We have a list of things in which there are exactly 4 or so categories of items. This list is fairly static, but slowly growing as new items get added. Think: a list of projects, or a mailbox, or purchase orders, etc.

In our interface, we're providing tools above the list of items to help the user find items they might be looking for. We're already providing a keyword search facility to cover the use case of finding a known specific thing.

We also want to provide some dynamic filter switches for the categories of items. The user can toggle those switches and show/hide respective categories. This is to support the use case of discovering items of an unknown category.

Here's our quandary - which search model do we support?

a) winnowing: user has big list of mixed categories, and hides one category at a time until they get a manageable list (winnowing the chaff). This would be facilitated by a multi-choice set of filters (eg. checkboxes).


b) focusing: user has a big list of mixed categories, and selectively shows one category in isolation (selective focusing). This would be facilitated by an exclusive-choice set of filters (eg. radio buttons).


What are some of the UX issues you can see with these two approaches? What are the factors you would consider to assist deciding which approach to take?

4 Answers 4


Looking at these two options as provided - winnowing and focusing:

The problem with winnowing is the risk of accidentally winnowing out the item you're actually looking for - i.e. throwing the wheat out with the chaff. You have to know what is safe to remove, which means you need to have an idea of what is most likely required to keep, and if that's the case, then isn't it annoying to start with all selected because you have to deselect all the others to hone in.

Focusing is more systematic, and does not suffer from the same problems as winnowing - you might skip past the selection with the desired item in, but that's slightly different.

Focusing allows you to look at what you think is the most likely category with a single click and if not found, to switch to another category with one more click.

But winnowing makes it less easy (more clicks) to select a single category if the user has a 'gut feeling' as to which is the right category. Subsequent failure to find an item in that category then requires deselection, and then selection of another (more clicks)

So, focusing initially seems the best bet.

However - there also side-benefits to winnowing, in that for a new user, it's a better way of presenting greater amounts of information - it allows exploration and discovery. Focussng is better for users who already have some knowledge of the content and can make those informed decisions that ease the process. A new user probably has no reason to focus on any one option over another. Winnowing increases the chance that the next time the process is undertaken, the user has a little 'previously learned information'.

Consider whether new users will be quickly getting to grips with the content, or whether this could be a slow process.

So Winnowing seems better for new users - focusing is better for experienced users.?

But there may be additional factors which make either of the processes more difficult. For example in the case of a larger number of categories, winnowing becomes increasingly harder, because, the reasons to discard one category have to be more firmly held; the confidence that the desired result is in the remaining categories is harder to determine (as opposed to being thrown out with the chaff); and the ease of back-tracking is reduced. In all likelihood, failure to find a desired item will result in reselecting all categories and starting again - a point of frustration.

In this case, the focusing option (already identified as the more systematic approach) should prevail, especially if the presentation of content is made easily traversable. For example, compare ease of scrolling up and down a long section of content versus paginated displays. Being more systematic, as the user progresses from one category to the next, the confidence that the desired result is within the remaining categories is increasingly greater with each progression.

So As complexity increases, systematic focusing prevails.

Consider each variable and implement accordingly.

  • 1
    There's one other use case to consider, and that is that the user might want to see multiple of the categories together, but not others. For example: broadcast messages, orders, deliveries, and alerts, but not social spam, marketing spam, or HR spam.
    – Erics
    Commented Oct 16, 2011 at 3:32
  • if you look at complex datasources and fuzzy user search behaviour lets look at how ebay does it: eg bit.ly/mWqu9P uses both: towards the top of the filter list we have mutually exclusive list items and towards the bottom are the checkboxes. Logic is "if search for object then FOCUS... if search for attribute of object then WINNOW"
    – colmcq
    Commented Oct 17, 2011 at 9:44

From my experience this is highly dependent on the users' understanding of the system. If the user is aware that attributes are mutually exclusive and has a clear task, workflow, and is an experienced user of the system then radio buttons may be the way forward. If the user does not have a clear mental model of the system and has fuzzy goals then checkboxes might be a better solution.

more: Radio buttons remove control from the user, which maybe desirable as an error prevention mechanism but checkboxes can simulate radio button exclusivity if the user selects one option. In other words checkboxes have dual functionality that radio buttons do not.

summary: use radio if task focused, user understands system use checkboxes if fuzzy task, multiple outcomes desired.

I would would love to see some evidence base behind this, but that's my 2pence worth.


It all depends on the data and what the categories are.

If you have a long list of categories, showing a single category is better - you can add a next category button that makes it really quick to go to the next category if you don't see what you want. With more categories as checkboxes you may need a lot of clicking to get the filter you want.

If the categories are unrelated, a user won't want to see them together except with "show all", so the radio buttons make more sense.

If some of the categories are closely related or contain very few items, you probably need the checkboxes to provide sufficient control.


I've seen a couple of times some hybrid cross of these two.

You create a list with checkboxes, and assign list item names to be links that act as radio buttons when clicked (exclusively leave only the selected item checked), while clicking on the checkboxes allows user to add more than one at the same time.

This is not a widely adapted UX in my opinion, it requires some clicking around to get familiar with how it works, but ive seen systems that couldn't work in any different form than this.

Making the list items react on hover, and act as links might help user understand the nature of this system by giving a visual feedback of each item.

  • I would have accepted this answer. It solves the problem mentioned by Roger that "winnowing makes it less easy (more clicks) to select a single category," and incorporates colmaq's key point that "checkboxes can simulate radio button exclusivity if the user selects one option." Commented Oct 17, 2011 at 1:07
  • Ah, tricky nuance here is that this answer is mostly about execution but doesn't address the main point of the original question of "What are some of the UX issues you can see with these two approaches? What are the factors you would consider to assist deciding which approach to take?". See the bottom of this blog post for more insights.
    – Erics
    Commented Oct 20, 2011 at 7:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.