Description
COMPANY Currys | YEAR 2020 |
ROLE Lead UX Designer | IMPACT 3x conversion lift |
The Problem
In 2020, Currys' in-store staff could recommend the perfect laptop case or HDMI cable — but online, customers saw static product pages with no guidance. Cross-sell rates were low, cart abandonment was high, and we were leaving money on the table.
The business wanted to replicate in-store expertise at scale through machine learning. My role was to design how those recommendations would actually work for users.
What I Did
I led the UX design for "Supercharge Attach" — a machine learning recommendation engine built in partnership with Syntasa.
My responsibilities:
Designed the recommendation interface patterns (bundles, cross-sells, contextual suggestions)
Created the UX framework for personalized vs. "natural attach" recommendations for new visitors
Defined placement strategy through A/B testing
Collaborated with data science to translate ML capabilities into user-facing experiences
I didn't build the algorithms — but I designed everything users actually saw and interacted with.
Key Design Decisions
1. Personalization should feel helpful, not creepy
We tested multiple presentation styles. Recommendations framed as "Customers like you bought..." outperformed "Based on your browsing history..." because it felt like social proof rather than surveillance.
2. New visitors need a fallback
Not everyone has browsing history. I designed a "natural attach" system using product relationship data so first-time visitors still received relevant suggestions — no cold start problem in the UX.
3. Context matters more than completeness
Recommendations on the product page performed 3x better than recommendations in a dedicated "suggested products" section. Users wanted help in the moment of decision, not a separate shopping task.
Results
METRIC | BEFORE | AFTER |
Product coverage with recommendations | 32% | 72% |
Add-to-basket rate (personalised users) | Baseline | 3x higher |
Add-to-basket rate (new visitors) | Baseline | 1.3x higher |
The system now serves millions of users daily and became the foundation for future personalisation initiatives at Currys.
What I Learned
The best AI is invisible. Users don't care about your nearest-neighbor model — they care about finding the right HDMI cable without hunting for it. My job was to make the technology disappear into a naturally helpful experience.
My Role
✓ UX design for recommendation interfaces
✓ Placement strategy and A/B testing
✓ Collaboration with Syntasa's data science team
