How a high-volume casino resort reduced grave shift fills by 33% in its first week with Kiosk Intelligence

The challenge

At a 24/7 resort with 2,000+ gaming machines, overnight operations never stop — and neither did the kiosk fill runs. Grave shift teams made frequent trips to the floor every shift, reacting to low kiosks as they found them. It pulled staff away from higher-value work and put steady pressure on cage operations.

The goal wasn’t just fewer fills. It was to eliminate grave shift fill operations entirely. But without data-driven forecasting, the team had no way to know when a kiosk would actually run out of cash.

The solution

Kiosk Intelligence’s machine learning model analyzed the property’s own historical transaction data — denomination usage, transaction rates, and time-of-day patterns — to predict exactly when each kiosk would run out of cash.

Instead of sending staff to the floor on instinct, the team could now plan fills around forecasted depletion windows. Grave shift crews went only when and where they were needed. The reactive scramble gave way to a proactive schedule, one built around how this specific property actually operates, overnight included.

The results

Measurable impact in the first seven days

One week in, grave shift fills were already down 33% year-over-year — a meaningful early signal for a team working toward eliminating overnight fills altogether. And the model is still learning. As forecasting accuracy sharpens against the property’s real-time transaction patterns, those numbers are expected to improve further.

Results reflect Week 1 performance data following go-live. Forecasting accuracy is expected to improve as the model refines against the property’s real-time transaction patterns over time.