E-Commerce Personalization at Scale
Designing Conversion Strategies for Wallis
Skill Constellation
Primary
Supporting
Emerging
The project focused on bridging the gap between high-volume marketing and generic website experiences. For Wallis, the challenge wasn't a lack of traffic; it was that a vast number of shoppers saw exactly the same experience, leading to significant revenue leakage.
As a Personalization Specialist embedded within Qubit's London team, I designed and implemented real-time behavioral interventions for the Wallis website, turning generic sessions into contextually relevant journeys.
Significant annual traffic across the entire online store.
Direct revenue growth via systematic behavioral triggers and campaign mirroring.
What the Data Revealed
Before designing interventions, I mapped behavioral data across the portfolio. Three critical friction points emerged as the primary sources of conversion loss.
Funnel Diagnostics
Mapped end-to-end behavioral data to identify the three highest-impact friction zones in the conversion funnel.
Evidence: Diagnosed 3 friction patterns — disconnected journeys, dead ends, and one-size-fits-all — that accounted for majority of revenue leakage.
1. The Disconnected Journey
Email campaigns drove massive spikes, but the website remained "unaware" of why the visitor was there. We found that discount code usage was significantly lower than open rates because the site failed to remind customers of the offers they'd clicked on.
2. Dead Ends
Empty bags and zero-results search pages functioned as exits. Instead of recovering the session, the site simply displayed a "Sorry" message, abandoning thousands of daily visitors at the point of intent.
3. One Size Fit Nobody
First-time international visitors saw the same homepage as loyal UK customers. Despite having the data, the sites provided no acknowledgement of history or context.
The diagnosis was clear: these websites weren't broken. They were generic. And in e-commerce, generic is broken — it just breaks quietly, one lost session at a time.
Building a Systematic Approach
To move beyond scattered A/B testing, I developed a four-layer framework that addressed friction at every stage of the funnel. This allowed us to deploy interventions systematically with speed and rigour.
Behavioral Segmentation
Designed audience segments from behavioral signals, not just demographics. Real-time triggers responded to in-session actions.
Evidence: Geo-specific social proof, threshold nudges, and history-aware personalization.
Real-Time Personalization
Built and deployed interventions that responded to in-session behavior — not pre-defined user segments.
Evidence: Four-layer framework deployed across all key funnels.
Rigorous Execution
Implementation wasn't just about UI changes—it was about statistical rigour. We introduced 10% blank-ad control groups (placebos) to calculate true incremental ROI, moving from correlation to causation.
We synchronized on-site messaging with a relentless daily promotional calendar, ensuring creative was always aligned with the trade plan.
A/B Testing & Statistical Rigor
Implemented 10% blank-ad control groups (placebos) to isolate true incremental lift from confounds.
Evidence: Shifted measurement from correlation to causation for every intervention.
Measuring the Lift
Significant revenue lift driven by targeted segments and geo-social proof.
Vast reduction in leaked conversion via mirroring.
Increased engagement through orienting onboarding.
Systematic deployment across all key funnels.
Personalization is helping us build relationships with our customers that translate into increased engagement, loyalty and revenue.
— Simon Pritchard, Arcadia Group Digital Director
Key Learnings
Complexity ≠ Impact
The most impactful interventions were often the simplest: code reminders and dead-end recovery. Clarity consistently beat sophisticated algorithms in driving immediate ROI.
Honest Scarcity Wins
Fake countdowns erode trust. Grounding urgency in real inventory data ensured that our signals remained useful and respected over the long term.
The Audience Lesson
This project taught me that personalization isn't just about adding more—it's about adding the right thing. For various segments, particularly less digitally native users, "good personalization" meant restraint and orientation. It shifted my perspective from extracting value to genuinely helping the user, making "Is this better for the person?" my primary design filter.
This case study documents personalization design conducted with Qubit for Wallis. Proprietary metrics have been generalized for confidentiality.