Simply put, filters are ways of sorting through your data. There are endless ways to slice and dice your data on the Stratifyd platform. You can apply filters to drill down into your data to compare subsets to one another, find insights, and discover trends. Here, we’ll cover some tips and tricks for using filters.
Tips and Tricks for Using Filters
Apply filters at different levels
You can apply filters to your data at different levels.
Global filters are applied to your entire workspace
Tab filters are applied to a single tab in your workspace
Widget filters are applied only to a given widget
Add multiple tabs for efficient analysis
Say you have multiple brands you’re tracking in the same data stream. If you wanted to review the data for just one of those bands, you could duplicate the original tab, rename it, and then apply tab-level filters for a specific brand.
To do a side-by-side analysis of two brands or products, copy an interesting widget and apply product-specific filters. This can be especially useful in word cloud or sentiment visualizations.
Use Boolean logic wisely
Introducing AND/OR/NOT logic through the Advanced Filter designer combines filters to make them more powerful.
AND returns only results that match all your filter criteria. This default setting when using the platform.
OR returns results that includes a single match for any of your filter criteria. This type of search will can only be used through the advanced filter options
NOT returns results where a filter does not match a specific criteria.
It’s considered a best practice to use Boolean searches against your taxonomy labels to include or exclude certain topics. Perhaps you are curious about customer sentiment but don’t want to include UK customers. Under the Advanced Filters menu, you can use the taxonomy path and visual editor to drag “NOT” to the UK label, and AND to the language and sentiment fields.
In this illustration, results must be in English and also have a sentiment value between 0 and 2, while the location must not be United Kingdom.
Filtering is best utilized when combining structured and unstructured fields. If you have dimensions like CSAT Score or Star Rating, those may be the best place to start slicing your data. You can then overlap unsupervised models and taxonomies to understand what is driving high and low ratings.
The order in which we apply filters does matter. If we see a spike in review volume, it would be best to start filtering on that time range. We can then take a look at how sentiment and conversation changed during that time. If we saw the spike but started our analysis by diving into sentiment, we could be filtering out critical data.
Tell a Story
Applying a set of filters to your data is called a story. Stories can be a very helpful way to store data anecdotes. As we discover stories and interesting findings in our data, we can simply save all filters we have applied and easily share with our team. This way, we don't have to remember on our own what click path we took. Also, if certain users will always be interested in a subset of data, we can save their filters to apply them all with one click.
Don’t contradict yourself
If some of your widgets appear blank, you may be adding conflicting filters. Check your filters at each level to be sure they are not canceling each other out. If they are, remove the conflicts and the widget should populate again.
A common mistake is a saved filter that refers to a specific topic in a theme detection model. If the model is updated, then that topic reference could become unavailable and result in blank widgets
Still have questions about using filters? Don’t hesitate to reach out!