This would be an alternate UI (i.e. on a different page) which replaced the facet-controls (i.e. the codelist browsing/ searching) with a series of questions and answers (one for each facet).
Each choice would be presented in turn (in descending order of entropy - i.e. the dimensions that narrow down the number of candidates quickest would appear first).
There would also be an option to skip, useful for when the filter isn't relevant to the user's requirements.
We'd need to write some narrative content (the questions and answers) so it wouldn't be easy to generalise to other dataset families, but it could lead to more intuitive options for the user.
We could update the results table each time a question was answered.
Instead of using heuristics to reach the point where data could be previewed (per data previews #83), we could just continue asking questions and answers until we reach a single cube, slice or observation - at which point we could present (or link to) that.
A major obstacle making this hard to use is the range of codelists under each facet.
With a wizard we might want to reduce the range of choices even if that meant sacrificing some detail. For instance, we might present the option of "European Union" and behind the scenes permit a range of definitions EU28, EU27, EU12 etc.
We should probably aim to harmonise products, industry geography and date before attempting to build the wizard (#81).
Having types/levels for codes might also help to structure the forms for each facet (e.g. "Which products are you interested in? Goods, Services or both?") (#80).
Extending the original description from the mockups...
This would be an alternate UI (i.e. on a different page) which replaced the facet-controls (i.e. the codelist browsing/ searching) with a series of questions and answers (one for each facet).
Each choice would be presented in turn (in descending order of entropy - i.e. the dimensions that narrow down the number of candidates quickest would appear first).
There would also be an option to skip, useful for when the filter isn't relevant to the user's requirements.
We'd need to write some narrative content (the questions and answers) so it wouldn't be easy to generalise to other dataset families, but it could lead to more intuitive options for the user.
We could update the results table each time a question was answered.
Instead of using heuristics to reach the point where data could be previewed (per data previews #83), we could just continue asking questions and answers until we reach a single cube, slice or observation - at which point we could present (or link to) that.
A major obstacle making this hard to use is the range of codelists under each facet.
With a wizard we might want to reduce the range of choices even if that meant sacrificing some detail. For instance, we might present the option of "European Union" and behind the scenes permit a range of definitions EU28, EU27, EU12 etc.
We should probably aim to harmonise products, industry geography and date before attempting to build the wizard (#81).
Having types/levels for codes might also help to structure the forms for each facet (e.g. "Which products are you interested in? Goods, Services or both?") (#80).