Quilmes runs 240,000 branded coolers across every corner of Argentina. Three to five percent vanish every year — and the alleged thieves were often the shop owners themselves. Recovery took weeks, when it happened at all. Everyone called it a theft problem. It took a former database analyst about a month to see it wasn’t one.
Six years of databases, first month of design
This was my first product design role. I arrived after six years as a database analyst, and the discovery research kept surfacing something familiar: nobody could say where any cooler should be. Inventory lived in warehouses and died in shipment. Maintenance swaps returned a different machine than the one that left. The legal papers linking a cooler to a shop owner were disconnected from both. That’s not a theft problem — it’s a broken chain of custody. A data integrity problem wearing a crime costume.
The 80/20 was a QR code
Before designing anything sophisticated, we solved the most basic problem that addressed 80% of the issue: every cooler got a QR code, scanned by the rep at delivery — activating it on site, linking machine, legal documentation, point of sale, and geographic position in one gesture. For the first time, a complete inventory existed. You can’t recover what you can’t locate.
Three channels, because people watch what they already watch
When a cooler goes dark, the first hours decide the recovery. The protocol notified three audiences through the channels they already monitored: field reps got the alert as a task inside Bees, ABInBev’s own system — worth points toward their monthly objectives, so first response was incentivized, not requested; CAPEX management got email; and every status change landed in the platform as an audit trail. If recovery failed, the paper trail made legal action viable — that was a core feature, not an afterthought.
Watching users ignore the feedback
Usability testing with CAPEX staff and a regional trade manager confirmed the layout decisions — highlight states between map and data table read as intuitive, settings were found without guidance — and produced one humbling finding: nobody read the system feedback. Confirmation messages were invisible. It reshaped how I’ve designed feedback ever since: if it matters, it can’t be polite.
Results, and what’s projection
Awareness-to-action went from two weeks to a three-hour average. Traceability went from nonexistent to complete. The $250,000 in annual savings is a projection, not an audited number — built from recovered reposition costs and unblocked sales, and I’d rather label it than launder it.
Two weeks → three hours
Awareness-to-action, average
$250K / year
Projected savings · recovered reposition costs + unblocked sales
The honest part
Not everything I explored at Sensify shipped. A redesign of the platform’s alert system stayed an exploration — the fatigue problem that motivated it became a field note instead: Alert Systems at Scale: When the Cure Is the Disease. The theft protocol shipped because it obeyed the lesson that took me six years of databases to learn and one project to apply: solve the record before the alarm.