Latest / 5 Minute UX / Use Open Card Sorts for New Navigation Structures
Transcript
- 0:08This lesson is Use Open Card Sorts for New Navigation
- 0:11Structures. Let's get started.
- 0:19In an open card sort, you hand participants a set of items
- 0:22printed on cards and ask them to group them in a way that makes
- 0:25sense. No categories are provided up front. Instead, the
- 0:27participant creates the groups and gives each one a name. They
- 0:30describe what belongs where based on their own mental model
- 0:33of the content. This stands in sharp contrast to a closed sort,
- 0:36where the high-level categories are already defined and
- 0:39participants simply place items into those pre-existing buckets.
- 0:42In the open variation, the structure emerges from the
- 0:45people doing the sorting, not from the researcher’s
- 0:47assumptions. The primary goal is to generate a site map or a
- 0:50hierarchy of content, categories, and subcategories.
- 0:53This structure organizes specific content types like
- 0:55articles, documents, videos, or photos. By letting participants
- 0:58define the logic, you capture how users naturally perceive
- 1:01relationships between different pieces of information. When the
- 1:04session concludes, you look for common patterns in how people
- 1:07grouped the items. You also note areas of confusion or
- 1:10disagreement. These insights reveal the underlying mental
- 1:12models that should drive your navigational structure.
- 1:21The right moment to reach for an open card sort is when you are
- 1:23determining a new navigational structure from scratch. You
- 1:26might also use it when making significant changes to an
- 1:29existing navigational structure, where the old hierarchy no
- 1:31longer fits the content. The core goal is to create a site
- 1:34map that establishes a clear hierarchy of content,
- 1:36categories, and subcategories. This approach is especially
- 1:39effective for complex content sources, such as a company
- 1:41intranet, where user mental models must drive the
- 1:43organization. In these scenarios, participants sort
- 1:46items like articles, documents, videos, or photos into groups
- 1:49that make sense to them. Because they create the groups
- 1:51themselves, the resulting structure reflects how people
- 1:54actually think about the information. This direct access
- 1:56to user mental models is why the technique outperforms predefined
- 1:59categories in these contexts. You are building the
- 2:01architecture based on genuine user understanding rather than
- 2:04assumptions. The outcome is a structure that feels intuitive
- 2:07because it emerged from real user behavior. This context
- 2:09sets the stage for the specific execution methods that follow.
- 2:18The standard way to run an open card sort is with a single
- 2:21participant, in person, sorting physical cards. This individual
- 2:24format keeps the mental model pure because the participant is
- 2:26not influenced by others in the room. You can also run these
- 2:29sessions remotely using digital tools that replicate the card-
- 2:32sorting experience. These remote setups allow you to recruit
- 2:34participants who are geographically distributed while
- 2:36still capturing the same individual decision-making
- 2:39process. Group sorts exist as a variation where participants
- 2:41work together to sort items, but this is not the typical approach
- 2:44for open sorts. In a group setting, the dynamic shifts
- 2:46toward consensus-building rather than individual mental modeling,
- 2:49which can muddy the data you are trying to collect. The primary
- 2:52goal remains understanding how a single user organizes
- 2:54information, so individual sessions are the norm for this
- 2:57technique. When executing remotely, you need tools that
- 2:59support both the sorting activity and the subsequent
- 3:01analysis. Platforms like OptimalSort or UXtweak provide
- 3:04the interface for participants to drag and drop digital cards
- 3:06into their own groups. These tools also capture the data in a
- 3:09format that makes analysis manageable, which is critical
- 3:11because open sort data is more complex to interpret than closed
- 3:14sort results. You are looking for patterns in how participants
- 3:17named and grouped their categories, not just checking
- 3:19off predefined labels.
- 3:26The real challenge with an open card sort isn't the sorting
- 3:28itself, it’s what happens after you collect the data. Since
- 3:31participants create their own groups and name them, you don't
- 3:33have a predefined set of categories to fall back on. This
- 3:36makes the analysis significantly more complex than a closed sort,
- 3:38where you can simply count how many items landed in each of the
- 3:41few provided buckets. Here, you’re looking for common
- 3:43patterns in how people are sorting the items across
- 3:45different participants. You’re also hunting for common areas of
- 3:48confusion or disagreement, which are your most valuable signals
- 3:50for information architecture decisions. If three different
- 3:53people put the "Help" article into "Pricing" but named that
- 3:55group differently, that’s a structural conflict you need to
- 3:58resolve. Because there’s no small set of possible categories
- 4:00to focus on, you have to interpret the semantic overlap
- 4:02in the labels participants chose. This is where manual
- 4:05analysis becomes critical, and resources like Donna Spencer’s
- 4:07spreadsheet provide the specific instructions needed to make
- 4:10sense of that open-ended data. The goal is to distill that
- 4:12noise into a coherent hierarchy, identifying where the user’s
- 4:15mental model diverges from your intended structure.
- 4:23So, when you’re ready to apply appropriate tools and analysis
- 4:26methods to interpret common patterns and confusion areas,
- 4:29you have a specific path. If you’re running the sort
- 4:31remotely, OptimalSort or UXtweak handle both the participant
- 4:34session and the backend analysis for you. They strip away the
- 4:37manual labor of chasing down individual grouping decisions,
- 4:39which is a significant time saver. For manual analysis, or
- 4:42if you need a more granular view of the data, look to Donna
- 4:45Spencer’s spreadsheet. It’s an excellent resource specifically
- 4:48for interpreting open sort data, and it comes with built-in
- 4:50instructions to guide your process. This is particularly
- 4:53helpful because open sorts lack the predefined high-level
- 4:55categories that make closed sorts easier to analyze. Here
- 4:58is the decision you face: do you need the speed of a closed sort,
- 5:01or the depth of an open one? Closed sorts are relatively easy
- 5:04to analyze because you can simply count how often items
- 5:06fell into specific buckets. With open sorts, you’re building the
- 5:09map from scratch, so your analysis tools need to handle
- 5:12that complexity. When you sit down to analyze your results,
- 5:14remember that you’re not just counting items; you’re
- 5:17identifying where user mental models diverge. That divergence
- 5:19is your most valuable data.
- 5:24Welcome to The Take. I'm joined by a fellow practitioner and
- 5:27together we're going to dig into Use Open Card Sorts for New
- 5:30Navigation Structures — what it means in practice and what's
- 5:33worth remembering.
- 5:35[HOST] I keep coming back to the idea that the structure emerges
- 5:39from the people, not the researcher. [LEARNER] That’s the
- 5:42part that feels risky in client work. [HOST] Why does it feel
- 5:46risky to you? [LEARNER] Because stakeholders often arrive with a
- 5:49hierarchy they’ve already drawn on a whiteboard. [HOST] So they
- 5:53want validation, not discovery. [LEARNER] Exactly. They see the
- 5:56open sort as a delay tactic. [HOST] But a closed sort just
- 5:59confirms their bias. [LEARNER] True, but it’s faster to present
- 6:03a clean chart of predefined buckets. [HOST] Speed isn't the
- 6:06same as accuracy. [LEARNER] I get that, but what about the
- 6:09analysis phase? [HOST] That’s where the real work happens.
- 6:13[LEARNER] Donna Spencer’s spreadsheet is helpful, but it’s
- 6:16manual. [HOST] Manual gives you nuance that algorithms miss.
- 6:19[LEARNER] Unless you use AI to cluster the labels first. [HOST]
- 6:23AI can find semantic overlaps quickly. [LEARNER] But it might
- 6:26miss the confusion behind a weird label. [HOST] Right. The
- 6:29confusion is the signal. [LEARNER] So we use AI to sort,
- 6:32then humans to interpret the outliers. [HOST] That hybrid
- 6:35approach works well. [LEARNER] It saves time on the obvious
- 6:39groups. [HOST] And frees you up for the tricky ones. [LEARNER]
- 6:42Like when three people put "Help" in "Pricing". [HOST] That
- 6:45divergence is valuable data. [LEARNER] It reveals a mental
- 6:49model mismatch. [HOST] And that’s what drives the
- 6:51architecture. [LEARNER] So the tool is just the starting point.
- 6:55[HOST] The insight is the end goal.