Goal-oriented Data Warehouse Quality Measurement

A Goal-orien ted F ramew ork for Data W arehousing Qualit y Measuremen t Cristina Cac hero and Jes ´ us P ardillo Departmen t of Soft ware and Computing Systems Univ ersity of Alicante, Spain {ccachero, jesuspv}@dlsi.ua.es Abstract. Requiremen ts engineering is kno wn to b e a k ey factor for the success of soft ware pro jects. Inside this discipline, go al-oriente d r e- quir ements engine ering approaches hav e shown specially suitable to deal with pro jects where it is necessary to capture the alignmen t betw een system requiremen ts and stakeholders’ needs, as is the case of data- w arehousing pro jects. Ho wev er, the mere alignment of data-warehouse system requirements with business goals is not enough to assure b et- ter data-warehousing products; measures and tec hniques are also needed to assure the data-warehouse qualit y . In this paper, we provide a mo d- elling framework for data-war ehouse quality me asur ement ( i ∗ D WQM). This framework, conceived as an i ∗ extension, provides supp ort for the definition of data-wareho use requirements analysis mo dels that include quan tifiable quality scenarios, defined in terms of w ell-formed measures. This extension has b een defined by means of a UML profiling archi- tecture. The resulting framew ork has b een implemented in the Eclipse dev elopment platform. Key w ords: UML, data-w arehouse, goal-oriented, i ∗ , measuremen t, re- quiremen ts, measurement, mo delling 1 In tro duction Data-w arehouse systems pro vide a m ultidimensional view of heterogeneous op- erational data sources in order to supply v aluable information to decision mak- ers. Its developmen t is usually based on the multidimensional mo delling, b e- cause of its intuitiv eness and its supp ort for high-p erformant queries [9, 10]. Since the data-warehouse in tegrates several operational data-sources, the design of m ultidimensional models has b een traditionally guided by supply-driven ap- proac hes [7, 8]. How ever, in order to assure the adequation of such designs to the information needs of decision makers, a requiremen t-analysis stage is needed. F or this stage, goal-orien ted frameworks ha ve pro ven specially suitable. The rea- son for this fact is tw ofold: First, goal-oriented framew orks provide constructs for the mo delling of large organisational contexts, whic h are the commonality in data-w arehouses. Second, they match the w ay in whic h decision makers express 2 C. Cachero, J. P ardillo themselv es, i.e. , in terms of general exp ectations or ob jectiv es that the data- w arehouse should supp ort. This suitability has b een materialised in prop osals suc h as i ∗ D WRA [12] or the one presented in [6]. Ho wev er, the inclusion of goals, although necessary , may not b e sufficien t to guaran tee the qualit y of data-w arehouse systems. Indeed, although a go od metho dology with accurate goal definitions may lead to go od and suitable data- w arehouse mo dels, man y other factors could influence their quality , suc h as hu- man decisions. It is thus necessary to complete data-warehousing metho dologies with measures and techniques for product quality assessment [16, 17, 2, 15]. One of the b est kno wn techniques in this sense, emphasised in w ell-known soft ware dev elopment processes such as the unifie d pr o c ess (UP) [11], is the definition of qualit y scenarios as part of the requiremen ts-analysis workflo w. Quality sce- narios define measures that serve to v alidate the requirement to which they are asso ciated. F urthermore, they sp ecify the context in which the measurement pro cess is to tak e place ( e.g. , it is not the same measuring p erformance with 10 sim ultaneous users than with 10,000). Qualit y scenarios turn requirements in to measurable requiremen ts. In order to mo del these measurable requiremen ts, in this paper w e extend i ∗ D WRA. The result of this extension is the i ∗ -based data-war ehousing quality me asur ement framew ork ( i ∗ D WQM). An imp ortan t adv antage of this frame- w ork is that, to our knowledge extent, it is the first prop osal that traces qualit y scenarios back to the originating stakeholders’ needs. Also, our prop osal stresses the role of an often forgotten actor in data-warehouse dev elopment: the quality manager. Quality managers are resp onsible for orchestrating and leveraging the differen t stak eholders’ in terests during the data-warehouse dev elopment. Suc h in- terests include certain quality restrictions that m ust b e resp ected for the pro ject to b e considered successful. It is imp ortan t to note how, contrary to other mea- sures prop osed by differen t authors for i ∗ diagrams [3, 4], our emphasis is not on the quality of the diagram p er se , but on the pro vision of mechanisms to mo del the lev els of quality required by the system under developmen t. i *DWQM i * for Data W arehouse Quality Measurement i *DWRA i * for Data W arehouse Requirements Analysis [5] Decision Maker Quality-assured Data W arehouse I N F O R M A T I O N Q U A L I T Y OUR PROPOSAL DERIVE Quality Stake- holder Fig. 1. Adding measurable quality scenarios for data-warehouse requiremen ts analysis Goal-orien ted Data W arehouse Qualit y Measurement 3 The remainder of the pap er is structured as follows: we present i ∗ D WQM next ( § 2) as an extension of i ∗ D WRA (see Fig. 1) that, for the sake of under- standabilit y , is also sketc hed. Both the measurement concepts and the chosen notation are then further illustrated with a sample application ( § 3). Our proposal has b een defined b y using the unifie d mo del ling language (UML) [14] profiling capabilities, whic h has p ermitted us to implemen t it in the Eclipse 1 dev elopment platform ( § 4). Finally , w e summarise the main contributions of this paper and outlines some future lines of researc h ( § 5). 2 Mo delling F ramework for Data-w arehouse Quality Measuremen t Empirical research shows that the definition of measures and the description of measuring efforts in literature suffer from the typical symptoms of an y relatively y oung discipline [1] and present man y flaws that compromise their completeness and consistency . In order to ov ercome these problems, in [5] a softwar e me a- sur ement ontolo gy (SMO) has b een prop osed. Until the new ISO/IEC 25000 standard series app ear 2 , SMO reflects a compromise solution to solve the many inconsistencies and gaps detected in standards and research prop osals. W e ha ve follo wed this on tology for the definition of i ∗ D WQM, in order to facilitate its adoption in the measurement domain. F or the sake of understandability , in Ap- p endix A the definitions of the ontology terms that hav e b een used along this pap er are formally repro duced. Interested readers ma y find further information ab out the whole ontology in [5]. 2.1 Mo delling Requiremen ts Analysis with i ∗ D WRA As it has been aforementioned, i ∗ D WRA is a data-w arehouse requirements anal- ysis framework that has pro ven useful for the discov ery of data-warehouse re- quiremen ts out of business goals. This purp ose is achiev ed by iden tifying the de- cisions that decision mak ers usually are faced to. The i ∗ D WRA framew ork has b een defined in tw o steps; first, in [12], a UML profile for i ∗ (the i ∗ profile, see T able 1, col. 4 & 5) has b een pro vided. This profile elegantly redefines i ∗ concepts and relationships [18] in terms of UML mo delling elements. These elements p er- mit to model both the organisational con text (b y means of the i ∗ str ate gic dep en- dency (SD) diagram) and the actors’ rationale (by means of the i ∗ str ate gic r a- tionale (SR) diagram) when interacting with the data warehouse. Suc h concepts include intentional elements –actors ( ), goals ( ), tasks ( ), softgoals ( ), resources ( ), and beliefs ( )– and intentional r elationships –inten tional de- p endencies ( ), means-end relationships ( ), task-decomp ositions ( ), and con tributions ( ). Ov er this i ∗ profile, the second step has consisted in adding specific seman tics for data-warehouse requirements analysis (see T able 1, col. 1–3). F or the sak e 1 URL: www.eclipse.org 2 Namely , softwar e pr o duct quality r e quir ements and evaluation (SQuaRE) 4 C. Cachero, J. P ardillo of simplicity , in this table only the i ∗ elemen ts that ha ve been extended by the i ∗ D WRA framework are listed. T able 1. Mapping i ∗ D WRA concepts in to the i ∗ framew ork Analysis Concept i ∗ D WRA i ∗ Profile Stereot yp e Notation Stereot yp e UML Strategic Goal Strategy +  strategy  Goal Class Decisional Goal Decision +  decision  Goal Class Informational Goal Information +  information  Goal Class Info. Requirement Requiremen t +  task  T ask Class Con text Resource +  context  Resource Class Measure Resource +  measure  Resource Class Let us now give an example to illustrate how to prop erly read T able 1: Let us assume that we wish to mo del a data warehouse information requirement (see col. 1) in i ∗ D WRA. F or this task, we would ha ve to use the Requirement stereot yp e (col. 2). This stereot yp e pro vides additional semantics and notation (col. 3) to the i ∗ task concept. i ∗ tasks are mapp ed into UML by means of the Task stereot yp e (col. 4) on the UML Class mo delling element (col. 5). This mo delling framework has b een the basis on whic h we ha ve p erformed a further extension to p ermit the definition of qualit y scenarios, whic h we hav e called i ∗ data-war ehouse quality me asur ement ( i ∗ D WQM) framework. It is worth noting that the Measure concept defined in i ∗ D WRA refers to the analysis me asur es employ ed during the decision-making pro cess supp orted b y the data-warehouse. Therefore, it m ust not b e confused with the measure concept that app ears in the con text of data-w arehouse quality scenarios, which w e will explore next. 2.2 Mo delling Qualit y Scenarios with i ∗ D WQM T able 2. Mapping of SMO concepts in to i ∗ D WQM SMO Concept Equiv alent i ∗ Elemen t Indicator, Derived Measure, Base Measure Goal Analysis Mo del, Measuremen t F unction, Measurement Method T ask En tity Class, Decision Criteria Resource A ttribute Belief The i ∗ D WQM framework enric hes i ∗ D WRA with the capabilit y of sp ecifying quan tifiable quality scenarios for the assurance of quality requirements asso ciated with data-warehouses. As we ha ve aforementioned, our prop osal is based on Goal-orien ted Data W arehouse Qualit y Measurement 5 SMO [5], in order to facilitate its understandability and help in its adoption b y qualit y stakeholders. Mapping Quality Stakeholders into A ctors. In order to mo del quality stakehold- ers in i ∗ D WQM, follo wing the i ∗ D WRA prop osal, we use the i ∗ actor mo delling elemen t. F or instance , qualit y managers are actors that are in charge of defining quan tifiable quality scenarios. Mo del ling Quality Sc enarios for Data War ehouses. i ∗ D WQM establishes a cor- resp ondence b et ween particular SMO measuremen t concepts and more gen- eral i ∗ D WRA concepts. Such mapping is presen ted in T able 2. In this ta- ble, we can observe how SMO measures ( Indicator s, Derived Measure s and Base Measure s) are mapp ed in to measurement Goal s that can b e achiev ed through certain Task s. These tasks are, namely , performing an Analysis Model , a Measurement Function or a Measurement Method , resp ectiv ely . The mea- suremen t concepts Entity Class and Decision Criteria are mapp ed into Resources , while the Attribute concept is mapp ed into a Belief in i ∗ D WRA. Similarly , i ∗ D WQM maps the SMO relationships in to i ∗ in tentional relation- ships in a hierarc hical manner, as can b e seen in Fig. 2. : Indicator : AnalysisModel : DerivedMeasure : MeasurementFunction : DecisionCriteria : BaseMeasure : MeasurementMethod : EntityClass Quality Stakeholder Has Uses Uses Uses Calculated with Calculated with Uses : BaseMeasure : MeasurementMethod Uses Uses : Attribute Is performed on Object Diagram for SMO Concepts i * Correspondence Goal T ask Resource Goal T ask Goal Goal T ask T ask Resource QualityScenario Belief Is performed on Fig. 2. Mapping a SMO o ccurrence for specifying qualit y scenarios with i ∗ D WQM The upp er part of this figure presents a SMO-based mo del that represents a generic quality scenario, while the low er part presents the equiv alent i ∗ mo del that has serv ed to i ∗ D WQM as a basis for further enrichmen t. In this figure, we observ e ho w, in SMO, an Indicator is related with an Analysis Model , which in turn has one or more Decision Criteria asso ciated. An indicator is in fact 6 C. Cachero, J. P ardillo a type of Measure that is made up of several other measures, b e them Derived Measure s or Base Measure s. Derived measures are related with Measurement Function s, while base measures are associated with Measurement Method s. In SMO, analysis mo dels, measurement functions and measurement metho ds can b e related with EntityClass es through Attribute s. The counterpart i ∗ D WQM relationships are presented in the lo wer part of Fig. 2. In this figure, we observe how indicators (mapped into goals) and anal- ysis mo dels (tasks) are related through a Means-End relationship. The same relationship app ears b et ween deriv ed measures (goals) and their correspond- ing measurement functions (tasks) and base measures (goals) and their corre- sp onding measuremen t metho ds (tasks). Another relev ant relationship is that of Task Decomposition that app ears b et ween analysis mo dels (tasks) and their related measures (goals) or decision criteria (resources), and also betw een mea- suremen t functions (tasks) and the asso ciated measures (goals), or betw een mea- suremen t metho ds (tasks) and the associated en tity classes (resources). Last, a Contribution relationship provides the attribute (b elief ) that p ermits the con- nection b et ween measuremen t metho ds (tasks) and entit y classes (resources). With this structure it is p ossible for quality managers to sp ecify the quality scenarios asso ciated with data-warehouse requirements. Mo del ling Dep endencies among Stakeholders. The quality scenarios mo delled with i ∗ D WQM must be connected with particular non-functional requiremen ts (softgoals) in the con text of a particular data-warehouse informational scenario. Suc h information requirements and the asso ciated softgoals provide resp ectiv ely the context and the rationale for the measurement activity . As we hav e afore- men tioned, i ∗ D WRA information requiremen ts are mo delled as analysis Tasks . Analysis tasks ma y hav e different softgoals asso ciated, which specify how the decision maker expects those tasks to b e performed. At this p oin t, existing qual- it y mo dels for data warehouses [15] are useful to choose among the set of non- functional requiremen ts that are typical of this type of applications. F rom the existing relationship b et ween softgoals and measures, a dep endency b et ween the corresponding actors can be inferred. In i ∗ D WQM, this dep endency is mo delled as an in tentional dep endency from decision makers’ softgo als ( de- p ender ) to quality stak eholders’ go al indicators ( dep ende e ) in order to achiev e a given qualit y scenario go al ( dep endum ) that a particular quality stak eholder kno ws how to measure. This mo delling solution can b e seen in Fig. 3. Mo del ling Me asur ement A ttributes. In the mapping presented so far, some SMO concepts, namely the measure c haracteristics Unit Of Measurement , Scale , and Type Of Scale are still missing. F or the mo delling of these concepts, i ∗ D WQM has made use of the notes mechanism enabled in i ∗ . F urthermore, the UML scaffolding that we hav e used to implement i ∗ D WQM ( § 4) further pro vides the necessary lev el of formalism to prop erly sp ecify these SMO concepts. Goal-orien ted Data W arehouse Qualit y Measurement 7 3 Case Study The case study c hosen to illustrate our approac h consists in a compan y selling automobiles across several coun tries. In this example (see Fig. 3), w e ha v e iden ti- fied a sales manager as an actor that has several information requirements to be fulfilled b y the data warehouse to be dev elop ed. During the requiremen ts discov- ery phase, w e ha ve iden tified that automobile sales b e incr e ase d is a strategic goal of the sales manager. F rom this strategic goal, sev eral decision goals ha ve been deriv ed: sales pric e b e de cr e ase d , pr omotions b e determine d , and so on. F o cusing on the first decision goal, we hav e obtained tw o information goals: automobile pric e b e analyse d and automobile sales b e analyse d . Concerning the first one, we ha ve recognised that the information requirement analyse automobile sale pric e is the means for achieving this decision goal, and for this analysis, the sales manager needs to chec k the prices and automobiles as fact and dimensions of the data w arehouse analysis, resp ectiv ely . AM_RepFlexLev (analysis model) Report (entity class) {RDT ,DST ,LCT}.unit="seconds" {RDT ,DST ,LCT}.scale="Natural" {RDT ,DST ,LCT}.typeOfScale=ratio RDT(R)<60 = OK (decision criteria) Analyse Automobile Sale Price «requirement» «context» Automobile «measure» Price Flexibility [Reporting] Report Flexibility Level (indicator) DST(R) (base measure) LCT(R) (base measure) DST(R)+LCT(R) (m. function) RDT(R) (derived measure) Automobile Sales be Increased Sale Price be Decreased Promotions be Determined Automobile Sales be Analysed Automobile Price be Analysed Sales «context» Date «decision» «information» «information» «strategy» «decision» «businessProcess» Sales Manager Quality Manager Time LCT(R) (m. method) Time DST(R) (m. method) Structural Complexity (attribute) Ad-hoc Reporting (quality scenario) Fig. 3. i ∗ D WQM model for the ad-ho c reporting quality scenario In addition, the analysis of the automobile sales price also needs the system to b e flexible, where by flexible we refer to the extent to which the data-warehouse soft ware facilitates ad-ho c rep orting [15] (see Fig. 3). The quality scenario as- so ciated with this softgoal has b een defined by the quality manager as follo ws: “A sales manager is able to design the required rep ort, based on her mental mo del of the data warehouse, in less than 60 seconds” (referred to as “Ad-hoc Rep orting” in Fig. 3). In Fig. 3, this quality scenario has b een mo delled with 8 C. Cachero, J. P ardillo the aid of a r ep ort flexibility level indicator that ev aluates the time it takes to the sales manager to design rep orts ad ho c . This indicator relies on a derive d me asur e called r ep ort design time ( RDT(R) ) that, measured ov er a given rep ort, returns the num b er of seconds that it tak es to the sales manager to actually de- sign the rep ort. The quality scenario establishes that no more than 60 seconds is an acceptable time interv al. This fact is captured in the de cision criteria asso ci- ated with the indicator. This measure is calculated through the sum of tw o base measures (measurement function), namely the data sele ction time ( DST(R) ) and the layout c omp osition time ( LCT(R) ). These measures are assigned v alues b y applying the corresp onding measurement metho d, which consists in both cases in timing the corresp onding tasks o ver a given report (the entit y class). The b elief in Fig. 3 indicates that these measures ev aluate the structural complexity attribute asso ciated with the rep ort. Last, the unit of measuremen t, scale, and t yp e of scale concepts are sp ecified as additional notes in Fig. 3. 4 Implemen tation The i ∗ D WQM has b een implemented as an extension of UML and has b een deplo yed in the Eclipse developmen t platform (Fig. 4). Sp ecifically , UML pro- vides a standard extension formalism: UML profiles. Thes e profiles consist of a set of stereotypes for particular UML modelling elements and some related tag definitions and constrain ts that, together, p ermit UML to host our mo d- elling language. The i ∗ D WQM profile is based on t wo preexisting UML profiles for modelling i ∗ diagrams adapted to the data-warehousing discipline, i.e. , the i ∗ D WRA and the i ∗ profile [12] (see Fig. 4). So far, we ha ve presented the map- pings that supp ort the definition of the necessary stereotypes for prop erly rep- resen ting the i ∗ D WRA modelling elements in UML. While some concepts hav e b een directly mapp ed to i ∗ elemen ts, others (namely , Unit Of Measurement , Scale , and Type Of Scale concepts), whic h in a pure i ∗ framew ork can be mo delled as notes, ha ve b een implemented in our profile as tag definitions asso- ciated with the measurement-related stereot yp es. In addition to the mo delling elemen ts, i ∗ D WQM also considers the required constraints (derived from SMO) that assure the righ t use of these stereotypes, e.g. , forcing that only analysis mo dels b e the means for achieving an indicator. In this wa y , we pro vide a coher- en t mo delling en vironment for (i) analysing data-w arehouse requirements and (ii) asso ciating a quantitativ e means (through measures) to assess their qualit y . 5 Conclusions In this pap er, we hav e presented i ∗ D WQM, a mo delling framework to sp ecify measurable qualit y scenarios that contribute to the assessment of the quality with which data-warehouse requiremen ts are achiev ed. The completeness and unam biguity of the framework is facilitated b y the use of a w ell-known Softw are Measuremen t On tology [5] for its definition. Moreov er, the UML scaffolding on Goal-orien ted Data W arehouse Qualit y Measurement 9 «profile» i * «profile» i *DWRA «import» «import» «use» «profile» i *DWQM Eclipse «platform» i * : «goal», «task», «resource», ... i *DWRA : «strategy», «decision», «fact», «dimension» ... i *DWQM : «indicator», «measure», «analysisModel», «attribute» ... + Semantics + T ag definitions + Constraints + Notation STEREOTYPES Fig. 4. The implemented UML profiling architecture for modelling with i ∗ D WQM whic h our approac h is based contributes to achieving the desired degree of p orta- bilit y . The use of our framework complemen ts existing goal-oriented approac hes for the dev elopment of data warehouses with several additional adv antages: – It increases the w eight of quality scenarios and quality managers in the mo delling pro cess. – It adds emphasis to the, often forgotten, measurable asp ect that should be alw ays associated with requiremen ts in order to decrease the risks associated with system dev elopment. – It pro vides a means to reason ab out how such measurement should take place, with the final goal of orc hestrating and lev eraging the different stake- holders’ in terests during the data-warehouse developmen t. – It provides traceability b et ween the quality scenarios and the particular stak eholders’ needs Additionally , the measuremen t domain also obtains b enefits out of b eing asso ciated with goal-oriented approaches, among which w e w ould like to stress out the provision by these approaches of a muc h ric her organisational context than the one pro vided by the SMO for the definition of measures. Although this framework has b een devised for its application to data ware- houses, their characteristics make us b eliev e that i ∗ D WQM can b e equally use- ful for other domains. This hypothesis constitutes one of our future lines of researc h. Last but not least, measuring models op en the path for mo del-driv en data-w arehouse developmen t framew orks (see e.g. [13]) to tak e them into account for the automatic generation of application tests. 6 Ac kno wledgements This w ork has been supported b y the pro jects: TIN208-00444, ESPIA (TIN2007- 67078) from the Spanish Ministry of Education and Science (MEC), QUASI- MODO (P A C08- 0157-0668) from the Castilla-La Mancha Ministry of Educa- tion and Science (Spain), and DEMETER (GVPRE/2008/063) from the V alen- cia Gov ernmen t (Spain). Jose-Norb erto Maz´ on and Jes ´ us Pardillo are funded by MEC under FPU gran ts AP2005-1360 and AP2006-00332, resp ectiv ely . 10 C. Cachero, J. P ardillo References 1. Lionel C. Briand, Sandro Morasca, and Victor R. Basili. An Op erational Pro cess for Goal-Driv en Definition of Measures. IEEE T r ans. Software Eng. , 28(12):1106– 1125, 2002. 2. Samira Si-Said Cherfi and Nicolas Prat. Multidimensional Schemas Quality: As- sessing and Balancing Analyzability and Simplicity . In ER (Workshops) , pages 140–151, 2003. 3. Xa vier F ranch. On the Quan titative Analysis of Agent-Orien ted Mo dels. In CAiSE , pages 495–509, 2006. 4. Xa vier F ranch, Gemma Grau, and Carme Quer. A F ramework for the Definition of Metrics for Actor-Dep endency Mo dels. In RE , pages 348–349, 2004. 5. F ´ elix Garc ´ ıa, Manuel F. Bertoa, Coral Calero, Antonio V allecillo, F rancisco Ruiz, Mario Piattin i, and Marcela Genero. T ow ards a consistent terminology for softw are measuremen t. Inform. Softwar e T e ch. , 48(8):631–644, 2006. 6. P aolo Giorgini, Stefano Rizzi, and Maddalena Garzetti. Goal-oriented requiremen t analysis for data warehouse design. In DOLAP , pages 47–56, 2005. 7. Matteo Golfarelli, Dario Maio, and Stefano Rizzi. The Dimensional F act Mo del: A Conceptual Mo del for Data W arehouses. Int. J. Co op. Inf. Syst. , 7(2-3):215–247, 1998. 8. Bo do H ¨ usemann, Jens Lech tenb¨ orger, and Gottfried V ossen. Conceptual data w arehouse mo deling. In DMDW , page 6, 2000. 9. William H. Inmon. Building the Data War ehouse . Wiley , 2005. 10. Ralph Kimball and Margy Ross. The Data War ehouse T o olkit . Wiley , 2002. 11. Craig Larman. Applying UML and Patterns : An Intr o duction to Obje ct-Oriente d Analysis and Design and Iter ative Development . Pren tice Hall, 2004. 12. Jose-Norb erto Maz´ on, Jes ´ us Pardillo, and Juan T rujillo. A Mo del-Driv en Goal- Orien ted Requirement Engineering Approac h for Data W arehouses. In ER Work- shops , pages 255–264, 2007. 13. Jose-Norb erto Maz´ on and Juan T rujillo. A MD A approac h for the dev elopment of data warehouses. De cis. Supp ort Syst. , 45(1):41–58, 2008. 14. Ob ject Management Group. Unified Mo deling Language (UML), version 2.1.1. http://www.omg.org/technology/documents/formal/uml.htm , F ebruary 2007. 15. F´ a Rilston Silv a P aim and Jaelson Castro. Enhancing Data W arehouse Design with the NFR F ramework. In WER , pages 40–57, 2002. 16. Man uel A. Serrano, Coral Calero, Houari A. Sahraoui, and Mario Piattini. Em- pirical studies to assess the understandability of data w arehouse schemas using structural metrics. Softwar e Qual. J. , 16(1):79–106, 2008. 17. P annos V assiliadis. Data War ehouse Mo deling and Quality Issues . PhD thesis, National T ec hnical Universit y of A thens, 2000. 18. Eric S. K. Y u. T ow ards Modeling and Reasoning Support for Early-Phase Require- men ts Engineering. In RE , pages 226–235, 1997. Goal-orien ted Data W arehouse Qualit y Measurement 11 A Soft w are Measuremen t Ontology T erms Definition T able 3. Excerpt of SMO term definitions employ ed in this pap er Concept Definition En tity Class The collection of all en tities that satisfy a given predicate A ttribute A measurable ph ysical or abstract prop ert y of an entit y , that is shared b y all the entities of an en tity class Scale A set of v alues with defined properties T yp e of Scale The nature of the relationship betw een v alues on the scale Unit of Measure- men t P articular quantit y , defined and adopted by conv ention, with whic h other quan tities of the same kind are compared in order to express their mag- nitude relative to that quan tity Base Measure A measure of an attribute that do es not dep end up on any other measure, and whose measurement approac h is a measuremen t metho d Deriv ed Measure A measure that is deriv ed from other base or derived measures, using a measuremen t function as measurement approac h Indicator A measure that is deriv ed from other measures using an analysis mo del as measurement approac h Measuremen t Metho d Logical sequence of op erations, described generically , used in quantifying an attribute with resp ect to a sp ecified scale. (A measurement metho d is the measurement approac h that defines a base measure) Measuremen t F unction An algorithm or calculation p erformed to combine t wo or more base or derived measures. (A measurement function is the measurement ap- proac h that defines a derived measure) Analysis Mo del Algorithm or calculation com bining one or more measures with asso ci- ated decision criteria. (An analysis mo del is the measuremen t approach that defines an indicator) Decision Criteria Thresholds, targets, or patterns used to determine the need for action or further inv estigation, or to describe the level of confidence in a given result