and Carol S. Saunders. Managing and Using Information System: A Strategic Approach, 3rd edition. 2006.

Power, D. "DSSResources.COM." A Brief History of Decision Support Systems. March 10, 2007. [On-line] http://DSSResources.COM/history/dsshistory.html (accessed October 1, 2007).

Raden, Neil. "Business Intelligence 2.0: Simpler, More Accessible, Inevitable." IntelligentEnterprise.com. February 1, 2007. [On-line] http://www.intelligententerprise.com/showArticle.jhtml?articleID=197002610 (accessed October 1, 2007).

Rawat, Rajeev. "The Ins and Outs of Open-Source Business Intelligence." The Data Warehousing Institute (TDWI.org). October 24, 2007. [On-line] http://www.tdwi.org/info.aspx?id=44190 (accessed October 24, 2007).

Shields, Joseph, and Jasmine Bharucha. "Measuring IT Returns: Are we counting chickens from eggs?" CuttingEdge, Infosys, 2003.

Turban, Efraim, Dorothy Leidner, Ephraim McLean, and James Wetherbe. "Information Technology for Management: Transforming Organizations in the Digital Economy." In Information Technology for Management: Transforming Organizations in the Digital Economy, by Efraim Turban. 2006.

Varner, Philip E. "The economics of open source software." 1999. [On-line] http://www.cs.virginia.edu/~pev5b/writing/econ_oss/ (accessed October 1, 2007).

Ventana. Open Source BI Executive Summary. Ventana Research, 2006.

Williams, Steve, and Nancy Williams. "The Business Value of Business Intelligence." Business Intelligence Journal, 2003, 32-43.

Page 41: Business Intelligence for Small Enterprises: An Open Source Approach

A-1

Appendices

Appendix A. Rawat’s criteria for assessing OS BI

Original list of the criteria for OS BI assessment proposed by Rajeev Rawat (2007): 1. Business Need

1.1. Urgency 1.2. Sponsor 1.3. Users 1.4. Bottlenecks 1.5. Data sources 1.6. Data quality 1.7. Data owners 1.8. Business needs

2. BI Functionality Requirement 2.1. Computations 2.2. Integration 2.3. Reporting 2.4. Time 2.5. Embed-ability/SOA 2.6. Repository 2.7. Dashboard/portal 2.8. OLAP UI 2.9. OLAP Server 2.10. ETL 2.11. End-User Ad Hoc 2.12. Alerts 2.13. Semantic Layer 2.14. Security 2.15. Scheduling 2.16. OSS Community connection

3. Risk Profile 3.1. Financial 3.2. Operational 3.3. Performance 3.4. Skills gap

4. Skills match

4.1. Business Users 4.2. Systems Users 4.3. Current vendor teams 4.4. New vendor team

5. Systems environment 5.1. Server and storage connectivity 5.2. Data integration and interchange 5.3. Application compatibility

6. Pricing

6.1. Licenses 6.2. Maintenance 6.3. Support 6.4. Transition costs 6.5. Upgrades 6.6. Renewals

7. Vendor maturity 7.1. Customer satisfaction 7.2. Install base 7.3. Requirements gathering process 7.4. Reference account

8. Solution Maturity 8.1. Availability duration 8.2. Install base 8.3. Tools compatibility 8.4. Standards compatibility 8.5. Mature, world-class reporting products 8.6. Ready-to-use out of the box 8.7. ISV-ready: embeddable & extensible 8.8. Professionally localized & internationalized 8.9. Meets complete continuum of BI needs 8.10. Reporting, Analysis, Data Integration

Page 42: Business Intelligence for Small Enterprises: An Open Source Approach

B-2

Appendix B. Survey Questions

PART A. GENERAL QUESTIONS

1. What types of enterprises are your company’s main target group (check all

applicable)?

Small Medium Large

2. What percentage of your customers are Small Enterprises?

%

3. What percentage of your customers are Medium Enterprises?

%

4. Does your company specialize only in BI solutions?

Yes, we provide only BI consulting, integration and related services.

No, we provide many strategic IT solutions; BI is the one of them.

5. Which BI functionalities usually included in the complete BI solutions you provided (check all applicable)?

Reporting Spreadsheet Services

Dashboard Ad Hoc Query and Reporting

OLAP Server OLAP UI (a.k.a. Slicing and Dicing)

ETL Data Mining

Alerts Scheduling

6. What other functionalities/applications would you consider as “must have”

part of BI suite?

Page 43: Business Intelligence for Small Enterprises: An Open Source Approach

B-3

PART B. MINI CASE QUESTIONS

7. Approximately estimate each of the BI costs based on the mini case (mini

case description included in separate file):

Cost Type Estimated Cost

Acquisition Costs

1. BI software licenses SEK

2. Consulting, installation, and configuration SEK

3. RDBMS or other non-BI system licenses SEK

4. Hardware and other infrastructure SEK

Operations Costs

5. Maintenance and Support SEK/year

6. Upgrades SEK/year

7. User Trainings SEK/year

8. What other cost types you used to considered as typical in BI costs calculation?

9. Which BI vendor(s) would you use in the assumed solution for the mini case? SAP (Business Objects) IBM (Cognos) Oracle (Siebel, Hyperion) SAS Microsoft Other vendors: Proprietary/in-house developed

Page 44: Business Intelligence for Small Enterprises: An Open Source Approach

C-4

Appendix C. Interview and Survey Results

Interview and web-based survey results:

Question

Resp

on

den

t 1

Resp

on

den

t 2

Resp

on

den

t 3

Resp

on

den

t 4

Resp

on

den

t 5

Resp

on

den

t 6

1. What types of enterprises are your company’s main target group?

Medium, Large

Large Medium, Large

Medium, Large

Large Large

2. What percentage of your customers are Small Enterprises?

0% 5% 5% 5% 0% 5%

3. What percentage of your customers are Medium Enterprises?

10% 15% 15% 20% 10% 15%

4. Does your company specialize only in BI solutions?

No Yes Yes Yes No Yes

Interview only results:

Question Interviewee 1 Interviewee 2

5. Which BI functionalities usually included in the complete BI solutions you provided?

All except Data Mining, Alerts, and Scheduling.

All except Alerts.

6. What other functionalities/applications would you consider as “must have” part of BI suite?

None None

7.1 + 7.3. Cost of BI software, RDBMS and other licenses 200.000 SEK 200.000 SEK

7.2. Cost of consulting, installation, and configuration 600.000 SEK 500.000 SEK

7.4. Cost of hardware and other infrastructure 100.000 SEK 150.000 SEK

7.5. Cost of maintenance and support 300.000 SEK 300.000 SEK

7.6. Cost of upgrades 0 SEK (included in support fee)

20.000 SEK

7.7. Cost of user trainings 20.000 SEK 10.000 SEK

8. What other cost types you used to considered as typical in BI costs calculation?

None None

9. Which BI vendor(s) would you use in the assumed solution for the mini case?

Microsoft PerformancePoint 2007 and SQL Server 2005

Microsoft PerformancePoint 2007 and SQL Server 2005

Page 45: Business Intelligence for Small Enterprises: An Open Source Approach

D-5

Appendix D. Business Case Description

BACKGROUND “MobileComet” is a content provider (i.e. ringtones, pictures, games, etc.) for

mobile devices. Sales operated only through partner mobile operators. At the

moment company has 10 employees and annual turnover of €500K.

All operational data of the company stored in the single source – RDBMS database.

This source contains information about contents, customers, orders, mobile

handsets and their characteristics, and partners (known as aggregators). The

simplified domain model is depicted in figure d.1.

Aggregators are 3rd party resellers, who usually work based on revenue sharing

model. However, aggregators’ interest rates and similar financial information are

not stored in database and available only at financial department.

Customer

Phone Number 1

Join Date 1

Order

Content Type 0..1

Content

Publish Date 1

Alias 0..*

Category 0..*

Status 1

Rating 1

Artist

Content Format

Content Type 1

1 0..*ordered_by

0..*

1

1

0..*

0..*1

belongs_to

Mobile Handset

UA Prof 1

0..*0..*

supports

Content

Aggregator

0..10..*

sold_by

0..1

0..*

provided_by

1

0..*

Figure D.1. Domain Model of Operational Database

Page 46: Business Intelligence for Small Enterprises: An Open Source Approach

D-6

REQUIREMENTS MobileComet would like to implement Business Intelligence solution which will

help them:

Monitoring of Key Performance Indicators (KPI) via Dashboard (e.g. customer base, revenue, profit, etc.)

Automatic report generation (e.g. financial reports and invoices for internal usage and 3rd parties)

Slicing and Dicing; Ad-hoc analysis and reporting (e.g. ability to create complex, multidimensional queries to provide answer to questions like: What is the monthly percentage of Samsung phones which failed to load java games within last 1 year?)

Provide easier, online and centralized access to all data types.

Archive operational data. Operational database does not need anymore keep data older than 3 months. Thus, reduce load of operational db and save resources of production system.

Only 5 employees need to have access to the BI system. At this stage, company

does NOT need any Data Mining tools. However, they plan to implement it in

future.

DATA SOURCES Even though company has centralised database, there is a lack of some data. To fill

gaps BI system should be able to access central DB as well as other data sources

listed below:

1. Central Operational RDBMS

2. Statistics provided by Mobile Operators (real-time, usually over SOAP/XMLRPC)

3. Financial Data (rarely updated excel files, e.g. cost of services, interest rates of aggregators)

REQUIRED BUSINESS PROCESSES At this stage company wants to get insight into two main areas:

1. Sales Information

2. Customer Information and Behaviour

DATA WAREHOUSE STRUCTURE OVERVIEW Data from operational sources captured, cleansed, transformed and loaded into

data warehouse on the nightly basis. Granularity level of the data warehouse

structures is a one minute. There are usually 5000-6000 orders per day in

operational DB.