Machine Losing Money Sao Paulo How to Trace Revenue Loss to Specific Time Windows
The most common complaint I hear from Sao Paulo operators is a version of this: “My machines are losing money. I can see it in the monthly totals — revenue is 15-20% below what it should be based on play volume — but I cannot identify when the loss is happening or which machine is affected.” Without time-window and machine-specific data, the operator cannot distinguish between a cheater operating on specific days, a machine with a hardware problem developing over weeks, or a systemic issue like power quality degrading over the month. The investigation stalls before it starts.
This article provides a practical method for tracing revenue loss to specific time windows and specific machines, based on the diagnostic approach I use when I arrive at a Sao Paulo venue for the first time. The method requires bus monitors for data collection but does not require full-time monitoring.
Step 1: Establish the Baseline — What Revenue Should Look Like Per Machine Per Hour
Before you can identify a loss, you need to know what normal looks like. Install bus monitors temporarily on the 5 highest-revenue machines for 7 days and collect hourly data: total credits played per hour, total payouts per hour, and payout percentage per hour. From this data, calculate the baseline for each machine: average credits per hour for each hour of the day (10:00 AM, 11:00 AM, etc.), average payout percentage for each hour, and standard deviation of payout percentage per hour — the range of normal variation. A machine that normally pays 80% with a standard deviation of 5% should raise an alert at 90% or above.
The baseline reveals patterns that are invisible in monthly totals. A machine may show normal monthly revenue but have a consistent dip every Thursday evening — specific to one day and one shift. Without hourly baseline data, that Thursday evening dip would never be identified because it averages out across the month.
Step 2: Compare Current Week to Baseline — Identify Anomalous Windows
After establishing the baseline, compare the current week’s hourly data for the same machines to the baseline. Flag any hour where: credits played deviate by more than 2 standard deviations from the baseline mean (a significant increase may indicate credit injection, a significant decrease may indicate player avoidance due to machine malfunction), payout percentage exceeds the baseline mean plus 2 standard deviations (potential payout manipulation), and consecutive hours of anomaly — 2 or 3 consecutive anomalous hours is a stronger signal than a single anomalous hour.
The flagged hours form a time window that tells you when to review surveillance video. Instead of reviewing 7 days x 12 hours = 84 hours of video, you review 3-5 flagged hours. This is the efficiency benefit of time-window identification.
Step 3: Cross-Reference With Surveillance Video for the Flagged Windows
Review surveillance video only for the flagged hours. Look for the same indicators described in the Rio tourist area article: unusually consistent wins, phone or device near machine, touching of non-player areas, staff interactions. Because you are reviewing only 3-5 hours instead of 84 hours, the review can be completed in a single afternoon rather than over several days.
If the video shows suspicious activity at the exact times flagged by the bus data, you have a cheating case with both data and video evidence. The two sources corroborate each other and form the foundation of a police report if you choose to pursue legal action. If the video shows no suspicious activity but the flagged hours persist, the cause is environmental or hardware-related.
Step 4: Environmental Testing During Flagged Hours
If the flagged hours are concentrated at specific times of day — always during the afternoon, always during the evening shift — perform environmental testing specifically during those hours. RF spectrum analysis during the flagged hours rather than at a random time (1,500-3,000 BRL for a specialist visit during a specific time window), power quality recording during the flagged hours (the analyzer records continuously, so it captures the entire day, but the analysis focuses on the flagged hours), and internal temperature and humidity logging during the flagged hours.
The time-specific environmental testing has higher diagnostic yield than generic testing. An RF signal present only during the evening hours — when a neighboring business operates its equipment — will be captured only if the spectrum analysis is performed during those hours.
Step 5: Machine-Specific Isolation
If the flagged hours affect one machine disproportionately but several machines share the same power circuit and physical location, the problem is likely machine-specific — component aging, connector oxidation, or a configuration issue. Isolate the machine by: swapping the machine with another machine of the same model from a different location, monitoring the swapped machine at the new location, and monitoring the replacement machine at the problem location. If the problem follows the machine, the cause is internal (hardware or configuration). If the problem stays at the location, the cause is external (environmental, power quality, or cheating). This swap test is one of the simplest and most powerful diagnostics available — it costs nothing except machine movement labor and definitively separates internal from external causes.
Frequently Asked Questions
Q: How long does the full 5-step tracing process take?
A: Step 1 (baseline): 7 days of data collection, but it runs automatically — no technician time. Step 2 (comparison): 1-2 hours of analysis. Step 3 (video review): 2-4 hours of reviewing flagged footage. Step 4 (environmental testing): 4-6 hours of specialist time. Step 5 (machine swap): 1-2 hours of labor. Total elapsed time: 8-10 days, total technician time: 10-15 hours spread across the period.
Q: Can I perform this process with only one bus monitor rotated across machines?
A: Yes, but it takes longer. Monitor machine 1 for 7 days, then machine 2 for 7 days, etc. For 5 machines, the baseline alone takes 5 weeks. If budget allows, purchasing 5 bus monitors for the baseline period reduces the timeline from 5 weeks to 1 week. After baseline, 1-2 monitors for ongoing monitoring is sufficient — the baseline data has already been established.