The Challenge
Crestone Water & Sewer District serves a remote mountain community at 8,000 feet elevation in Colorado's San Luis Valley. With only two certified operators managing both water and wastewater systems, the district faced a critical problem: they couldn't trust their SCADA data.
Lead Operator Maria Rodriguez described the situation: "We were getting 35 to 45 alarms every night. High pressure. Low chlorine. Pump failure. Tank level. After years of checking these alarms and finding nothing wrong 80% of the time, we'd started ignoring them. But we knew that was dangerous—one day, a real emergency would come through and we'd miss it."
Specific Pain Points
Sensor drift was invisible. The district's chlorine analyzer was gradually reading 0.3 ppm high, but operators had no way to detect this drift. They discovered it only during annual calibration—by which point they'd been over-dosing chlorine for months, wasting chemicals and money.
No audit trail for troubleshooting. When treatment performance degraded overnight, operators couldn't determine what had changed. "Was it a valve someone adjusted? A setpoint that drifted? We had no record of what happened, so we were guessing," Maria explained.
Alarm overload created paralysis. With dozens of alarms daily, operators couldn't distinguish signal from noise. Critical issues like actual pump bearing wear were buried among false temperature alerts and spurious pressure spikes.
"I was spending 2-3 hours every shift investigating alarms that turned out to be sensor glitches. That's time I should have been doing preventive maintenance or optimizing our processes. Instead, I was chasing ghosts."
The Solution: Data Trust Foundation
In January 2025, Crestone became one of Upstream AI's first Colorado pilot utilities, deploying the Trust Foundation tier. The implementation took a single afternoon—installing the edge agent on their existing SCADA PC and configuring data connections.
30-Day Baseline Period
For the first month, Upstream AI simply observed. The system learned Crestone's specific operational patterns: when operators typically ran pumps, what constituted "normal" pressure fluctuations, how chlorine residual varied by time of day and season.
"They told us not to expect anything for 30 days," Maria said. "That was actually reassuring—we weren't guinea pigs for some untested system. They were taking time to understand OUR specific operation, not just applying a generic template."
Trust Scores Changed Everything
On day 31, Maria logged into the dashboard and saw something different: every sensor reading now had a 0-100% trust score displayed next to it.
The chlorine analyzer that had been drifting? Trust score: 42%. The system flagged it immediately with a note: "Probe reading inconsistent with free chlorine calculated from dose rate and contact time. Recommend calibration check."
A pressure sensor that had been triggering false high-pressure alarms? Trust score: 28%. Note: "Sensor shows step-change response to temperature, indicating electrical interference. Physical pressure gauges read normal."
"Suddenly I wasn't guessing anymore," Maria explained. "When I saw Trust: 94%, I knew I could act on that data. When I saw Trust: 35%, I knew to verify manually before making any process changes."
The Results
Alarm volume: 42/night → 2.5/night
94% reduction in alerts. The remaining 2-3 alerts per night were genuine issues requiring attention, not false positives from sensor drift or electrical noise.
Chlorine probe drift detected early
The system flagged the drifting analyzer at 0.2 ppm error—before it became a compliance or cost issue. Previous detection method (annual calibration) would have missed 11 months of over-dosing.
Pump bearing wear predicted 6 days early
Vibration and temperature trends indicated bearing failure in Well Pump 2. Operators scheduled a planned bearing replacement during normal hours for $2,800. Emergency replacement would have cost $18,000+.
Operator confidence restored
"I'm sleeping through the night now," Maria said. "I know if my phone rings, it's real. And I know that if something's trending wrong, I'll get advance warning—not a 3 AM emergency."
Unexpected Benefits
Board presentation improvement. At monthly board meetings, Maria can now show infrastructure health scores and trending data. "The board used to ask 'Is everything okay?' and I'd say 'I think so.' Now I can show them: Pump health 87%, treatment efficiency 91%, sensor data quality 94%. They trust our operation more because we have data to back it up."
Compliance documentation. When state inspectors visited for the annual sanitary survey, the complete audit trail impressed them. Every operational decision—valve adjustments, pump startups, chemical dose changes—was logged with timestamp and operator ID. "They said it was the best documentation they'd seen from a utility our size," Maria noted.
Training new operators. When Crestone hired a part-time assistant operator in March, the audit trail became a teaching tool. "I can show him 'This is what I did and why' for any situation in the past 90 days. It's like having a playbook of real operational decisions with full context."
"The trust scores saved us from a compliance violation. Our pH probe was reading 7.2, which was fine. But the trust score was 31%. When we calibrated it, the actual pH was 6.9—just barely in compliance. Without that warning, we would have kept trusting the bad reading and eventually dropped below our permit limit."
ROI Analysis
Annual Cost: $7,200 (Trust Foundation tier at $600/month)
Documented Savings (First 6 Months):
• Prevented pump bearing failure: $18,000
• Reduced chemical over-dosing (chlorine probe drift caught early): $2,400
• Operator time saved (38 hrs/month × $35/hr × 6 months): $7,980
• Avoided false alarm truck rolls (estimated 2-3/month × $200 × 6 months): $3,000
Total documented savings: $31,380
Return on Investment: 436% in first 6 months
District Manager Tom Chen added: "The $7,200 annual cost was easy to justify to our board because we prevented one major failure in the first six months that would have cost more than twice that. Everything else is pure savings."
What's Next
After six months on the Trust Foundation tier, Crestone is evaluating an upgrade to the Operator Co-Pilot tier in 2026. "Now that we trust the data, we're ready for the AI predictions," Maria said. "If the system can predict equipment failures and treatment anomalies with the same accuracy it's shown for data quality, that's a huge operational advantage for a two-person shop like ours."
The district is also exploring the Energy Optimization add-on. "Our electricity bill is about $48,000 a year," Tom noted. "If we can save even 10% through intelligent pump scheduling, that's $4,800 with no upfront cost. The shared savings model makes it a no-brainer."