Taxonomy Restructuring: Organizing Complex Inventory via User Intent Data

Role: Senior Product Designer (Contract)
Company: Artfinder
Timeline: 8 weeks

Goal: Redesign Artfinder’s main navigation to improve discoverability and SEO alignment.

Overview

Artfinder's main navigation was hiding their inventory. Users couldn't browse by specific subjects, styles, or mediums from the homepage—they had to land on broad listing pages and filter manually. This created friction and made the marketplace feel smaller than it actually was.

I was brought in to redesign the information architecture and navigation based on what users actually purchased, not just what they browsed.

My contribution: Analyzed Google Analytics purchase vs. visit data, collaborated with an SEO agency on category naming, corrected art world terminology, designed 5 IA iterations, and delivered a 2-level navigation structure that exposed subcategories for the first time.

The Problem

What I found in week 1:

The navigation was too simple:

  • Only top-level categories visible (e.g., "Shop," "Inspiration")

  • No way to navigate directly to "Portraits" or "Abstract paintings" from the homepage

  • Users had to click into broad listing pages, then filter manually

  • This made Artfinder feel like it had limited inventory

The bigger problem: Visits ≠ Purchases

I pulled Google Analytics data and found a critical insight:

The insight: People browsed "Nudes" out of curiosity but rarely purchased. Meanwhile, "People and portraits" ranked lower in visits but higher in purchases—meaning buyers knew what they wanted and went straight to checkout.

The strategic question: Should navigation prioritize browse behavior (what people click) or purchase behavior (what people buy)?

My decision: Prioritize purchase behavior. Navigation should surface what converts, not just what attracts clicks.

Research & Discovery


What I did:

1. Google Analytics audit:

  • Analyzed category page traffic, bounce rates, and conversion paths

  • Identified which subjects/styles drove purchases vs. curiosity clicks

  • Found that users relied heavily on search because navigation was too shallow

2. Competitor analysis (Saatchi Art, Rise Art, Singulart):

  • All competitors exposed 2-3 levels of navigation hierarchy

  • Saatchi used mega menus showing subjects, styles, and mediums

  • Rise Art prioritized "Shop by Subject" as primary navigation

  • Singulart emphasized curated collections + browseable taxonomy

Key insight: Every major art marketplace made subcategories visible in navigation. Artfinder was the outlier.

3. SEO agency collaboration:

  • Worked with an external SEO agency who provided keyword research and category naming recommendations

  • Identified high-volume search terms that should be reflected in navigation

  • Received reports on optimal category structure for crawlability

4. Taxonomy audit:

  • Found art terminology was inconsistent or incorrect

  • Examples: "Impressionistic" (wrong) → "Impressionism" (correct), "Surrealistic" (wrong) → "Surrealism" (correct)

  • Corrected terminology to match proper art world conventions (important for credibility with artists and collectors)

5. Stakeholder workshops:

  • Aligned on business priorities: increase discoverability, improve SEO, maintain artist relationships

  • Confirmed: Artists categorized work correctly, but buyers couldn't find it due to shallow navigation

Process


The Solution: Purchase-Driven IA

Design principle:

Prioritize what people buy over what people browse.

This meant:

  • "People and portraits" gets prominence despite lower browse volume (because it converts)

  • "Nudes" gets de-emphasized despite high curiosity clicks (because it doesn't convert)

  • High-purchase subjects become top-level navigation items

The new structure:

2-level hierarchy:

Top Level: Medium/Category
└─ Second Level: Subject, Style, Medium filters

Example:
PAINTINGS
├─ By Subject: Portraits, Landscapes, Abstract, Flowers, Animals, Architecture ├─ By Style: Impressionism, Surrealism, Minimalism, Photorealistic, Expressionism └─ By Medium: Oil, Acrylic, Watercolor, Mixed Media
SCULPTURE
├─ By Subject: Figurative, Abstract, Animals, Nude
└─ By Medium: Bronze, Stone, Wood, Metal

What changed:

  • Before: Users clicked "Shop" → landed on generic listing page → had to filter manually

  • After: Users see "Paintings → Portraits" directly in navigation → land on exact category → faster path to purchase

The 5 iterations:

I created 5 IA versions, iterating based on:

  • SEO agency recommendations

  • Purchase data prioritization

  • Stakeholder feedback on artist/buyer needs

Stakeholders approved version 5 because it balanced:

  • SEO-friendly category naming

  • Purchase-driven hierarchy

  • Artist terminology accuracy

Key Design Decisions

1. Purchase rank determines navigation order

Instead of surfacing "Nudes" as the #3 category (based on visits), I positioned "People and portraits" higher because it drives more revenue.

Why this matters: Navigation is prime real estate. Buyers should see high-converting categories first.

2. Expose subcategories in navigation

The old nav forced users to click blindly into "Shop" hoping to find their desired subject. The new nav shows all subjects/styles/mediums upfront.

Why this matters: Reduces cognitive load, increases perceived inventory, improves SEO (more internal links to category pages).

3. Correct art world terminology

Changed "Impressionistic" → "Impressionism," "Surrealistic" → "Surrealism," etc.

Why this matters: Credibility. Art collectors and serious buyers expect proper terminology. Incorrect labels signal amateur curation.

4. SEO-aligned category names

Used SEO agency recommendations to match category names to actual search queries.

Example insights from SEO report:

  • "Modern art" had higher search volume than "Contemporary art" → prioritized "Modern" in labels

  • "Landscape paintings" searched more than "Landscape art" → adjusted phrasing

  • Subject-based URLs (e.g., /paintings/portraits) performed better than style-based URLs (e.g., /art/realistic)

The Process

  • Google Analytics audit (visits vs. purchases)

  • Competitor benchmarking

  • SEO agency kickoff

  • Taxonomy audit

  • Created 5 versions of information architecture

  • Tested different hierarchy models (subject-first vs. medium-first)

  • Stakeholder reviews and refinement

  • Designed navigation components in Figma

  • Created desktop mega menu and mobile drawer variations

  • Documented responsive behavior

  • Created taxonomy guide for content team

  • Documented category naming conventions

  • Handed off to engineering for implementation


Conclusion

What shipped:

✅ 2-level navigation hierarchy exposing subjects, styles, and mediums
✅ Purchase-driven category prioritization based on GA conversion data
✅ SEO-optimized category names from agency recommendations
✅ Corrected art terminology for credibility and accuracy
✅ 5 IA iterations showing rigorous design process

What happened after I left:

The navigation structure I designed launched and remains in use on Artfinder today. While I didn't measure post-launch metrics, the fact that it's still live suggests:

  • It solved the discoverability problem (users can now browse subcategories from homepage)

  • It improved SEO (more category pages exposed for crawling)

  • It aligned with business goals (stakeholders approved and shipped it)

What I'd Do Differently

1. Push for a phased launch with measurement

I designed the full IA but left before launch. If I could redo this, I'd negotiate to stay through launch week to:

  • Set up proper before/after analytics tracking

  • Run A/B tests on different category orderings

  • Validate the purchase-driven prioritization with real data

Learning: For IA projects, push to stay through launch + 2 weeks of data collection. Otherwise, you never know if your decisions were right.

2. Test category names with users, not just SEO

The SEO agency gave great keyword data, but I didn't validate category labels with actual Artfinder users. Did buyers understand "Impressionism" vs. "Impressionistic"? I assumed correct terminology = better UX, but didn't test it.

Better approach: Run 5-8 user interviews asking: "Where would you find X type of art in this navigation?" Validate assumptions.

3. Document the "visits vs. purchases" insight more thoroughly

This was the core strategic insight—yet I documented it quickly in a slide. Should have created a full analysis showing:

  • What % of "Nudes" visitors converted vs. "Portraits" visitors

  • Average order value by category

  • Repeat purchase rates by subject

Why: Stronger documentation = stronger case for future navigation decisions.

Key Learnings

Browse behavior ≠ purchase behavior

High traffic doesn't mean high revenue. Navigation should prioritize what converts, not just what attracts clicks. This seems obvious in hindsight, but most teams default to "put the most-clicked items first."

Shallow navigation hides inventory

Forcing users to click into broad listing pages and filter manually makes your marketplace feel smaller than it is. Exposing subcategories in navigation = perceived abundance.

Art world terminology matters

Incorrect labels ("Impressionistic") signal amateur curation. In niche marketplaces, domain expertise in language choices builds credibility with both buyers and sellers.

Iteration is the process

5 IA versions wasn't excessive—it was necessary. Each iteration refined the hierarchy based on different constraints (SEO, purchase data, stakeholder input, terminology accuracy). Version 1 was wrong. Version 5 was right.