A Business Architecture framework for industrialisation and standardisation in a modern National Statistical Institute Piero Demetrio Falorsi1, Giulio Barcaroli2, Alessandra Fasano3, Nadia Mignolli4 1
Italian National Institute of Statistics (Istat), e-mail:
[email protected] Italian National Institute of Statistics (Istat), e-mail:
[email protected] 3 Italian National Institute of Statistics (Istat), e-mail:
[email protected] 4 Italian National Institute of Statistics (Istat), e-mail:
[email protected] 2
Abstract In recent years, many European and international statistical organisations have initiated developments to restructure their statistical production process, with the aim of improving its efficiency and the ability to produce outputs that better satisfy user needs. In this framework, the main topic of this study is to describe the path towards the design of a Business Architecture (BA) suitable to meet the specific requisites of the Italian National Institute of Statistics (Istat). Like other National Statistical Institutes, for several years Istat has been engaged in a series of complex challenges, dictated on the one hand by the requirement to increase the production of statistical information and its quality, on the other hand by the need to reduce both the total cost for its production and the respondent burden, so as to work in a more efficient and optimised way. In order to find proper solutions and increase the overall efficiency, since 2010 Istat has launched a Programme called Stat2015, that is focused on the modernisation of the whole statistical production procedure.
Keywords: Modernisation; value chain; Service Oriented Architecture (SOA).
1. Background Among the main Stat2015 Programme objectives there is the achievement of modernisation prerequisites, such as standard IT tools and methods, audit activities to certify and validate methods and software, change management approach with regard to ICT area (see Figure 1). Nonetheless, the most important issues are standardisation and industrialisation of production processes based on reuse (of data, methods and processes) and on the adoption of a model founded on shared services within a service-oriented architecture (SOA) framework. One of the most relevant obstacles to the success of such an ambitious project lies in the fact that Istat is presently characterised by multiple organisational models
1
(concerning financial, technological, regulatory sectors, etc.) that tend to be inconsistent.
Figure 1 – Stat2015 new model of production process Source: Our processing, 2012.
The use of different vocabularies and terminologies sometimes can lead to conflicting descriptions of the same entities. Production processes in most cases are also organised following the logic of non-integrated stovepipe models. This lack of homogeneity of language and organisational approaches makes the dynamics of change expensive and complex and draws the attention to the importance of achieving a unity of views on the current situation, so that each section of the Institute may undertake a work of innovation consistent with the objectives that should be achieved. It is therefore necessary that Istat adopts a common language in order to enable all its components both to conceptualise the given situation (“as is”) and the one to be reached at the end of the evolution process (“to be”). The description of the two states (present and future) also allows to design a path towards possible changes in a more rational and measurable way, defining specific actions involving different skills that need to interact within a shared view of a tangible progress. In this framework, this work aims at formulating a Business Architecture “to be” model that considering the lack of a European and international standard is based on established and well-experimented best practices. In particular, the BA model considered to be the most appropriate for Istat is inspired both by the Dutch Central Bureau of Statistics (CBS) (Bredero et al, 2009) and by the international standard GSBPM (Generic Statistical Business Process Model) (UNECE/Eurostat/OECD, 2009). CBS was carefully chosen after a comparative study of the different approaches of various National Institutes/Offices of Statistics. The Dutch example, indeed, has been operating for several years and proving to be very successful: it clearly identifies four main business functions which are all relevant at the governance level, splitting policy and management. It distinguishes clearly the design activities from those related to current production process. Therefore, it is compliant with one of the fundamental principles underlying process automation (or industrialisation), i.e. making the statistical process as efficient as possible. An industrialised process, indeed, is based 2
on the independence of its implementation stage from the high competence of human resources involved in its design. Afterwards, significant changes have been made to CBS model, concerning mainly: - the activities defined within the four different business functions that, when possible, have been aligned with those set out in phases and sub-processes proper of the GSBPM, recognised as a real standard and adopted at international level; - the basic principles and the general scheme, that have been customised and readapted to Istat specific situation, also taking into account the introduction of some important innovations, such as the Web 2.0 approach, and some important international initiatives, as the launch of a harmonised programme of standardisation within the whole European Statistical System. The consistency of Istat BA model with the GSBPM allows to reach a good level of compatibility also with other international standards for official statistics, including the Generic Statistical Information Model (GSIM) (UNECE – HLG-BAS, 2012). The scenario just outlined brings out the relevance of implementing an Enterprise Architecture (EA), which allows a standard approach for the representation and management of organisational changes. EA can be defined as the reference model by which an organisation operates and is structured to achieve its objectives, whereby each lower layer is governed by the higher one by means of a requirement chain and of modelling processes. In this way, it is possible to move from the level of conceptual representation to more and more operational and technological ones. In general, it is a one-way diagram, with the possibility of introducing feedback mechanisms important for the development of solutions. An interesting application example of this model to the reality of a National Statistical Institute is the one proposed by the United Nations Economic Commission for Europe (UNECE), in which: - Business Architecture, with the relative Process Model, represents a fundamental node which ensures a bridging function between the enterprise architectural level and the operational level. In this context, production processes are defined and integrated in such a way that all inputs are properly assembled to produce the required outputs; - the second node includes production processes and related methods, together with data and metadata information flows between different parts of the production process; - the last node of this conceptual scheme concerns the technological layer, i.e. systems and tools of the Information Technology (IT) which constitutes the technical solution for production process implementation.
3
Figure 2 – The Enterprise Architecture logic in a “to be” model of a National Statistical Institute Source: Our processing from UNECE - High Level Group-Business Architecture in Statistics (HLGBAS), 2012.
BA is therefore called to play a central role in a programme as complex as that of industrialisation and standardisation of statistical information production.
2. Fundamental principles Through BA it is possible to outline a conceptual scheme for a complete representation of the organisation (with its fundamental principles), useful to the achievement of change goals. The whole reference system is led by BA fundamental principles that become real guidelines for the implementation of each business domain action. In detail, the following nine different governing principles, suitably defined and addressed specifically to Istat, ensure the success of the Institute BA model: 1. the statistical process is a logical and value chain of activities clearly separated in different Business Domains (Strategy, Design, Management and Implementation), which are controlled by specifically defined rules; 2. the whole statistical process is output-driven; 3. it is necessary to standardise all statistical processes utilising SOA, so as to maximise the benefits arising from the reuse and adoption of standards contained in the Repository of standard Methods and Guidelines (RMG); 4. it is necessary to industrialise statistical processes, ensuring the independence between design and implementation; 5. no regular production activities can be conducted without having completed the design of referential metadata of model, process and rules; 6. it is necessary to clearly distinguish between the data actually processed and the metadata describing definitions, quality and process activities. The set of data and metadata represent a shared asset to be optimised;
4
7. the Repository of Data and Metadata (RDM) and the Repository of Tools and Applications (RTA) represent an Istat asset. It is necessary to exactly define its products, so as to ensure that they can be re-used, traceable, mutually related to their different versions; 8. it is necessary that the internal statistical information flow goes exclusively through the Repository of Data and Metadata (RDM) infrastructure; 9. it is necessary that the production process be implemented on the bases of generalised tools and the reuse of the applications contained in the Repository of Tools and Applications (RTA), developments from the scratch should be limited to what is not already available. In addition, infrastructure elements of the “to be” model are defined. In a nutshell, the information produced at its various stages of processing should be conceived as a common good, to be stored in the available infrastructure of the Repository of Data and Metadata (RDM). Statistical methods applied to processes should be recognised as standard and made available in the Repository of standard Methods and Guidelines (RMG). In the same way, the set of production applications of the Institute, i.e. those necessary to ensure the realisation of statistical information to be disseminated to external users, should be also considered as a common good to preserve and share, maximising the use and reuse, through the infrastructure of the Repository of Tools and Applications (RTA). This repository has to contain three distinct categories of software: - generalised IT tools; - reusable applications; - ad hoc applications, non-reusable. Shared services and standards should also be defined and implemented, with regards to: - data and metadata; - methodologies; - processes; - Information Technology (IT); - the way to organise all the activities. Once the main aspects of the above elements have been identified, the explanation of BA follows the logical sequence during which are defined: - main areas (or business domains) characterising Istat organisation; - logical chain of statistical processes; - process activities; - input and output of each activity, i.e. specific information products, depending on the context that is reflected in both data and metadata; - value chain of BA, which is also the conceptual model underlying specific developments of the other EA components, such as the technological ones related to the IT architecture.
5
3. Business domain products and activities The main step in the definition of Istat BA is to identify homogeneous areas with respect to the nature of the information processed and/or services that insist on this information. Using a language typical of this organisational approach, these areas are called business domains and are defined guaranteeing an independence from the current organisational structure of the Institute, so as to ensure their stability with regard to any future institutional processes of reorganisation. In detail, as established by fundamental principle number 1, four business domains are defined in the following: - Strategy; - Design; - Management; - Implementation. Each business domain is characterised by specific activities with a close connection with the GSBPM, except for the Strategy since it is not considered in this standard model. In particular, Strategy activities are: - S1. Maintenance and consolidation of strategic relations; - S2. Budget definition; - S3. Strategic Planning; - S4. Internal and external data source management; - S5. Policy definition for process improvement; - S6. Project portfolio management and Capacity management. Design activities are: - D1. Determine needs for statistical information (GSBPM sub-processes 1.1, 1.2, 1.3, 1.4 and 1.6); - D2. (Re)Design statistical outputs (GSBPM sub-processes 2.1 and 2.2); - D3. Check data availability/(Re)Design data sources (GSBPM sub-processes 1.5); - D4. (Re)Design production system and rules (GSBPM sub-processes 2.6 and from 2.3 to 2.5). Management activities are: - M1. Planning (GSBPM sub-processes 3.3); - M2. Monitoring (GSBPM sub-processes 9.1 and 9.2); - M3. Adjustment (GSBPM sub-processes 9.3). Implementation activities are: - I1. Tool and application reuse/development and release for the production (the whole phase 3 of GSBPM); - I2. Collect: preparatory stage (sample selection; set up collection) (GSBPM subprocesses 4.1 and 4.2); - I3. Collect: run and finalise data collection, administrative source acquisition, standardisations (GSBPM sub-processes 4.3 and 4.4); - I4. Process: integration, classification and coding, editing, imputation, new variables and statistical units derivation (GSBPM sub-processes from 5.1 to 5.5); - I5. Process: calculate weights and aggregates (GSBPM sub-processes from 5.6 to 5.8); 6
- I6. Analyse: validate, apply disclosure control and finalise outputs (the whole phase 6 of GSBPM); - I7. Disseminate: produce and release dissemination products (the whole phase 7 of GSBPM); - I8. Storage in the Repository of Data and Metadata (RDM) (the whole phase 8 of GSBPM). Each business domain supplies the necessary products for carrying out statistical processes. Strategy provides the framework (i.e. all the methods, processes and rules) for the statistical process control and organisation. Its products include: budget, regulations, agreements with other bodies/agencies, strategic planning, standards, capacity management, reports from the management (including risk management, measurements of performance, audit and quality control, Portfolio of projects). Design produces the meta-information essential for the functional organisation and for the statistical process control. Its products comprise: technical designs, action patterns, instructions, process indicators and their description. Management utilises the control information in real time. Its products embrace: the scheduling of activities, description of results, state implementation, reports on the achieved quality and plans for the improvement and adjustment of procedures carried out. Implementation realises the transition from the initial sources to the statistical information. Its products include: data archives and the metadata that describe them, as well as application tools used in the process.
4. The value chain of the Business Architecture conceptual model The statistical process represent a logically ordered chain of activities that can also be considered as a value chain, as each step increases the value of the statistical product. Business Domain activities and products are the main elements of statistical process/value chain. At a high level of abstraction, activities are processes that can be distinguished in the statistical value chain, and products represent objects. If processes are ordered in a logical way, it becomes possible to represent the process chain, which starts with the stakeholders and concludes with users (see Figure 3). In Figure 3, rectangles represent the four business domains, rhombuses are specific activities or business services (processes), cylinders symbolise information products (data and metadata) or applications, while large ovals indicate actors and external entities. The labels of each activity are indicated inside the little white ovals. Although the scheme of BA model is static, the statistical process is dynamic by nature and is realised through the iterative repetition of chain parts, which also depend on the possibility of reusing processes and/or information products. Changes to a node in the step sequence can impact on all subsequent ones, but should have no direct impact on the previous stages. Although the scheme of BA model is static, of course the statistical process nature is dynamic and is realised through the iterative repetition of chain parts, which also depend on the possibility of reusing processes and/or information products. 7
Changes to a node in the step sequence can impact on all subsequent ones, but should have no direct impact on the previous stages. The chain can be seen as composed by the union of two parts, each of which consists of two Business Domains: Strategy and Design, Management and Implementation.
Figure 3 – The Business Architecture conceptual model Source: Our processing 2012, from Statistics Netherlands BA Model and GSBPM.
Activities related to Strategy and Design are realised only the first time that the statistical product is approved and projected. The activities related to Management and Implementation are repeated on a regular basis in each processing iteration for the release of specific statistical products.
5. Future steps The BA conceptual model illustrated above allows to overcome the limits of the “as is” Istat situation: - the existing information/organisational model is very complex; - most of the statistical processes are still organised as stovepipes, i.e. vertically integrated in silos only partially communicating with other processes; this emphasises heterogeneity of procedural and methodological approaches, which is not always justified by the covered topic variety; - there is a lack of standards regarding processes, methods and technologies; 8
- there is a great redundancy of data and applications; - it is often necessary to provide ad hoc developments for the harmonisation of applications industrialised in different environments. For this reason, in this document is outlined a “to be” model of Business Architecture, which offers an integrated view of the statistical production process, that can facilitate the realisation of Istat industrialisation and modernisation. This permits the achievement of an adequate organisational flexibility, ensuring the independence between Design and Implementation, in order to make available skilled resources so as allocating them to innovations. These resources, indeed, in a not industrialised situation can run the risk of being involved in a suboptimal way, e.g. in the redesign of existing tools or in the mere repetition of current and structured processes. This becomes crucial especially in the present economic context: the acquisition of additional resources is not likely and consequently innovations can be achieved only through an efficiency retrieval. In the medium period, the BA conceptual model can be made operational through a road map properly designed and scheduled, focusing particularly on the implementation of some of the basic infrastructures provided, both in terms of procedures (such as the management of the Portfolio, the compliance assessment of the several statistical production processes with the BA and EA principles, with their subsequent validation, etc.) and in terms of shared services (the Repository of standard Methods and Guidelines, the Repository of Data and Metadata, the Repository of Tools and Applications). In this road map planning phase, a particular attention is focused on the realisation of Istat Business Architecture giving priorities to the definition of: - the governance, to provide the strategic directions, to coordinate, to validate and to monitor all the necessary activities, also involving some Istat reference committees (e.g., methodology, quality, innovation and research etc..); - the organisational structure, for the different operational phases and to advise the governance arrangements for the basic infrastructure implementation; - the communication process, to share BA principles within Istat and to disseminate their knowledge. As illustrated in the present study, the Business Architecture can provide an important support to an integrated and ordered implementation of all the necessary innovations useful to carry out the industrialisation and modernisation process. This involves necessarily an active participation of Istat and the sharing of broader initiatives aiming at increasing integration at European and international level.
References Bredero, R., Dekker, W., Huigen, R. and Renssen, R. (2009). Statistics Netherlands Architecture; Context of change. Discussion paper (09017), CBS Statistics Netherlands, The Hague Heerlen. Commission of the European Communities (2009). Communication from the Commission to the European Parliament and the Council on the Production method of EU statistics: a vision for the next decade.
9
Sponsorship on Standardisation (2012). Progress report, Theme 6.10, 15th Meeting of the European Statistical System Committee, Luxembourg, 15th November. Statistics Netherland (2012). Strategy to implement the vision of the High-level Group for Strategic Developments in Business Architecture in Statistics, Conference of European Statisticians, Sixtieth plenary session, Paris, 6-8 June, Item 8 of the provisional agenda. UNECE (2011). Some ideas on standardisation by layers in official statistics, http://www1.unece.org/stat/platform/display/hlgbas/Some+ideas+on+standardisati on+by+layers+in+official+statistics. UNECE/Eurostat/OECD (2009). Generic Statistical Business Process Model (GSBPM Version 4.0), Work Session on Statistical Metadata (METIS, April 2009). UNECE, High-Level Group for Strategic Developments in Business Architecture in Statistics (HLG-BAS) (2012). Generic Statistical Information Model (GSIM) http://www1.unece.org/stat/platform/display/metis/Generic+Statistical+Informatio n+Model+(GSIM).
10
