optimizacion-conversion

Card Sorting: Guide to Better Web Architecture

Adrià Vidal6 min read
card sortingUX researchinformation architectureCROusability

What is card sorting

Card sorting is a UX research technique where participants organize content into groups that make sense to them. The goal is to discover how users expect to find information on a website, app, or any navigation system.

The technique is straightforward: participants are presented with cards (physical or digital) labeled with page names, products, or categories, and asked to group them. Based on those groupings, the design team builds or validates the information architecture.

It is one of the most cost-effective user research techniques because it requires few participants (15-20 is usually enough) and delivers directly actionable results. If your navigation menu is causing friction, a card sort can reveal exactly where the problem lies.

Types of card sorting

There are three main variants, each suited to a different stage of a project:

TypeDescriptionWhen to use it
OpenUsers create their own categories and group cards freelyWhen designing architecture from scratch or exploring mental models
ClosedCategories are predefined; users only assign cardsWhen you already have a structure and want to validate it
HybridCategories are predefined, but users can create new onesWhen you want to validate a structure while leaving room for discoveries

Open card sorting

This is the most exploratory type. There are no predefined categories: participants group cards and name each group. It is ideal for the early stages of a project or when you are redesigning the entire navigation of an eCommerce site.

The risk is that it generates a lot of variability between participants, making analysis more complex. That is why a minimum of 20-30 participants is recommended to get clear patterns.

Closed card sorting

Here the categories already exist. Participants only decide where each card fits. It is faster to run and analyze, and perfect for validating a structure before launch.

If a high percentage of users assign a card to an unexpected category, you have a clear signal that your labeling or hierarchy needs adjustment.

Hybrid card sorting

It combines the best of both: it offers predefined categories but allows creating new ones. It is the most flexible option and the one we recommend when you already have a reasonable structure but suspect there are gaps.

When to use card sorting

Card sorting is not a technique you apply at any time. These are the scenarios where it adds the most value:

  • Navigation redesign: before reorganizing the main menu or product categories
  • New section launches: to decide where to place new content within the existing structure
  • Findability issues: when analytics data shows users cannot find what they are looking for
  • Platform migration: to take advantage of the technical change and improve information architecture
  • Catalog expansion: when an eCommerce goes from 500 to 5,000 SKUs and the original structure no longer scales

Card sorting results perfectly complement quantitative data from tools like Hotjar or Microsoft Clarity, which show where friction exists, while card sorting explains why.

Card sorting tools

Optimal Workshop

Optimal Workshop is the go-to tool for remote card sorting. Its OptimalSort module supports open, closed, and hybrid studies with remote participants. It automatically generates dendrograms and similarity matrices, speeding up analysis.

It also includes TreeJack (for tree testing) and Chalkmark (for first-click testing), allowing you to validate information architecture from multiple angles within the same platform.

Maze

Maze has evolved from a prototype testing tool to a full research platform. Its card sorting module is newer but integrates with the rest of the research workflow: you can combine card sorting with usability testing in the same study.

Other options

ToolBest forApproximate pricing
Optimal WorkshopDedicated IA studiesFrom 99 USD/month
MazeIntegration with prototypingFrom 99 USD/month
UserZoomEnterprise and owned panelsCustom pricing
Miro / FigJamIn-person card sorting or workshopsFree / from 8 USD/month
UXtweakBudget-friendly alternativeFrom 60 EUR/month

For small teams or limited budgets, an in-person card sorting session with sticky notes and a whiteboard is still perfectly valid. What matters is the method, not the tool.

How to analyze the results

Similarity matrix

The similarity matrix shows how often two cards were placed in the same group. High values indicate that users perceive those pieces of content as related. It is the first analysis you should run.

Dendrogram

The dendrogram visualizes hierarchical clusters: which cards are grouped first (high similarity) and which are joined later (low similarity). It helps you identify the "natural cuts" for defining main categories and subcategories.

Key metrics

  • Agreement rate: percentage of participants who assign a card to the same category. Above 70% you have clear consensus.
  • Ambiguous categories: cards that are split between two or more categories roughly evenly. These need attention.
  • Outliers: cards that no participant groups consistently. They may indicate confusing content or items out of context.

Impact on information architecture and conversion

Well-designed information architecture reduces friction in the customer journey and has a direct impact on business metrics:

  • Lower bounce rate: users find what they are looking for faster
  • More pages per session: intuitive navigation invites exploration
  • Higher conversion rate: less friction in product discovery means more users reaching checkout
  • Fewer support tickets: when information is where the user expects it, there are fewer inquiries

Card sorting is a modest investment (a full study can be run in 1-2 weeks) with a high return, especially in eCommerce with large catalogs where findability is critical for conversion.

If you combine card sorting with usability testing and heuristic analysis, you have a solid research process that grounds design decisions in real data, not internal opinions.

Common mistakes

  1. Using too many cards: more than 60 cards fatigues participants and contaminates results. Between 30-50 is the optimal range.
  2. Ambiguous labels: if the card name is unclear, participants group based on their interpretation of the name, not the actual content.
  3. Too few participants: with fewer than 15 people, patterns are unreliable. With 20-30 you get robust results.
  4. Not complementing with tree testing: card sorting tells you how to group; tree testing tells you if the resulting hierarchy works in practice.
  5. Ignoring outliers: cards that nobody knows where to place often reveal naming issues or content that needs restructuring.

Conclusion

Card sorting is one of the UX research techniques with the best cost-to-impact ratio. It allows you to build information architectures based on the mental models of your actual users, not on your organization's internal logic. For any redesign or navigation improvement project, it is a step you should not skip.

If you need to improve your website's conversion through a better user experience, at Boost we work with research and experimentation methodologies to optimize every touchpoint. Get a free initial assessment with Scan&Boost.


Adrià Vidal — Boost · Conversion Rate Optimization

Adrià Vidal

Adrià Vidal

CEO & Founder

Founder of Boost. Specialist in digital analytics, CRO, and artificial intelligence applied to digital business optimization.

Related articles

optimizacion-conversion4 min read

Amplitude: Product Analytics Guide

Amplitude is the product analytics platform that lets you understand user behavior in real time. Learn how to set it up, what metrics to track, and how to...

Card Sorting: Guide to Better Web Architecture | Boost