is it OK to filter results of a previous filter? If so, how should it work?

Question

First time asker, forgive me if I get anything wrong. I'm working on a big-data enterprise product that has many filters to allow the user to find what they want in a table with thousands of rows and 20+ columns. Every column is filterable, and columns might include "Price Range," "Weight," etc. Out of these dozens of filterable columns, two in particular are what I'd like to ask about: "Category" and "Subcategory." These two filters, as you can imagine, have a close relationship.

As the UI works right now, Subcategory's available options to filter from are unaffected by Category's selection. A user could filter by the Dairy category, for instance, and still see Subcategory options that might include "Cookies" or "Beef", for example.

What is the convention that I should follow in this case? Keep the filters independent of each other (like all other filters) or make these filters dynamically populate one another (either one way or both ways)? I'd be interested in your thoughts/any examples in the real world you see where some filters affect other filters, and other filters don't affect others, all in the same table.

Thank you in advance!

Answer 1

What do the users assume?
I would be quite confused if I select "vegetables" (if that is a category" and still see the different meats. But I am only one and a designer, to boot.
Please test this with new potential users.

Design a prototype where the second column adapts to the selcetion of the first and again test this with new potential users. Think about animations or other design "tricks" that clarify the dependency.

Of course, you have to consider the existing users who are already used to the established way. In my experience, if the new way is more logical and easy to perceive, users will adapt quickly (there will be some resistance, though)

Established users are used to how the table works, but usually you always have to consider new users as well. If the established design does not work well, the onboarding of new users is more difficult, it may even cause resistance.

You have to weigh the costs of keeping the design and changing it. Having test data will facilitate a good decision.

Answer 2

I second the sentiment in another answer: might there be a taxonomy problem here if the user can select the Dairy category and see meats and cookies? Obviously without more context and knowing your users and their goals I can’t say, but if I was doing a standard grocery task with my own mental model of what “grocery categories” look like, that would confuse me.

You might want to consider doing a card sort study with all of your categories and subcategories and see how your users would organize things (if that’s not already been settled). That could help you to determine not only which should be categories versus sub categories, but how to organize things and what paths people might expect to take with their filtering terms to get certain results.

This SOUNDS like it would be a case for simple additive filtering, as if it were a multi-term search; I believe this is how some spreadsheets like Excel do it when you select filters on multiple columns: the first filter you apply results in displaying only the rows which include that term in that column; the next filter you apply is applied the same way but only to those results of the first filter, resulting in a smaller list (there is probably more nuance possible, but I am a pretty simple Excel user).

If the order in which those filters were applied might provide different results, giving the users some visual indication of that order (likely at the top of the table and/or near where the filters are applied) could be a helpful guidepost:

Dairy > Cheese > Goat Milk

I have seen this implemented with each term in a “capsule“ with a remove control, allowing the terms to be moved around or removed from any point in the chain:

Be careful, however… Depending on how you do this and display the results, you can definitely confuse people.

As for testing, you really should do some before choosing a direction. If you truly have no access to any of your users or ability to test with them, you may be able to learn something from past research or analytics data about how people find what they need. There is so much more context needed here, but you might be able to find people who work with your users to test with as proxies (grocery clerks? Caterers? sales people?); you could test with your own internal stakeholders who are closest to these users; you could also perform an expert review with a group of other UX people could give you good insights as well, though you will need to be very clear up front about all of the context of use and who your users are.

But especially if you have an entrenched existing user base, try not to send anything out into the wild without doing some reasonable testing first!

Just a few thoughts… Hope it is helpful.