Oftentimes, all of your data will be in one file and will need to be split out in order to build comparisons. Once you've uploaded your master data file into the platform, you can create subsets of your data (splits) one at a time or in bulk via the data library.

In this article:

The importance of metadata

While the text component of a data set will always be the central object of analysis in Relative Insight, associated metadata points (e.g. demographic details of survey respondents) can be used to split out a master data file in different ways and build multiple comparisons using the same data. In doing so, you can build a 360° view of your audience of interest.

In the example below, we could build at least four unique comparisons using this single data set. In each case the text being analysed is the same, it is just being grouped differently.

  • Gender

  • Age

  • Location

  • Like vs dislike

Creating splits in bulk

E.g. creating a unique data set for each country from which you've received survey responses

Bulk splitting can be used to create a maximum of 20 data sets at once.

  1. Within the relevant project folder in the data library, select the checkbox next to the dataset you want to split

  2. Select the advanced split icon from the bulk action menu at the top of your data list

  3. In the metadata column, select which pieces of metadata you want to use as the basis for creating your splits. For each metadata point selected, a unique data subset will be created that only includes text associated with that attribute. The subsets you're creating will be previewed in the right-hand pane.

  4. Ensure the setting at the top of the right-hand pane is set to 'Split all results', and if desired rename the new data sets before completing clicking 'Split' to finish

Combining metadata points into a single split

e.g. group together survey responses from different countries into regional datasets.

In some situations, you might want to group together several metadata points and their associated text.

To do this, follow the above approach selecting only the metadata points you want to group together. Once you've done this, change the setting at the top of the right-hand panel to 'Combine all' before clicking 'Split' to complete the action.

Multi-condition splits

e.g. creating a subset of your data only including survey responses from respondents for which location = USA and age = under 25

This operation currently supports 'AND' logic only, meaning if you define multiple conditions only the pieces of text that meet all of them will be included in your new data set.

  1. Hover over the applicable data set and select ‘Split’

  2. Use the dropdown menu to define a rule for your split then click 'Add Rule'

  3. Add additional rules by clicking 'Add another'. A preview of the data set size will be presented in the bottom-right corner of the modal. When you're done, click 'Finish it'

  4. Give the new data set a name and click 'Create the split'

Did this answer your question?