Chapter 1:
Benchmarking & Early Label Discovery
Introduction
To define my slow-fashion top-level categories, I first benchmarked Amazon’s “All” menu and then ran a small open card sort to align labels with real shopper mental models—ensuring my navigation matches user intent, boosts findability, and reduces click depth.
1. Benchmarking Amazon’s “All” Menu
Objective
Understand how Amazon balances breadth vs. depth and header vs. link styling to keep its “All” menu scannable.
Approach
Focused on two unclickable category headers—Digital Content & Devices and Shop by Department—and one of their child links each: Kindle E-readers & Books and Clothing, Shoes, Jewelry & Watches.
Drew a mini-sitemap illustrating breadth (sibling count) and depth (hierarchy levels).
Amazon “All” mega-menu (left), “Kindle E-readers & Books” menu (center), and “Clothing, Shoes, Jewelry & Watches” menu. (right). Click images to enlarge.
Mini-sitemap of the same branches showing breadth = 3 and depth = 2–3; “(+ n more)” nodes indicate additional categories.
Key Insights
Task-Centered Grouping
Amazon’s “All” menu organizes by user intent (e.g., Shop by Department; Digital Content & Devices; Programs & Features; Help & Settings), not strict product hierarchies—reflecting what shoppers seek to do, not how Amazon internally classifies items.
Broad‐and‐Shallow Structure
Branches cap at 3 levels (e.g., Digital Content & Devices → Kindle E-readers & Books → Kindle Unlimited), favoring a wide, shallow hierarchy that keeps the menu scannable and quick to navigate.
“See All” Fallback
Two of five sections (Shop by Department and Programs & Features) show four clickable top-level category links followed by a “See all” link, offering a concise preview plus one-tap access to the full list—balancing orientation with full browse support.
2. Early Label Discovery via Card Sort
Validate which top-level categories shoppers naturally think of when grouping a small set of product cards drawn from my Slow Fashion Product Catalog Taxonomy case study.
Approach
Conducted an open card sort with 14 product items using UXtweak and three participants.
Product list:
A-Line Dresses
Blouses
Bucket Hats
Cotton Skirts
Crossbody Bags
Linen Pants
Loafers
Oblong Scarves
Panama Hats
Raffia Sun Hats
Sandals
T-Shirts
Tote Bags
Wrap Dresses
Instructed participants to group the product cards into categories that make sense to them, then to name each category.
Collected raw groupings, computed a similarity matrix, and generated two dendrograms — Actual Agreement Method (AAM) and Best Merge Method (BMM) — to see which items clustered strongest.
Similarity matrix showing consensus percentages for each item pairing.
AAM dendrogram (100% merges highlight unanimous clusters).
BMM dendrogram (best-merge clustering for smaller samples).
Key Insights
Five clear categories emerged.
Note: Consensus is calculated using (votes for dominant label ÷ total participants) × 100.
Dresses
Wrap & A-Line at 100% consensus.
Tops & Bottoms
Shirts, Blouses, Skirts, and Pants at approximately 67% consensus.
Note: Linen Pants had 0% unanimous agreement—but 2 of 3 still grouped them with ‘Tops & Bottoms,’ reinforcing the benefit of a unified garment category despite split labels.
Footwear
Sandals and Loafers at 100% consensus.
Bags & Accessories
Crossbody and Tote Bags, and Scarves between 67–100% consensus.
Hats & Headwear:
Bucket, Panama, and Sun Hats between 67–100% consensus.
Unified garment category
Participants preferred a combined “Tops & Bottoms” over separate Shirts vs. Pants categories at the top level.
Accessory grouping
Bags and non-wearable accessories naturally co-occurred, justifying a merged “Bags & Accessories” category.
3. Key Takeaways
User-Centered Grouping
Benchmarking Amazon’s menu confirmed that top-level headers need to reflect user tasks, not just back-end product hierarchies—so my navigation will lead with purpose-driven buckets.
Optimal Shallow Depth
Keeping category hierarchies to three levels maximizes scanability.
Validated Categories
Card sort unanimously clustered items into Dresses, Footwear, etc., giving me five intuitive top-level categories to pursue in the next tree-test.
These findings directly informed my navigation taxonomy—ensuring that top-level categories mirror user mental models, minimize misclicks, and ultimately support a more efficient path-to-product discovery.
4. Next Steps
I adopted five categories—Dresses; Tops & Bottoms; Footwear; Bags & Accessories; Hats & Headwear—as the foundation of my slow fashion catalog’s navigation, then moved on to paper-based tree testing to refine click-through paths.