H&M · frog Design
Locate in Store: bridging online browsing with in-store stock
Customers couldn't trust that items shown online were actually in stock at their local store. I traced the problem to stale inventory data and paired an interactive map with real-time shelf scanning.
- Role
- UX Strategist
- When
- 2021 – 2023
- Also on Behance
- Original deck ↗
Want to ask about this project?
AI answers grounded in Anshula's CV and portfolio.
H&M was proactively experimenting with a way to bridge online and offline shopping, betting that the boundary between the two channels was about to matter a lot more. 'Locate in Store' started as that bet: could customers move seamlessly between their preferred H&M store and the online platform, and easily find what they wanted once they got there.
The problem
Research across the US, China, and Sweden confirmed the bet was right. In Sweden, 15% of customers who mentioned the store described it as "messy," and 6% said products were hard to find. In China, feedback was harsher: stores were described as feeling "more like a warehouse than a fashion store."
But the deeper issue, surfaced through staff interviews, was that the in-store inventory system itself only updated once or twice a day. Any navigation feature built on top of stale data would confidently point customers toward stock that might already be gone.
The constraint
Three approaches were piloted in real stores: RFID tags for product tracking, smart cameras for shelf monitoring, and autonomous store bots that could scan aisles and assist customers. Each addressed a piece of the problem, but the team had limited time to ship something that actually fixed the root cause rather than just adding a nicer interface on top of the same outdated system.
The tension, and the call I made
The obvious move was to build the map and ship it. Customers had been asking for exactly that. But a map is only as good as the data behind it, and an inventory system updating once or twice daily meant the map would inherit that staleness.
I made the call to pair the map with store bots, which physically scanned shelves to keep inventory data current in real time, treating real-time accuracy as a precondition for the map being trustworthy, not a nice-to-have layered on afterward. RFID was scoped for a later phase.
Approach
I led customer research across the US, China, and Sweden, interviewing both customers and store staff. That research is what surfaced the real root cause. Staff didn't describe the map as the missing piece; they described the inventory system itself as outdated. That reframing, from "customers need a map" to "the underlying data needs to be real-time first," is what shaped the final solution.
What shipped, tied to outcome
- Store bots scanning shelves in real time → drove the 15% inventory accuracy increase within the first two months, directly solving the staleness problem staff had flagged.
- Interactive map built on that now-accurate data → drove the ~12% increase in cart conversion on located products and a 25% reduction in overall product search time.
- A/B testing and painted-door tests post-launch → validated real usage patterns and surfaced further refinements.
- Experimentation metrics and decision frameworks built alongside it → cut pilot-to-deployment time by 40%, letting the team assess and scale high-potential retail concepts faster.
What I'd revisit
Risks around impact on online orders and balancing digital and physical experience were identified but not fully resolved at launch. They needed ongoing team and stakeholder discussion rather than being solved upfront.
At a glance
- Improved Inventory Accuracy by
- ↑0%
- Improved Cart Conversion by
- ↑0%
- Reduced Search Time by
- ↓0%
* Results measured across targeted pilot stores.
Want to talk through this kind of work? Let's talk.
More work
