Experiences from the Australian Bureau of Statistics (ABS) Generic Statistical Business Process Model (GSBPM) and its contribution to modelling business processes Experiences from the Australian Bureau of Statistics (ABS) Trevor Sutton

Outline “Industrialisation” and the need for a strategic focus on statistical business processes Introduction to GSBPM as a reference model Recent developments related to the GSBPM Practical ABS experiences in applying the GSBPM Some other high level considerations when modelling business processes Questions First section provides the link back to Brian Pink’s keynote talk about HLG-BAS, industrialisation etc.

Strategic Context HLG-BAS Strategic Vision Industrialisation includes: We have to re-invent our products and processes and adapt to a changed world Industrialisation includes: Common processes Common methodologies Common tools Facilitating commonality through agreeing and applying “industry” frameworks and standards Recognizing all statistics are produced in similar ways A common language (eg common reference model) for identifying and describing processes is essential to supporting industrialisation (including standardisation, harmonisation and collaboration). GSBPM is an example of an agreed “industry” framework (in this case in regard to statistical business processes) for facilitating commonality.

Implications in regard to process We must be able to review our statistical business processes (SBPs) at a strategic level in order to determine their fitness for purpose & value add improve, integrate, reuse, transform, industrialise, standardise, harmonise Each SBP must be described (including modelled) in a manner which facilitates comparison with other SBPs (locally and internationally) In order to facilitate standardisation and reuse, SBPs should be described independently of the statistical methods and IT tools currently used to perform them As per definitions used in the GSBPM, a “statistical business process” is a set of activities undertaken by a producer of official statistics that results in data outputs. Statistical business processes consist of phases (eg design, collect, process, disseminate) and sub-processes within each phase. For example, the business process is imputation, the method is historical unit or hot deck imputation and the IT tools is SAS Or the business process is apply disclosure control, the methods are perturbation or special uniques analysis, the IT tool is SAS or SUDA (Special Unique Detection Algorithm)

The following slides explore, in a simplified manner, the relationship between the GSBPM the objective of harmonising statistical business processes, and Industrialization of methods and technologies at the practical level Information on GSIM GSIM is a reference model (still under development, rather than mature like GSBPM) which is focused on the statistical information objects which are used and/or produced by statistical production processes. It defines the information objects (e.g. concepts, classifications, variables) required to drive those processes, as well as the outputs (e.g. statistical data) and the outcomes (e.g. process metrics) from them. GSIM aims to standardise terminology and modelling regarding statistical information. This will assist organisations to develop statistical information management solutions (e.g. systems and frameworks) and facilitate interagency understanding and collaboration. When operationalized, GSIM will support interoperability between new methodologies and technical developments within the industry of producers of official statistics. GSIM will not attempt to harmonize the definitions of instances of the information objects (e.g. Person or Household - instances of Statistical Unit) but will provide the structure to allow their comparison.

Harmonised Statistical Business Processes The design of each particular statistical business process (eg production of quarterly national accounts) is shaped by relevant conceptual frameworks (eg system of national accounts, analyses of user requirements and priorities for meeting these requirements). The result of the design is a business process which is able to operate in practice on a sustainable and reliable basis uses fit for purpose methods and technologies produces the practical outputs (products) that are required Statistical business processes therefore bring together conceptual and practical aspects and join the top half and the bottom half of the diagram together.

Harmonised Statistical Business Processes Provides a common categorisation and set of terminology for describing/defining statistical business processes Harmonised Statistical Business Processes

Harmonised Statistical Business Processes Provides a common categorisation and set of terminology for describing/defining statistical business processes Harmonised Statistical Business Processes Drive consistent business requirements for industrialised methods and technologies

Introducing GSBPM Originally based on the business process model developed by Statistics New Zealand Three rounds of international consultation led by CES Steering Group on Statistical Metadata (also known at the METIS Steering Group) Added Archive and Evaluate phases Terminology and descriptions made more generic Currently Version 4.0 The one page diagram is most often used but complete documentation (in three languages) and other resources are available on the web. CES=Conference of European Statisticians Languages : English, French, Russian Other resources : Presentations, papers, a discussion forum

Why do we need GSBPM? To define, describe and map statistical processes in a coherent way To standardize process terminology To compare / benchmark processes within and between organisations This facilitates collaboration. To identify synergies between processes To inform decisions on systems architectures and organisation of resources GSBPM does not attempt to go to a level of detail sufficient to fully define statistical processes as actually implemented by NSIs. Nevertheless, when documenting detailed definitions of actual processes, GSBPM provides a common point of reference and a common set of terminology. 10

Structure of the Model Process Phases Sub-processes (Descriptions) (statistical business process) Phases Sub-processes (Descriptions) (Descriptions) : Descriptions are not shown in the diagram but within the detailed documentation there is a description of each sub process. National implementations may need additional levels of detail The documentation recognises over-arching statistical processes (The first two are most closely related to the model, and are therefore shown at the top of the diagram also) Quality management Metadata management Statistical framework management Statistical programme management Knowledge management Data management (process independent) Process data management Provider management More general over-arching processes include: · Human resource management; · Financial management; · Project management; · Legal framework management; · Organizational framework management; · Strategic planning.

Applicability All activities undertaken by producers of official statistics which result in data outputs Producing statistics from raw data (micro or macro-data) Revision of existing data / re-calculation of time-series Development and maintenance of statistical registers A set of activities that fulfills these conditions is termed a statistical business process National and international statistical organisations Independent of data source, can be used for: Surveys / censuses Administrative sources / register-based statistics Mixed sources

Not a linear model Sub-processes do not have to be followed in a strict order It is a matrix, through which there are many possible paths, including iterative loops within and between phases Some iterations of a regular process may skip certain sub-processes It is not a blue print for your organisation For example, a sample survey may implement different processes to a survey using mixed sources (i.e admin plus survey data)

Recent Developments : Modelling business processes beyond the scope of GSBPM This is an example from Korea. Eurostat have a similar reference model spanning business processes beyond GSBPM. The HLG-BAS Workshop in November concluded that more work should be done in this field. Taking one step further back, it would be possible to develop a reference model showing how processes undertaken by NSIs relate to processes undertaken by data providers and data consumers. (A NSI is not a closed system.) Opinions vary, but a common perspective is that the need to model other NSI business processes does not imply GSBPM needs to “get bigger”. Other layers of modelling could complement GSBPM, leaving it fit for its current purpose.

Outcomes from METIS Workshop Held 5 – 7 October 2011 The GSBPM will not be revised in the short term Future work will focus on work on data and metadata flows in GSBPM METIS is the UNECE group on Statistical Metadata METIS “owns” GSBPM They have a workshop ever 12 – 18 months, the most recent one was focused on GSBPM

Applying GSBPM to an NSI GSBPM is a reference model, which has been be used in a number of ways: Agency adopts it “as is” Agency adopts a version of it Agency maps existing process model to it For more information see National Implementations of GSBPM

Applying GSBPM in ABS ABS sees the GSBPM as a cornerstone for a more generic reference architecture. It can be utilised to facilitate and enhance communication and understanding, and ultimately, sharing and collaboration across agencies Bottom left : Six NSIs mapped their high priority needs for new processing solutions to the GSBPM, helping identify common needs for collaboration purposes. Bottom right : A map from the Sharing Advisory Board (a collaboration facilitation group sponsored by the UNECE) showing existing collaborative projects mapped to the phases of the GSBPM that they address.

Applying GSBPM in ABS (2) GSBPM was formally adopted by the ABS in 2010. It is our primary reference model for statistical business processes. It is used in corporate planning and as a cornerstone of ABS Enterprise Architecture. Corporate planning use includes activity distribution survey and Strategic Investment Committee

Applying GSBPM in ABS (3) Early adopter of GSBPM in ABS was the Prices System Improvement Project. Project aims to design an end-to-end system for 5 Price Indexes. GSBPM was used as a guide to harmonise processes across the 5 Price Indexes

Applying GSBPM in ABS (4) In February 2010, the ABS announced the Information Management Transformation Program. A key element of this Program is business process transformation. The approach is to have workshops: to analyse and map a range of current collection processes to develop aspects of the "to be" environment from a functional, end to end perspective These activities will utilise GSBPM as a reference model.

Applying GSBPM in ABS (5) Create common frame Select sample Common Frame Create survey frame QEWS Frame Forms design and approval Forms Load sample to PIMS Label files Dispatch Significance Editing Time series analyses NAB and FAS sign off Time series to PPW Collection and IFU Data Paradata (Collection Information) Clean Data Time Series Databases Published Data Business Process Steps Business Output and Input Artefacts Derivation Processes Sample MRR Proof of Concept 2010/11 Core case study was elements of statistical business process for Quarterly Business Indicators Survey (QBIS) Part of IMTP project is to build a POC to demonstrate the infrastructure that is being developed by other parts of the Program. The statement of purpose for the POC is to "Establish fundamental supporting infrastructure to manage metadata in a coherent manner; to make this metadata available for active use and reuse; and to drive processes and systems across the ABS and beyond." Using GSBPM as a framework for the statistical production cycle, the POC demonstrates the benefits of the Metadata Registry and Repository (MRR) infrastructure which registers and stores metadata used in the statistical production process. - GSBPM gives us a generic set of business processes that go in to the statistical cycle - So, for any collection we should be able to map (at least at a high level) our statistical business processes to the GSBPM - Here we have some high level processes for QBIS, and as we move through the statistical cycle we can map our processes to the GSBPM The POC has implemented a Statistical Workflow Management System (SWMs) to ensure efficient end to end flows of statistical metadata within the statistical production process. These metadata driven processes are managed by the SWMs, and the metadata inputs and outputs to the GSBPM based processes are registered and stored by the MRR. This enables the discovery and reuse of processes and metadata, and ensures the efficient flow of metadata throughout the statistical business process.

Summary of experiences Very valuable as a common reference model facilitating comparability within & across NSIs Use as a consistent high level reference model for statistical business process, eg framework & context when presenting training about statistical production processes tracking resources (eg staff effort and other costs) directly related to statistical production useful when designing quality management for the statistical production process (eg positioning quality gates) useful point of reference when cataloguing, assessing and managing various methods and IT systems available to support statistical production Examples of its use as a high level reference model listed here are in addition to those described in previous slides.

Summary of experiences (2) Staff can be unsure about intent Not a template for designing statistical business processes Details of processes and workflows as implemented in practice are less generic Not a blueprint for the “ideal” statistical business process It provides some value as a reference model for all statistical business processes, but value tends to be greater for some types than others eg a better (and more obvious) fit for “traditional” business and household surveys vs compilations (eg National Accounts) and processes using administrative sources Need to keep its scope in mind Don’t try to use it as a reference for business processes that don’t fit the criteria Eg when modelling the process an NSI uses to determine human resource needs and recruit/train staff accordingly

Other high level considerations when modelling business processes Ensure the roles of GSBPM as a reference model are understood. Actual business processes often do not map simply to the GSBPM. In these cases, document relationships between the process as modelled and the GSBPM Do not simplify modelling of business processes simply to better align them with the GSBPM The best approach is a partnership between business staff and staff expert at analysing and modelling business processes (a centre of excellence). Don’t expect statistical business staff to produce consistent, high quality models on their own (but must include them) Don’t rely only on IT modelling skills Must clearly separate “As Is” and “To Be” modelling of business processes both are usually important there may be changes to process (eg to move to a process that current methods and/or IT cannot support) there are very likely to be changes to methods and/or technology used to implement processes GSBPM is one small but important element of a framework and toolkit for modelling statistical business processes.

Other considerations (2) Ensure practical benefits of investing in modelling is apparent to business areas Start with well defined plans for using and maintaining the information, not just for gathering it The case is strong where modelling inputs directly to business process re-engineering which delivers greater levels of automation and reliability and simplifies change Carefully select methods and tools used for modelling. Consider factors such as ease of use integration with software used for related purposes the simplest solution which is fit for purpose “powerful” & “advanced” is not always best!

Questions?