United Nations Economic Commission for EuropeStatistical DivisionUnited Nations Economic Commission for EuropeStatistical Division

Introducing the GSBPM

Steven ValeUNECE

steven.vale@unece.org

Background

Statistical production has traditionally been organised by topic, e.g. transport, trade, …

Financial pressures are encouraging new ways of thinking

Some statistical organisations are moving towards a process-based approach

Others are considering a matrix approach

Topics

P r o c e s s e s

Background

Defining and mapping business processes in statistical organisations started at least 10 years ago• “Statistical value chain” X• “Survey life-cycle” X• “Statistical process cycle” X• “Business process model” X

Generic Statistical BusinessProcess Model

What is the GSBPM?

A flexible model that describes and defines the set of business processes needed to produce official statistics

Standard framework and harmonised terminology to help statistical organisations• Modernise statistical production processes• Share methods and components

Who created the GSBPM?

The Conference of European Statisticians Steering Group on Statistical Metadata

Part of the Common Metadata Framework• www.unece.org/stats/cmf

How was it created?

Based on the business process model developed by Statistics New Zealand

Added phases for:• Archive (inspired by Statistics Canada)• Evaluate (Australia and others)

Three rounds of comments Terminology and descriptions made

more generic

Result: GSBPM Version 4.0 Released April 2009

Updating the GSBPM

By 2013 many organisations had adopted GSBPM – and given feedback

Related new models and standards had been developed

Agreement that GSBPM needed to be “refreshed” to maintain relevance

Changes between v4 and v5

Policy of minimum change• Only if wide agreement and good business case

Removed Phase 8 (Archive)• Archiving happens throughout processes

Improved documentation• Less survey-focused• Links to newer standards e.g. GSIM• More detail where requested by users

Result: GSBPM Version 5.0 Released December 2013

Why do we need the GSBPM?

To define and describe statistical processes in a coherent way

To compare and benchmark processes within and between organisations

To make better decisions on production systems and organisation of resources

What will this mean for resources?

Applicability (1)

All activities undertaken by producers of official statistics which result in data outputs

All statistical domains National and international statistical

organisations

Applicability (2)

Development and maintenance of statistical registers

All types of data source:• Surveys / censuses• Administrative sources / register-based

statistics• Mixed sources• “Big Data”

Structure of the Model (1)

Process

Phases

Sub-processes

(Descriptions)

Structure of the Model (2)

National implementations may need additional levels

Over-arching processes• Quality management• Metadata management• Statistical framework management• Statistical programme management• ........ (others – see paper)

Key features

Not a linear model Sub-processes are not followed in a strict

order It is a matrix, through which there are

many possible paths

Key Features (2)

In theory the model is circular:• Evaluation can lead to

modified needs and design In practice it is more like

a multiple helix:• There may be several

iterations of a process underway at any point in time

Uses of the GSBPM

Managing statistical programmes Cost / resource allocation Documenting statistical processes Framework for quality assessment Sharing statistical software

Example 1

Australian project to improve production of prices data

Identified activities within GSBPM sub-processes

Some phases “out of scope”

Example 2

Another price index, this time for legal services in Norway

Note the difference in phases / sub-processes covered• E.g. no data integration or classification in

this process

5.8Finalise data files

5.6Calculate weights

Price index for legal services

Quality Management / Metadata ManagementSpecify Needs

Design Build Collect Process Analyse Disseminate

1.1Determine needs for information

• Eurostat regulation• Ministry of Justice• Trade association• Major companies

1.2Consult & confirm needs

• See 1.1

1.6Prepare business case

1.4Identify concepts

1.3Establish output objectives

1.5Check data availability

• Provided in regulation

• Variables• Classifications

• Project plan• Finances

• Choose companies (population) together with trade association

2.2Design variable descriptions

• See Eurostat regulation• Agree on level of detail• Translation

2.6Design production systems & workflow

2.4Design frame & sample methodology

2.3Design data collection methodology

2.5Design statistical processing methodology

• Paper & electronic questionnaires• Scanning

• Establish population• Make sample plan• Stratification by size

• Scanning paper questionnaires• Electronic data capture

• Micro controls• Weighting• Estimation

3.1Build data collection instrument

• Standard first page• Test paper & electronic

3.2Build or enhance process components

• Integrated system for editing and estimation• Sample register

3.6Finalise production system

3.4Test production systems

3.3Configure workflows

3.5Test statistical business process

• Meetings with the various data capture divisions

• Test micro-controls

• Documentation• Training• Meetings with 1st and 2nd line support

• No field test or pilotcarried out

4.1Select sample

• Companies selectedaccording to sample plan

4.2Set up collection

• Time plan• Preprinting, packing, sending, scanning etc.

4.4Finalise collection

4.3Run collection

• Packing & sending• Receive electronic data• Scan papir• Update received reg.• Send warnings & fines

5.1Integrate data

• No integration carried out

5.2Classify and code

• No classifying or coding carried out

5.4Impute

5.3Review, validate & edit

5.5Derive new variables & statistical units

• Duplicates• Regional and product views• Discover outliers• Correct microdata

• No imputation carried out

• Weights are established in phase 2• Weights are stored per survey

• Publication variables• Observation units = analysis units

6.1Prepare draft outputs

• Index & 4 sub-indices• Confidence intervals• Coefficient of variation• Tables & graphs

6.2Validate outputs

• Check most important companies have replied. Iterative with 4.3• Check variation coefficient.

6.4Apply disclosure control

6.3Scrutinize & explain

6.5Finalize outputs

• Consult trade association.• Trade trends• Any previous periods• Other price indexes

• Aggregation & publication level

• Final consistency checks• Supporting internal documentation

7.1Update output systems

• Load data into StatBank

7.2Produce dissemination products

• Daily statistics• About the statistics• Tables & graphs

7.4Promote dissemination products

7.3Manage release of dissemination products

7.5Manage user support

• Statistical calender• Approval routines• ssb.no

• Daily statistics

• Web statistics• Mail or telephone • Contact 1.1 users

• Data from electronic &paper questionnaires• Deliver data to processing

5.7Calculate aggregates

• Data files are finalised

2.1Design outputs

General dissemination policy:• About the statistics• Tables in StatBank• Text & graphs

• Aggregates are calculated

Example 3

Irish tourism survey Mapped GSBPM to 78 survey process

steps

Example 4

Denmark – quarterly survey on employment in construction

Established, regular survey• Just phases 4 to 7

GSBPM as a Framework for Documentation

“The GSBPM provides a structure for organizing documentation, promoting standardisation and the identification of good practices”

GSBPM v5.0

February February 2014 – 2014 – pilot surveys descriptionpilot surveys description

Labour statistics Industry Statistics

10-12 June 2014, Nizhny Novgorod, Russia

Belarus: Using GSBPM 5.0 to describe the existing statistical production processes

Results:

Identification of gapsgaps in the existing processes

10-12 June 2014, Nizhny Novgorod, Russia

LackLack of necessary documentation

Existence of unsettled unsettled processes

The GSBPM is used by more than

50 statistical organisations

worldwide

KSBPM – Republic of Korea

Beyond statistics: Data archivesGeneric Longitudinal Business Process Model

The future of the GSBPM

The GSBPM will have to change over time to reflect new realities• But not too often!

Collect ideas and feedback to inform future revisions

More detail? Broader scope?

Detail vs Generic - Norway

Detail vs Generic - Thailand

GSBPM+ GAMSO

Generic Activity Model for Statistical Organisations

More information

GSBPM Wiki

www1.unece.org/stat/platform/display/GSBPM