Session 1 – Communication of Statistics
Lessons Learned:
- Complicated and fancy designs of communication material may not improve things (3)
- Communication is an essential part of survey design
- We should have a marketing strategy on the collection side
- Better communication makes data collection more efficient and effective (higher response)
- Automated/interactive FAQ for online surveys improves response rate
- Be pro-active and stay in contact with respondents
- Ask respondents if they “think what you think they think”
- Interaction with customers and good and active user support is important
- Use mail sort exercise for testing design of mailing packages
- Business registers could be good communication tools
- Importance of thorough testing of data collection instruments and dissemination tools
- Use in-house data for targeted marketing
Future work:
- Develop tools to measure the (cost) effectiveness of our interventions made to improve response (2)
- Best practices in how we communicate and reach respondents
- - Best practices for communication with respondents (especially for initial contact)
- - Research/testing different ways of reaching different respondents
- - Use of social media to reach-out to respondents
- - Find out more about personalized customer portal
- - What incentives improve response rates
- - Direct marketing to respondents who did not finish survey (‘re-marketing’)
- How to transition from traditional surveys to web based approaches (best practices)
- More examples of linking online collection with dissemination sites
- Thorough testing standard tools and business processes for data collection
- With change from fixed to mobile phones, do we need to rethink sample frameworks (no geographic location, household vs individual)?
- Guidelines on how to create e-mail databases and safeguard privacy
Session 2 – New Tools and Methods
Lessons Learned:
- By focusing on a particular type of Big Data, we can be very practical about what needs to be solved
- Use open source algorithms, specialist companies and not internal IT, standard query system for mobile phone data, work together with providers
- Big Data provides big opportunities but need mixed sources to validate algorithms and results (not the solution to everything)
- Big Data doesn’t solve all problems and might raise new ones
- New technologies and data sources need new skills (and new staff?)
- Outsourcing or partnerships can fill gaps in in-house knowledge
- Paperless is the way to go
- Greater use of tablets is worth to follow-up on
- Use open source and share algorithms
- Harmonization and standardization of tools, taxonomies and software needed
- Phone data and other electronic sources concern have a technical component and a legal/regulatory component
Future work
- Work on international level MOU with mobile phone companies and partner with them using political support (data important for the public good)
- Evaluate whether to make contracts with private sector or make them provide access by law
- Explore how to partner with data suppliers and get aggregate instead of micro-data
- Cost benefit of using Big Data sources
- How to overcome public concerns about privacy and Big Data from NSOs
- How to make data from new sources quality ensured and how to build partnerships
- How to team up IT staff and statisticians to improve and enable innovation
- More work on data integration of new sources
- Organize sharing of algorithms and software demos as workshop session
Session 3 – Data Source Management and Risk Monitoring
Lessons Learned
- Advantages and disadvantages of the different models (subject vs process) are still not clear
- As we are still at an early stage with Big Data, collaboration is important
- Planning is important but monitoring is key
- Follow the whole production system to see where most gains can be made
- Data collection needs a good management process and system
- Potential uses of new data sources, including web-based interviewing
- CAWI needs a different approach than traditional collection methods
- A Big Data framework already exist (Eurostat)
- In order to harness new data sources statistical offices have to be agile to incorporate new IT and software
- Reduce burden by using administrative and tax reporting through common taxonomy to get statistics
Future Work
- Continue cooperation and coordination with all actors on Big Data activities and keep sharing experiences
- Good practices in systematic approaches to risk management (not just for Big Data)
- How to tailor traditional methods to CAWI and what is the optimal mix
- How to incorporate virtual interviewing in web surveys
- Share experiences and ideas on standardized business reporting
- Have standardized ledgers and business reporting based on existing taxonomies.
- How to modify administrative sources to be more useful for statistics and reduce reporting burden
- What are the advantages and disadvantages of subject/silo versus process oriented approaches and how do you manage change from one to another or combine them into a matrix approach