Empty Cart
adidas International B.V.
AI Personalisation
A Dead End, Redesigned
An empty cart is easy to dismiss as a non-state. Nothing there, nothing to design for. But at the scale adidas operates, that moment is anything but neutral; it's where millions of journeys quietly stall.
The business read it as a conversion problem. The fix seemed obvious: fill the space with products, push harder, recover the session. But before committing to that direction, the team took a step back.
Session analytics, funnel data, and qualitative research pointed to something more nuanced. A significant share of users landing on an empty cart weren't disengaged, they were mid-journey. They'd returned after a break, cleared their cart intentionally, or switched devices. The empty cart wasn't the end of a session. It was an experience dead end: a moment with no clear path forward, no context, no continuity.
An empty cart is easy to dismiss as a non-state. Nothing there, nothing to design for. But at the scale adidas operates, that moment is anything but neutral; it's where millions of journeys quietly stall.
The business read it as a conversion problem. The fix seemed obvious: fill the space with products, push harder, recover the session. But before committing to that direction, the team took a step back.
Session analytics, funnel data, and qualitative research pointed to something more nuanced. A significant share of users landing on an empty cart weren't disengaged, they were mid-journey. They'd returned after a break, cleared their cart intentionally, or switched devices. The empty cart wasn't the end of a session. It was an experience dead end: a moment with no clear path forward, no context, no continuity.
That reframe shifted the design question entirely
from
"how do we fill the cart faster"
to
"how do we help users re-enter their own journey?"
from
"how do we fill the cart faster"
to
"how do we help users re-enter their own journey?"
Setting the North Star
Before any solution framing, a North Star vision was defined for the empty cart experience; one that the whole team could orient around. The guiding principle was simple but deliberate: the empty cart should never feel like a dead end. It should feel like a system that knows who you are and where you were heading.
This vision reframed the empty state from a gap in the funnel into an active touchpoint, a moment to demonstrate that adidas understands its customers beyond the transaction. From there, success metrics were defined not just around conversion, but around re-entry rate, session continuation, and personalisation relevance. Having that shared North Star made prioritisation decisions easier and kept the team aligned when competing stakeholder interests pulled in different directions.
Before any solution framing, a North Star vision was defined for the empty cart experience; one that the whole team could orient around. The guiding principle was simple but deliberate: the empty cart should never feel like a dead end. It should feel like a system that knows who you are and where you were heading.
This vision reframed the empty state from a gap in the funnel into an active touchpoint, a moment to demonstrate that adidas understands its customers beyond the transaction. From there, success metrics were defined not just around conversion, but around re-entry rate, session continuation, and personalisation relevance. Having that shared North Star made prioritisation decisions easier and kept the team aligned when competing stakeholder interests pulled in different directions.
Aligning Before Building
Marketing, Product, and Engineering each had different instincts about what this space should do. These tensions were surfaced early through structured workshops: mapping user intent signals, ranking opportunity areas by impact and feasibility, and landing on trade-offs the whole team owned together.
The output wasn't a wireframe. It was a shared decision framework; a North Star for the interaction, agreed KPIs, and explicit choices about what this moment should and shouldn't try to do. That upstream clarity is what made execution fast and late-stage scope changes rare.
Marketing, Product, and Engineering each had different instincts about what this space should do. These tensions were surfaced early through structured workshops: mapping user intent signals, ranking opportunity areas by impact and feasibility, and landing on trade-offs the whole team owned together.
The output wasn't a wireframe. It was a shared decision framework; a North Star for the interaction, agreed KPIs, and explicit choices about what this moment should and shouldn't try to do. That upstream clarity is what made execution fast and late-stage scope changes rare.
The Personalisation Approach
The core design decision was to replace a generic empty state with a dynamically personalised re-entry point: one that reflected this user's journey, not a default product grid.
Using signals from across the user's history (previous orders, returns, exchanges, saved items, size preferences, and style patterns) the experience was designed to surface what was actually relevant at that moment. Not a recommendation engine bolted onto an empty page, but a considered set of re-entry paths that made the experience feel oriented rather than blank.
Research shaped how this was approached. Personalisation works best when it's quiet, contextual, and just there: reducing effort, confirming intent, helping users move forward. Discomfort doesn't come from data use itself; it comes from surprise, over-specificity, or timing that feels off. That finding drove the design direction: broad, task-aligned signals over high-precision, high-visibility ones. Relevance that feels natural, easy to act on, easy to ignore.
The logic also accounted for context. A user returning mid-checkout needed different signals than someone opening a fresh session. Exception states were mapped and designed for individually, rather than collapsed into a single template.
The underlying principle: an empty cart should feel like picking up a conversation, not starting over.
The core design decision was to replace a generic empty state with a dynamically personalised re-entry point: one that reflected this user's journey, not a default product grid.
Using signals from across the user's history (previous orders, returns, exchanges, saved items, size preferences, and style patterns) the experience was designed to surface what was actually relevant at that moment. Not a recommendation engine bolted onto an empty page, but a considered set of re-entry paths that made the experience feel oriented rather than blank.
Research shaped how this was approached. Personalisation works best when it's quiet, contextual, and just there: reducing effort, confirming intent, helping users move forward. Discomfort doesn't come from data use itself; it comes from surprise, over-specificity, or timing that feels off. That finding drove the design direction: broad, task-aligned signals over high-precision, high-visibility ones. Relevance that feels natural, easy to act on, easy to ignore.
The logic also accounted for context. A user returning mid-checkout needed different signals than someone opening a fresh session. Exception states were mapped and designed for individually, rather than collapsed into a single template.
The underlying principle: an empty cart should feel like picking up a conversation, not starting over.
Outcomes
Unmoderated usability testing validated the direction before launch. Post-launch analytics confirmed it.
Cart exit rate dropped from 40% to under 30% +4,000 incremental cart additions in the tracked period +0.7% CVR uplift across targeted segments. Beyond the conversion metrics, the project established a scalable personalisation pattern that could be extended into future post-purchase and re-engagement experiences — a reusable foundation, not a one-off fix.
Interested in the full case study? Contact us for walking through the case study in detail or explore how similar thinking can be applied to your organization.
Unmoderated usability testing validated the direction before launch. Post-launch analytics confirmed it.
Cart exit rate dropped from 40% to under 30% +4,000 incremental cart additions in the tracked period +0.7% CVR uplift across targeted segments. Beyond the conversion metrics, the project established a scalable personalisation pattern that could be extended into future post-purchase and re-engagement experiences — a reusable foundation, not a one-off fix.
Interested in the full case study? Contact us for walking through the case study in detail or explore how similar thinking can be applied to your organization.
