We have certainly learned some lessons from the implementation of the integrated metadata system, which has been more systematic in the last six years, some because we have seen that our options have had a positive effect and others because we have realised the form they should have taken in order to be more successful. We are even making some changes in the formal circuits of some subsystems with a view to greater efficiency and quality in the results obtained.
Involvement of the institution's top management was fundamental and the tie-in of the creation of documentation with formal and standardised procedures has been an excellent way of keeping documentation up to date at the SP.
Designing a metadata system not only requires considerable knowledge of statistical production, but also means leaving behind some habits acquired in this area. A great capacity for abstraction and tidy, integrated thinking is also necessary. An institution has specialists with all these capabilities but not always with all of them at the same time. The teams chosen to implement these systems must consist of specialists with different profiles among those mentioned, because they complement each other. The IT technicians who develop applications must participate from the start.
We believe that it is essential to develop prototype systems before final implementation. Prototyping is the best way to test a system's design, detect strong and weak points and come up with experience-based alternatives for the weak points. When designing a system like this, it is very hard to give an appropriate description of all its functions without prior experience. Even the workflow of procedures may need some adjustments.
Must be given training to statisticians, not only in the use of applications but also, and above all, about the concepts underlying the system and workflow of procedures. The introduction of the position of survey manager has fostered cooperation and dialogue between production, metadata and dissemination. The distribution of terminology associated with each metadata subsystem is having a beneficial effect at the SP as it encourages the use of a language common to all profiles using the system.
After the classification subsystem was made available to the general public, we began to receive some complaints about its usability and decided to conduct some usability tests. The test results showed us the difficulties that people experienced when using the system and we decided to redo some of the navigation in the consultation application. When we implement the methodological documentation subsystem, scheduled on the beginning of 2008, we have decided to conduct usability tests in the prototype phase of the consultation and publication application so that we do not need to redo any parts of the system after it goes into production.