AI Personalization in eCommerce: 2026 Guide
What AI personalization in eCommerce is, types of personalization (content, product, price), tools like Dynamic Yield, Nosto and Algolia, key metrics, and...

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.
There are three main variants, each suited to a different stage of a project:
| Type | Description | When to use it |
|---|---|---|
| Open | Users create their own categories and group cards freely | When designing architecture from scratch or exploring mental models |
| Closed | Categories are predefined; users only assign cards | When you already have a structure and want to validate it |
| Hybrid | Categories are predefined, but users can create new ones | When you want to validate a structure while leaving room for discoveries |
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.
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.
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.
Card sorting is not a technique you apply at any time. These are the scenarios where it adds the most value:
Card sorting results perfectly complement quantitative data from tools like Hotjar or Microsoft Clarity, which show where friction exists, while card sorting explains why.
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 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.
| Tool | Best for | Approximate pricing |
|---|---|---|
| Optimal Workshop | Dedicated IA studies | From 99 USD/month |
| Maze | Integration with prototyping | From 99 USD/month |
| UserZoom | Enterprise and owned panels | Custom pricing |
| Miro / FigJam | In-person card sorting or workshops | Free / from 8 USD/month |
| UXtweak | Budget-friendly alternative | From 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.
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.
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.
Well-designed information architecture reduces friction in the customer journey and has a direct impact on business metrics:
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.
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
What AI personalization in eCommerce is, types of personalization (content, product, price), tools like Dynamic Yield, Nosto and Algolia, key metrics, and...
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...
Criteo is the leading dynamic retargeting platform for eCommerce. Learn how its technology works, what formats it offers, and how to integrate it into...