Statistics Finland: use of GSBPM

Statistical business process model

In 2009, Statistics Finland decided to implement a customized model based on level 1 of the GSBPM v4.0. However, in March 2015, Statistics Finland adopted GSBPM v.5.0 as the model for statistics production processes as such.

Due to the improvements in GSBPM v5.0 and the need to easily communicate our practices, customizations were no longer seen as necessary. The following two pictures illustrate our process structure.

GSBPM Implementation

Since 2009, the GSBPM has mainly been used by management as a general description of the statistical process. After that the model has gained traction in project work.  From spring 2015 onwards, the model and its use has been communicated to the entire staff.

In practice, we have produced an internal document describing the GSBPM in the context of Statistics Finland. For example, a distinction between different types of individual production processes (i.e. logical processes) and the aggregate production process encompassing them (i.e. abstract process) is made. Furthermore, different modes of data collection - and to some degree, also different strategies in processing - are considered as different logical sub-processes.

The GSBPM is being utilized in

• Time usage and expenditure monitoring. In statistics production, working hours and direct costs are allocated to roughly 70 statistical processes and process steps (closely related to level 1 of the GSBPM).
• Project planning. For proposed projects, we identify the process steps involved in the proposal. This information is useful in improving the focus of the project and to identify common or similar tasks in different proposals, possibly leading to combining or otherwise changing the structure of the projects.
• Project work. Current practices and processes to-be are described according to the GSBPM. Storyboarding-techniques have been proven to work well with the GSBPM.
• In work documentation. The GSBPM is used as a common structure and vocabulary. The goal is to improve the usability of working documents for new personnel and persons outside the particular statistical domain.
• Development of information architecture. The GSBPM is a useful starting point in defining the logical information architecture, i.e. how classifications, registers, samples, collected data, processed data, analysed data and the results are stored.
• In IT development to identify common services and components. Our goal, as in perhaps in all NSI’s, is to gradually move from domain-specific architectures to a service oriented architecture.

We are planning to describe standard or template work processes for different types of statistical processes. This would help in standardizing processes and in reducing the time needed for process modelling in projects.

Enterprise architecture and the GSBPM

Process orientation and strategic management are hard to implement if the managerial roles are structured entirely around functions and statistical domains. This is clearly visible in project planning, where domain-specific orientation and standardization are often in conflict.

In March 2015, Statistics Finland adopted an enterprise architecture oriented approach in order to gain more strategic control of the development of new services, processes, technologies etc. The approach is based on process management for the statistical process with portfolio management for customer services, respondent services and ICT. Five key responsibilities were created:

• Architecture owner (Deputy Director General)
• Process owner (Deputy Director General of Statistics Production)
• Customer portfolio owner (Director of Communications and Information Services)
• Respondent portfolio owner (Director of Data Collection)
• ICT portfolio owner (Director of Information Technology)

We expect the new roles to result in improved customer focus and standardization.

Metadata are used to inform the users of data what the statistics describe and how they are realised. With data descriptions, classifications and definitions of variables it is ensured that statistics are mutually comparable. Quality and methodological descriptions help the user to assess the reliability of the statistics and their suitability to various use purposes.

At the same time, metadata are needed in order to organise production of statistics in an efficient way. Timetables, data concerning the state of the data set and supplementary notes to the data set are necessary in interaction between those involved in the statistics production. Process indicators, such as data concerning the effectiveness of editing, help to focus measures on the right matters.

In the production processes, the same metadata have to be used again in different process stages. For example, variable descriptions needed in table headings can be produced by inheriting them from those descriptions that were made in data planning and that have possibly been updated in the course of the production process.

Persons involved in the production of the statistics at the practical level do not often even realise they are working with metadata. Metadata are processed similarly in the course of work as other information. Metadata exist but they are not necessarily managed and described in the best possible way, so they cannot be exploited to the full.

In future development the importance of metadata and its application will be stressed in every phase of statistical processes. At present, there are information systems which use both statistical and process metadata in all the phases of data production, but it is by no means standard in data production. For example, annual census data are processed by using metadata in every production phase from data collection to data dissemination. As a whole, metadata are mostly used in the data dissemination phase. That is the area where the development of metadata systems has been stressed also because of the wide use of the Internet as a data dissemination medium.

There are of course business processes which help the budgeting and the follow-up of the yearly action plans of the statistical units also including the action plan of the statistical office itself.

In personnel administration there are various business processes for managing human resources and their optimal allocation to various projects. For instance, time allocation systems are in place, consisting of work time allocation to different tasks or projects, overseeing absences, etc. (sick leaves, holidays).

These business systems include a great amount of metadata about the projects and personnel. At present, they are not as optimally integrated into statistical process management as they could be, but that could be remedied in the future.