Collecting online: How Statistics Netherlands moved its business Surveys onto the internet Ger Snijkers thanks to Deirdre Giesen, Hank Hermans, Myra Wieling Overview 1. Overview of developments at Statistics Netherlands: a. Electronic data collection b. Organising data collection and data sharing 2. Case study: Moving the Structural Business Survey to the web a. Developing and testing the questionnaire b. Communication strategy to launch the survey and improving web-take up rates c. Implementation of the survey Lessons learned and challenges 3. Looking into the future: future challenges 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 2 Brief history of electronic data collection at SN Before 2000: • 1984 First PCs • Mid 1980s Blaise 1.0 • 1993 E-reporting for International Trade: IRIS - disk-by-mail (floppy disks, CDs), now offline using downloadable software • Mid 1990s Electronic Data Interchange: EDIsent - • 1998 Simple establishment surveys - 15 October 2013 software installed on business’ systems e-mail OSS Seminar, Statistics New Zealand, Wellington 3 E-mail questionnaire Business Tendency Survey 1999 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 4 Brief history of electronic data collection at SN (cont) After 2000: • 2000 Short Term Statistics (STS) - • 2003 • 2006 • 2006 Transportation Survey STS – New HTML forms Structural Business Surveys - 2007 first electronic forms (HTML) Pilot offline Structural Business Survey in the field - • 2009 • 2014 offline downloadable software Business Surveys online: Transportation Survey • Next - New integrated system for online surveys - Blaise 5 EDI using XBRL integrated in questionnaires • 2005 • 2008 Experiments for household surveys (on-line) Household surveys on the internet (on-line) 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 5 Transportation Survey 2003 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 6 Offline SBS Survey 2007 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington more to come 7 Online Transportation Survey 2009 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 8 XBRL-based system (future) 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 9 Health Survey 2010 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 10 Business Data Collection at the moment A few numbers: • • • • • 450 different questionnaires 400 mixed-mode: electronic & paper > 500,000 business units every year > 750,000 questionnaires on average each year 700,000 electronic questionnaires available 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 11 Drivers for using Web for Business Surveys • From the data collection agency perspective: - data quality: cleaner (fewer edits) and more complete cost-efficiency: cheaper and less editing timeliness: faster • From the business perspective: - internet penetration for businesses is high much of the data held by businesses is already in digital form many of the regulatory and administrative tasks they perform are already online can be easier to do than paper (if designed well) This only holds is surveys are designed well! - the whole communication process: from end to end 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 12 Conclusions so far Until 2007: • Each survey for its own: stove-pipe approach • No coordinated/harmonised design • Have it in the field quickly • Focus on technology, not on methodology/communication No efficient questionnaire design and communication: • Internally: every survey starts all over, no learning curve • Externally: no professional questionnaire and survey communication: - various designs - respondents have to learn every questionnaire again 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 13 Lessons learned • Analyse contextual factors: - Internal constraints and processes Response processes within businesses • Focus on usability = success factor: - internally AND externally technology AND methodology perspective “We suspect that many of the survey organisations that introduce web questionnaires forget that it is not the technology in itself, but how it is utilised that determines the result” ... both internally and within businesses From: Haraldsen & Couper (2013) • Apply Deming cycle – PDCA: 15 October 2013 also C & A: Evaluate and Improve OSS Seminar, Statistics New Zealand, Wellington 14 Lessons learned (cont’d) • Good design is to create benefits for both sides: - 15 October 2013 Create internal conditions to be successful (under control) > One Data Collection Division Tailor to business conditions (not under control; try to influence) > Moving SBS to the Web OSS Seminar, Statistics New Zealand, Wellington 15 Overview 1. Overview of developments at Statistics Netherlands: a. Electronic data collection b. Organising data collection and data sharing 2. Case study: Moving the Structural Business Survey to the web a. Developing and testing the questionnaire b. Communication strategy to launch the survey and improving web-take up rates c. Implementation of the survey Lessons learned and challenges 3. Looking into the future: future challenges 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 16 Data Collection over the years < 1994 1994 2000 Stove pipes Data Collection 2000 2000 2007 Present Partial Centralisation 15 October 2013 Data Collection Now OSS Seminar, Statistics New Zealand, Wellington 17 Organizing Data Collection in one Division Process-and-knowledge Driven approach Clustering of practices and knowledge • Focus on how knowledge is being used, for social and business surveys • Pilot 2007; effectuated 2011 Front desk: 12 fte • Client relations • Survey design Design: 35 fte • Questionnaire • Sampling • Training 15 October 2013 Survey deployment: 240 fte • CAPI / CATI Interviewing 92 26 118 fte • Planning & control • Fieldwork logistics • Support • Special services 122 fte OSS Seminar, Statistics New Zealand, Wellington • Fte’s: - 375 (2008) - 277 (2012) • Budget: € 18.5M • 145 surveys 18 Drivers for centralizing data collection Internal drivers • Achieving efficiency by: ∘ Abolishing redundant processes, workflows, and activities for social and business surveys - Monitoring processes, workflows, and activities ∘ Abolishing redundant systems and tools - One data collection management system Maintaining as less systems and tools as needed • Implementation of Data Collection Strategy (2005) External drivers: • Professional external focus • Reducing response burden 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 19 The 2011 Data Collection Strategy: Retrieving and returning (first version in 2005) 3 steps: 1. Re-use of available data - 2. Use of new registers and other secondary sources - 3. Multi-source designs Traditional government-based registers Big/organic data Data on the internet (web-crawlers) Primary data collection: 1. 2. 3. • • 15 October 2013 Data sharing & data warehousing EDI technologies, like XBRL Web surveys traditional modes: paper, CATI, CAPI Using new technologies Reciprocity: report back to respondents OSS Seminar, Statistics New Zealand, Wellington Mixed-mode designs 20 Data collection: Developments over the years Summary and conclusions • From stove-pipes to a coordinated system of data collection • From uni-mode surveys to multi-source/mixed-mode data collection designs • From single-survey managers to managers of integrated sets of statistics • From local decision making to corporate decision making • From instrumental data collection to a communication perspective We are still in the middle of this process! A cultural shift needs years to change! But … are we moving too slow? 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 21 Overview 1. Overview of developments at Statistics Netherlands: a. Electronic data collection b. Organising data collection and data sharing 2. Case study: Moving the Structural Business Survey to the web a. Developing and testing the questionnaire b. Communication strategy to launch the survey and improving web-take up rates c. Implementation of the survey Lessons learned and challenges 3. Looking into the future: future challenges 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 22 Moving to the Web: two basic approaches • Copy all paper questionnaires to the web in a short period of time, in one or various ways: - PDF, offline electronic questionnaire, online questionnaire Quick results, but how is the questionnaire utilised? High risks of: - web Qs not being used - negative effects on public image Pushing problems and costs downstream Evaluate and improve, if you still get the chance ... • Develop one approach and apply to all surveys: - investing in the organisation, technology, and methodology investing in the ‘respondent experience’: survey communication investing in PR: harmonised approach Takes time and costs money Higher chances of good results: web Qs actually used 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 23 Moving the SBS questionnaire to the web: approach Approach by Stats Neth.: 2nd approach • Develop one approach: - Good results for SBS … No systematic evaluations and improvements • that is applied to all surveys: Some follow-up, but not consistent, due to: - organisation not yet ready (one of the main factors for project failure) - still existing stove-pipes Needs a long-term vision that is followed up Assurance that investments are not lost Do it right: create benefits for both sides (or don’t do it all)! It is not going to be easy! 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 24 Moving the SBS questionnaire to the web: Project Project goals and constraints: • Develop a web questionnaire - same contents mixed-mode design: coherent paper and web support completion process motivate respondents to use this mode into the field: March 2006 • Start of project: June 2004 - 15 October 2013 we had one year, until Sept. 2005 OSS Seminar, Statistics New Zealand, Wellington 25 Dutch Annual Structural Business Survey Some characteristics: • Annual survey of economic activity, three parts: - revenues and costs summary of business accounts: profits and losses industry specific specifications • Mandatory • 75.000 business each year: - Sample of small businesses, bigger businesses each year • Until 2006: paper questionnaire (A4 booklet): - 15 October 2013 Length may differ, 20 pages is typical Items are grouped in sections Completion process: complicated and difficult: lots of details asked for, mismatch of definitions, various business departments involved, design issues OSS Seminar, Statistics New Zealand, Wellington 26 Vernieuwde huisstijl: 2. Onderzoek naar de PS –Q oud 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 27 Vernieuwde huisstijl: 2. Onderzoek naar de PS –Q oud 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 28 3. 4 projects 3. Improving paper questionnaire –Q nieuw 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 29 Developing the electronic questionnaire Developing and testing in five stages: 1. Develop and test the prototype - pre-tests to test usability: 3 waves field agents, business respondents 2. Revision of questionnaire - (1-1-’06) additional usability tests 4. Implementation of field pilot 5. Implementation of survey 15 October 2013 (1-9-’05) expert reviews 3. Testing of revised questionnaire - (31-1-’05) OSS Seminar, Statistics New Zealand, Wellington (1-3-’06) (1-3-’07) 30 The electronic questionnaire: the prototype Developing paper Q and testing this prototype helped in thinking about a good design for an eQ 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 31 Testing the prototype Research issues: 1. How does the e-Q work in practice? - Completing the questionnaire Question-and-answer process Usability 2. What features should be included to make it easy to use? - Respondent friendly: ‘Computer-assisted’ tools User demands 3. How should the eQ be designed in relation to the paper Q? - 15 October 2013 The same or a different design ‘look-and-feel’ of paper and e-form OSS Seminar, Statistics New Zealand, Wellington 32 Testing the prototype: Research issues 1. How does the Q work? Laborious and complex process • Long, complex questionnaire (≥ 25 items) • Complex completion process: - several sessions, several informants kick-and-rush behaviour • Imagine ... a respondent sitting behind his/her computer ... Respondent got lost in the questionnaire 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 33 Illustration of usability “Medieval tech support” A new system is introduced: The ‘book’ (“beek”), in stead of rolls of paper. The monk Ansgard has called the helpdesk and is waiting for someone to come … 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 34 Illustration of usability “Medieval tech support” It illustrates what we found in the pre-tests for the electronic questionnaire … 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 35 Testing the prototype: Research issues 2. Features to make it easy to use? • What am I supposed to do (next)? - • How is the questionnaire built up? - • Easy to download, install, complete, send data back It is one process: going to the internet – submitting data, Clear instructions and explanations (but not read) Show how the questionnaire is structured: overview Help to find the way in the questionnaire No hidden rules, no unexpected functionalities Where am I? What did I do so far? - Provide overview of the completion process Clear navigation, no scrolling Printing function Usability, navigation and overview: design of Tax forms Choose a design people know: don’t be original! 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 36 Testing the prototype: Research issues 3. Design of paper and web Q? The computer is different than paper • The web Q reacts to the respondent • Reading from the screen is different • Navigating and getting an overview works differently • Kick-and-rush behaviour, even stronger than on paper • The use of computer-assisted functionalities; • the respondent expects the computer to react Start with paper, then add functionalities 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 37 Testing the prototype: Conclusions of pre-test waves • Visual design Clear and logical: “Don’t make me think!” Simple, transparent, consistent No hidden and unexpected functionalities • Support the completion process, end to end Other mode, other features, other visual design Use computer functionalities: automatic calculations Use layout and usability principles that are known to R (country specific) • Tailor to kick-and-rush behaviour Split Q into small sections, small tasks Short and clear explanations 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 38 Lessons learned: Web Design Guidelines Web questionnaire design is communication design Questionnaire designer = Communication designer Questionnaire communication design 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 39 The revised questionnaire Based on: • Pre-test results • Expert reviews Iterative process with - Professional designer - Questionnaire designers - Methodologists • A user-friendly design was put first, not the IT tool • New prototypes designed in Power Point 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 40 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 41 The revised questionnaire: additional pretests • 10 concurrent in-depth interviews • Usability and user friendliness had been improved respondents enjoyed working with the questionnaire they could handle the task: they recognised the design and usability even though … the task had not changed 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 42 Field pilot Goals: • Implementation of web questionnaire - - test usability and completion process in the field: downloading – completing – sending-in data test communication strategy to get high web take-up rates • Test the whole process - logistics for paper and web flows • Response rates and web take-up rates? • Data quality - 15 October 2013 data editing, mode effects? OSS Seminar, Statistics New Zealand, Wellington 43 Field pilot Set-up: • March-July 2006 • 7200 businesses, 5 industries • Communication strategy: - internet portal: www.cbs.nl/productiestatistiek advance letter with user name and password leaflet to introduce web questionnaire and explain why the survey is conducted - paper questionnaire not mentioned - only 2e or 3e reminding letters include paper questionnaire In the Netherlands businesses have to report electronically to the Tax Office. 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 44 Leaflet 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 45 Field pilot Results: • Web questionnaire worked well • Web take-up rate: 80%, total response rate: comparable • No mode effects, but more research is needed • Internal processes worked well, but we missed checking the link to data editing. SBS questionnaire and communication design was implemented by the end of 2006, and run in 2007 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 46 Communication strategy: guidelines How to increase web take-up rate: • Easy access to the web version combined with restricted access to the paper version • Offer real advantages • Making the respondents aware of the web option • Offer practical guidance • Computer competence and motivation play a minor part (if designed well) 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 47 For more guidelines … 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 48 Implementation of SBS Survey • Running from 2007 onward • About 75.000 business receive this questionnaire every year Web take-up rates: • Dutch Annual SBS, 2007: 84 % Total response rate did not change (80%) • Dutch average 2008: 51% 3 largest Dutch business surveys (2008): 77 % (STS: 63%, SBS: 80%, Intrastat: 100%) 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 49 Implementation of SBS Survey https://formulieren.cbs.nl/downloads per survey, per period, per business unit the log on codes 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 50 The portal http://formulieren.cbs.nl 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 51 Welcome page 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 52 In the index items that are complete and correct according to the person who is filling in the form are ticked off 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 53 Click on the ? and then a more extensive explanation appears 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 54 Print of the SBS questionnaire 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 55 The confirmation of receipt 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 56 Paper version SBS questionnaire 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 57 SBS design in production Using SBS design as basis for other surveys • Visual design principles of new paper and web questionnaires starting point for new style guide - for all 450 business questionnaires for various modes: paper, offline, online • Development of QUAT: Questionnaire Application Tool - 15 October 2013 project 2005-2006 OSS Seminar, Statistics New Zealand, Wellington 58 SBS design in production Questionnaire Application Tool (QUAT): • Generic Tool for generating questionnaires: - Editing of questionnaires, letters, and attachments Uses data bases of populations and variables Allows for tailored questionnaires, per stratum Allows for generating questionnaires in mixed-mode design: paper, online, offline Based on one set of visual design principles for each mode • Uniform corporate identity concerning questionnaires and letters (one look-and-feel) • Input coordination with regard to variables and classifications • QUAT planner: for planning the dispatching of Qs 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 59 SBS design in production • SBS design still in production • But, offline system is out of date: - runs only on Windows PC – lot of complaints from Apple users Software to be downloaded – no open source software • QUAT is still running • New system to be developed, on the same principles 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 61 Overview 1. Overview of developments at Statistics Netherlands: a. Electronic data collection b. Organising data collection and data sharing 2. Case study: Moving the Structural Business Survey to the web a. Developing and testing the questionnaire b. Communication strategy to launch the survey and improving web-take up rates c. Implementation of the survey Lessons learned and challenges 3. Looking into the future: future challenges 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 62 Future challenges for statistics Global developments: • Emergence of the information society • Globalization • Shift in the balance of power between NSIs and businesses • • Erikson, Haraldsen & Snijkers, 2012, UNECE Seminar, Geneva Snijkers & Haraldsen, 2013, BLUE-ETS final Conference, Brussels 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 63 Challenges in modernising statistics Based on global developments and data coll. strategy: • Expanding usage of registers - data sharing and data warehousing assessing quality of register-based statistics • Blending designed and organic data - data integration theories • Modernising current business survey designs - tailoring design to the business context, including collecting data from globalised establishments using paradata adopting a communication perspective: “Improving the Respondent Experience” • Producing statistics in a globalized world 15 October 2013 developing new statistical indicators market position of an NSI? OSS Seminar, Statistics New Zealand, Wellington 64 Consequences … • Challenges are too big for individual NSIs: - collaborate on international level - collaborate with universities • Position in the information market: Our added value? - “slow” or/and “fast” statistics? accurate or/and timely statistics? - NSIs have access to and can combine many data sources: new statistics - Bench marking of non-NSI statistics - National Data Archive? National Data Service Center? 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 65 Future challenges for statistics Two final statements: We are at the beginning of a new era in statistics production: The era of data integration Are we moving too slow? The future is already here! 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 66 References • Beukenhorst, D.J., and D. Giesen, 2010, Internet Surveys at Statistics Netherlands. Paper presented at the 2nd International Workshop on Internet Survey Methods, Statistics Korea, 8-9 september 2010, Deajon, South korea. • Haraldsen, G., and M. Couper, 2013, How to Design Effective Business Web Surveys, Course at the Nordic Meeting, 14 August 2013. • Haraldsen. G., and G. Snijkers, 2013, Modernisation and Quality of Business Statistics: The NSI Perspective. Paper presented at the 3rd European Establishment Statistics Workshop, 9-11 September 2013, Nuremberg, Germany. • Groves, R.M. (2011), Three Eras of Survey Research. Public Opinion Quarterly, Vol. 75, pp. 861-871. • Erikson, J., G. Haraldsen, and G. Snijkers, 2012, The Future of Statistical Data Collection? Challenges and Opportunities. Paper presented at the UNECE Seminar on New Frontiers for Statistical Data Collection, 31 October-2 November 2012, Geneva, Switzerland. 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 67 References • Snijkers, G., H. Haraldsen, J. Jones, and D. Willimack, 2013, Designing and Conducting Business Surveys. Wiley, Hoboken. • Snijkers, G., H. Hermans, and R. Göttgens, 2011, Data Collection and Sharing at Statistics Netherlands, Paper presented at the CES meeting, 14-16 June 2011, UNECE Geneva, Switzerland. • Snijkers, G., E. Onat, E., and r. Vis-Visschers, 2007, The annual structural business survey: Developing and testing an electronic form, Proceedings of the 3rd International Conference on Establishment Surveys (ICES-III), Montreal, June 18–21, American Statistical Association, Alexandria, VA, pp. 456–463. • Wieling, M., 2013, Electronic business questionnaires at Statistics Netherlands, Presentation at Statistics Netherlands for guests from Statistics Canada, 7 June 2013, Heerlen, Netherlands 15 October 2013 OSS Seminar, Statistics New Zealand, Wellington 68