Integrating Customer Requirements into Data Product Design

Transcription

1 114 Econom Informatics, 2003 Integrating Customer Requirements into Data Product Design Ec., Ph.D. candidate Otilia PÎRLOG Successful data products are ones that meet customer needs, are innovative and that offer value. Qualit Function Deploment offers a continuous process for en suring that all the different qualit issues are considered during the whole design and production chain. It provides a plane view of designed process and is tpicall applied to small sstems. Its etension for large sstems consists of few tree-dimensional cascading structures. This is an important tool for understanding the customer and deploing a structured manner throughout the data life ccle. A set of matries links customer desires, data qualit and sstem functions with subsstems / procedures dedicated to generate data product desired. The formaliation of the methodolog given here demonstrates the capacit of using it like a ke component to assist in creative thinking and problem solving. The subsstems / procedures proposed to be used are related with the corresponded costs. D ata Qualit Function Deploment for Large Sstems The compleit of data product administration involves man sstems. The main goal of data qualit is to translate the customer s demands (Voice of Customer) into designed targets. Qualit Function Deploment (QFD) invests time in planning to reap a profit in a shorter overall development ccle. The requirements are decomposed in a logical fashion, from the level of customer wants to the detailed process, using a set of correlated matries. The central matri mediates the customer wants with supplier plans to provide them. This basic matri can be epanded to provide additional insight to the supplier and cascaded to identif process parameters that must be controlled. There are man varieties of QFD and man variations of the charts used. Data Qualit Function Deploment (DQFD) is a method of translating customer desires throughout the entire data product ccle. It relates ideas to ideas, ideas to data, data to data process. This is one of the most powerful approaches for delivering a valuable data product to customers. It helps to achieve high qualit data products b improving data design and process qualit. It can also be viewed as the allocation of requirements to subsstems in that a subsstem must meet a set of requirements. DQFD is a pointed wa of listening to customers to understand eactl what the want and then, using a logical sstem, to determine how to best fulfill those requirements with available recourses. B relating these customer demands to the technical requirements (Voice of Developer), a set of target values for each technical requirement is logicall derived. It addresses a set of dimensions: customer desire, data qualit, functions, subsstems, procedures. Since requirements and technolog largel dictate the compleit of the sstem, analsis is focused on the compleit of data. Because the performance increases, the first unit cost of large sstems has been escala t- ing. The primar cost parameters are the sie and the compleit of sstems. In order to estimate and optimie the cost of data product, we have to anale a lot of. The solution is Data Qualit Function Deploment for Large Sstems (DQFDLS). It links the concurrent engineering process, the robust design process and the costing process to the data qualit features. The effect is to generate a tightl project management process of high dimensionalit which flushes out issues earl to provide a high qualit, low cost, and, hence, competitive data product. The Three -dimensional Model The first understanding of the basic structure for DQFD is giving a plane view with multi-

2 Econom Informatics, ple correlations in all directions. This tec h- nique demonstra tes the meaning of its name, a matri of matries. With regard to DQFDLS, the picture is changed in a set of three-dimensional derived each-another shapes. This offers the possibilit of a better Customer desires look inside the sstem and to decide witch wa is the best one for our design. Ever le - vel of anale consists of a tree -dimensional representation (figure 1) in which are presented those categories of information that are correlated at that moment. Sstem functions B1 The start is represented b the customer desires, the qualit demanded b him. The customer desires are placed on the ais of a matri. A data qualit characteristic is a mea s- urable attribute b which one can measure whether a customer is getting the demanded qualit. For each customer desire are settled the data qualit. Qualit cha r- acteristics are defined through brainstorming to generate an affinit diagram. After forming a tree diagram of the chosen data qualit, those on the lowest level are placed on the other ais of the analing matri. Each data qualit are compared with each customer desire to determine the tpe of correlation: there is no correlation (value=0), a weak correlation (value=1), a moderate correlation (value=3) or a strong correlation (value=5). B summariing the customer desires for the specific data qualit characteristic it provided a value for that characteristic. This ma be interpreted as the value of a data qualit characteristic for a specific customer desire valuation. Mathematicall speaking, the vector of values of customer desires is transformed into vector values for data qualit, using the customer desire / data qualit correlation matri (A1 matri). A1 C1 Requirement Fig. 1. Requirement Definition The process is continued with identifing functions. A function is something the sstem must do to ensure the demanded qualit. Usuall it is defined in the form <verb, noun>. Qualit and sstem functions intersect and define a requirement variable of the form <function, attribute> (C1 matri). Each requirement variable is a measurable attribute for which a correlation eists between the function and the data qualit characteristic. Requirement can be fied to create requirements or the can be used as design guidelines for improving the sstem. If these can be related through equations, the provide a parametric behavioral description of the product. and functions can be ranked in terms of transformed customer value to determine which are the most important. If resources are constrained, then the most priorit can be given to those with the highest customer value. The same wa is followed to correlate the functions with customer desires using the customer desire/function correlation matri (B1 matri). The determined ais for the first stage of anale are: - of data product for satisfing customer desires; - customer desires;

3 116 Econom Informatics, functions used to achieve data. The process is continued b earmarking subsstems / procedures to the determined functions. This forms the new spatial structure of Subsstems/ Procedures the model (figure 2). It can also be viewed as the allocation of requirements to subsstems in that a subsstem must meet a set of requirement var iables. Sstem functions B2 Now we are able to see the correlation that eists between data and subsstems / procedures allocated to rea lie them (A2 matri). Because of the wa we used, the result of these designed subsstems / procedures will be qualitative data that will satisf the customer requirements. A2 C1 Requirement Fig. 2. Allocated Procedural Support Subsstems/ Procedures Ever designed project has to take into account the level of necessar resources. The new step of the analing process is given b the replacing of the sstem functions with the other dimension consist of the cost of qualit (Figure 3). Cost of data qualit COST1 This tree-dimensional picture praises costs in two appearances: first, the relationship that eists between costs and subsstems / procedures (COST1 matri) and secondaril, between costs and the obtained result, data qualit (COST2 matri). The first aspect shown represents an eas support for appling the activit based costing methodolog. A2 COST2 Cost Fig.3. The Cost of Data Qualit All these spatial structures finalie the prototpe support sstem. The team involved in the design of data product will evaluate it and will simulate different alternatives. This isn t a difficult approach because of the fleibilit delivered b the prototpe support sstem. In the same time, the result of this methodolog allows the suitable maintenance. Because of the wide range of epertise required for large sstems, it becomes evident

4 Econom Informatics, the need to decentralie the approach b decomposing the sstem into meaningful subgroups with lower interaction. The geometric nature of the method represents a natural medium in which to perform this decomposition and manage the interaction. The Formaliation of the Method A ver important aspect of using DQFDLS is the capabilit to translate the formal relatio n- ships in a mathematical form. The correlative method we provide here can be used as a real instrument in order to simulate different alternatives of the designed project. Finall, the best one will be select. Ever spatial structure presented above is corresponded with a set of tree correlated matries. To advance from one stage to the other means to change onl a spatial dimension. From mathematical point of view this is similarl with the keeping of one matri like support for the new representation and to add to it the two new derived matries. The first set of matries is shown in Figure 4. Let F1 Fn2 be sstem functions, Char 1 Char n1 be data qualit and Des 1 Des n3 customer desires. Therefore n1 denote the number of data qualit identified, n2 denote the number of sstem functions and n3 the range of the set of customer desires. To individualie a variable we will use i=1 n1 for data qualit, j=1 n2 for sstem functions and k=1 n3 corresponding to customer demands.. Sstem functions F 1 F n2 Customer desires Des 1 Des n3 C1(1,1) C1(1,n2) A1(1,1) A1(1,n3) C1( i, j ) A1( i, k ) Char 1 Char n1 C1(n1,1) C1(n1,n2) A1(n1,1) A1(n1,n3) B1(1,1) B1(1,n2) Customer desires Des 1 Des n3 B1(n3,1) B1( k, j ) B1(n3,n2) Fig. 4. Set of es to Anale Customer Desires The net set of matries is built accordingl with the previous manner, b recovering the place of customer desires with subsstems / procedures decided (P 1 P n4 ). Figure 5 presents the new set of determined matries, where n4 denote the number of subsstems / procedures and p is used to identif an element, p=1 n4. Sstem functions F 1 F n2 Subsstems / procedures P 1 P n4 C1(1,1) C1(1,n2) A2(1,1) A2(1,n4) C1( i, j ) A2( i, p ) Char 1 Char n1 C1(n1,1) C1(n1,n2) A2(n1,1) A2(n1,n4) Subsstems / procedures B2(1,1) B2(1,n2) P 1 P n4 B2( p, j ) B2(n4,1) B2(n4,n2) Fig. 5. Set of es to Anale Subsstems / Procedures The last step of anale involves the cost of data qualit. The connection between the last and the actual stage is given b A2 matri. It will be the base of the new structure. In this case n5 is the range of ident ified categories of cost and c particularies an element of the added matries, where c=1 n5 (Figure 6). Using the double view of projected costs,

5 118 Econom Informatics, 2003 matries COST1 and COST2, it is possible to simulate multiple scenarios and evaluate them in order to identif the best technical solution with the lowest cost level. Subsstems / procedures P1 Pn4 costs Cost1 Costn5 Char1 Charn1 costs Cost 1 Cost n5 A2(1,1) A2(1,n4) COST1(1,1) COST1(1,n5) A2( i, p ) COST1( i, c) A2(n1,1) A2(n1,n4) COST1(n1,1) COST1(n1,n5) COST2(1,1) COST2(1,n4) COST2( c, p) COST2(n5,1) COST2(n5,n4) Fig. 6. Set of es to Anale the Data Qualit Cost The adherent cost to a data characteristic is computed b summariing corresponding costs: COST1 j = COST1( i, c ) The level of cost for ever subsstems / procedures is computed in the same manner: COST2 p = COST2( c, p ) These sets of matries denote the eas maintenance of designed sstem. Regarding to large sstems, this approach delivers also a strong support for decomposing the designed sstem. This is a ver important appearance for the designing team. It suggests also the capacit of using the accumulated eperience for some data qualit and to pa more attention for the rest of them. There are evident benefits of presented method: - improves focus on customer needs: - helps prioritie customer requirements; - helps identif competitive gaps (benchmarking); - reduce late design changes; - promotes teamwork; - ensures inter-departmental communication; - facilitates concurrent engineering. References [Aung02] Aungst,S.,Barton,R.,Wilson,D.T.: Integrating Marketing Models, Institute for the Stud of Business Markets, The Pennslvania State Universit, ISBN Report [Dean98] Dean, E.B.: Qualit Function Deploment from the Perspective of Co mpetitive Advantage, 1998, akao.larc.nasa.gov/dfc/ qfd.html [Daha02] Dahan,E.,Hauser,J.: The Virtual Customer, Journal of Product Innova tion Management, September, [Fran02] Franke,N.,Von Hipped,E.: Satisfing Heterogeneous User Needs via Innovation Toolkits: The Case of Apache Securit Software, Massachusetts Institute of Tec h- nolog (MIT) Sloan School of Management, Cambridge, MA, W orking Paper, [Rolp00] Rolph, P., and Bartram, P.: The Information Agenda: Harnessing Relevant Information in a Changing Business Environment, London, Management Books, 2000.