UX Research Term

Similarity Matrix

A similarity matrix (also called agreement matrix or co-occurrence matrix) is a grid showing how frequently participants placed pairs of cards in the same group during card sorting.

How It Works

Each cell shows the percentage of participants who grouped two specific cards together:

  • 100%: All participants grouped these cards together
  • 50%: Half of participants grouped them
  • 0%: No participants grouped them

Visual Representation

Often displayed as a heatmap:

  • Dark/hot colors (red): High agreement - strong relationship
  • Light/cool colors (blue): Low agreement - weak relationship
  • Diagonal is always 100% (card grouped with itself)

What It Reveals

Strong pairs (80-100%):

  • Clear consensus these belong together
  • Core of a category
  • Use for naming categories

Moderate pairs (50-79%):

  • Some agreement but not universal
  • May belong together with caveats
  • Consider context

Weak pairs (0-49%):

  • Little consensus
  • Probably don't belong together
  • May fit in different categories

Example Use Case

"Shopping Cart" + "Checkout" = 92%
→ Strong relationship, definitely same category

"Shopping Cart" + "Help" = 12%
→ Weak relationship, different categories

"Help" + "FAQ" = 88%
→ Strong relationship, group together

Using the Matrix

  1. Identify clusters: Look for hot spots (high agreement blocks)
  2. Find category boundaries: Where agreement drops off
  3. Spot outliers: Items with low agreement across the board
  4. Validate structures: Compare against proposed navigation

Advantages

  • Quantitative measure of relationships
  • Easy to spot patterns visually
  • Can export for further analysis
  • Complements dendrogram analysis

Free Card Sort provides interactive similarity matrices with every study!

Try it in practice

Start a card sorting study and see how it works