I started Fleetpio because I watched a single compressor failure wipe out $60,000 in fresh produce inventory at a cold storage client I was responsible for. The failure wasn't sudden. Reviewing everything afterward, the warning signs were there for weeks in the unit's sensor data. Nobody was watching it. There was no system that would have translated those readings into a scheduled maintenance visit before the unit failed. We just got paged at 2am, showed up, and then spent the next three days dealing with the insurance claim.
That was the moment I started building what eventually became Fleetpio. It took two years of learning the hard way what it actually takes to build useful predictive maintenance software for refrigeration fleets — not just sensor ingestion, but the complete workflow from anomaly signal to technician dispatch with correct parts. We're open for business now, and I want to share what we learned along the way.
Why This Problem Is Harder Than It Looks
When I started, my mental model of predictive maintenance was: sensors → data → alert → fix. That model is correct in principle. The execution is where it gets complicated.
The first thing we discovered is that refrigeration sensor data is far noisier than any of the academic literature on condition monitoring suggests. A commercial walk-in cooler compressor starts 10–20 times per day. Each startup has a vibration transient. Defrost cycles cause temperature and pressure swings that look alarming if you don't account for them. Different units of the same model have different baseline signatures because of installation differences, refrigerant charge history, and the specific wear state they started the monitoring period in.
If you apply a single threshold to a fleet of 30 units, you will generate hundreds of false alerts per month. Within two weeks, the operations manager will have learned to ignore the alert emails. You've spent money on sensors and software and made the problem worse because the system cried wolf.
What works is per-unit baselines, built during a calibration period, with persistence logic that requires an anomaly to appear consistently over time before triggering a work order. The statistical work to get that right across different compressor types, refrigerant systems, and operating environments took most of the first year.
The Dispatch Problem We Didn't Expect
The second thing we learned is that generating a good alert is necessary but not sufficient. The alert needs to land in a workflow that produces a scheduled service visit before the failure window closes.
We ran an early version of Fleetpio with a few pilot users where we were generating alerts but leaving dispatch to the operator's existing workflow. Alerts went into email. Email went into an inbox. Inbox got checked sometimes. By the time a work order was created and a technician was scheduled, the three-to-four-week lead time we'd carefully detected had compressed to one week or less — and in several cases, the unit failed before the visit happened.
The lesson: predictive maintenance software that doesn't connect directly to dispatch workflow isn't really predictive maintenance software. It's an alert dashboard. Alerts sitting in a dashboard that requires a human to manually create a work order will always lose to the urgency of today's reactive calls. The lead time we generate is only valuable if the scheduling workflow uses it automatically.
That realization sent us back to build the work order automation layer — the part of Fleetpio that takes a confirmed anomaly trend, generates a work order with failure mode classification and parts recommendation, and puts it directly into the dispatch queue with a scheduling deadline. That took another six months to build correctly.
Why We Built for Food Logistics Specifically
Predictive maintenance is a broad market. HVAC, industrial machinery, automotive, wind turbines — the sensor monitoring and ML principles are similar across domains. So why commercial refrigeration for food logistics?
Partly personal: my background is in cold storage field operations, not industrial manufacturing. I understand the specific operating environment — the duty cycles, the refrigerant types, the failure modes, the compliance pressure that food safety regulations create, the economics of a food logistics operator trying to avoid both spoilage events and expensive service contracts.
But there's also a structural reason. Commercial refrigeration for food logistics occupies a gap between two existing categories. Industrial-scale condition monitoring (Augury, Fluke, Emerson's condition monitoring systems) is built for large rotating equipment — pumps, turbines, motors — and priced for enterprise manufacturing. A food distribution operator with 50 refrigeration units isn't the target customer for those platforms. On the other side, generic CMMS software (Fiix, UpKeep) handles work order management but doesn't generate condition-based alerts from sensor data. Our customers were stuck choosing between expensive enterprise tools that don't fit their scale and generic maintenance software that doesn't address the problem they actually have.
Fleetpio is purpose-built for the 20-to-200-unit refrigeration fleet running in a food logistics context. The compliance documentation output is designed to support FSMA preventive controls records, not generic maintenance logs. The alert calibration is specific to commercial refrigeration compressors — scroll and reciprocating types at the duty cycles and refrigerant systems common in food distribution. The pricing is per fleet unit rather than per seat, which aligns better with how these operators think about their asset base.
What We've Learned About Customer Problems
Two years of building and talking to fleet managers and service operations managers have clarified what the real problems are — which sometimes differ from the problems I thought I was solving when I started.
Emergency call reduction is the most visible ROI driver, and it's real. But what operations managers talk about more than emergency costs is the mental load of uncertainty. Not knowing which units in a fleet are healthy and which are developing problems. Getting called at 2am. Fielding the call from a client whose delivery was late because a unit went down overnight. Managing an operation where the equipment can fail at any moment and you have no early warning. Fleetpio doesn't just reduce the frequency of those events — it creates visibility that changes the job from reactive firefighting to systematic planning. That change in operational posture is something operations managers describe as valuable independent of the dollar savings.
The other thing we've learned: food safety audit preparation is a more significant benefit than we initially understood. When an auditor asks for maintenance records and temperature excursion documentation, operators who use Fleetpio can pull that audit trail from the system in minutes. Operators who are managing maintenance records across paper logs, technician emails, and a partially populated CMMS spend hours preparing for audits and worry about what gaps they might have missed. That's time and stress that directly affects how these operators run their business.
What's Next
We're focused on the Mountain West initially — Colorado, Utah, New Mexico, Arizona. That's where we know the operating environment, where our initial customer relationships are, and where the food logistics density supports building a fleet of reference customers who can speak to real outcomes.
We're a small team. Priya runs sensor engineering and anomaly detection, Carlos runs product and has a field service background that keeps us honest about what's actually useful in a dispatch workflow. I'm running the business and staying as close to customers as possible — still going on-site with new customers to watch how they use the product and where the friction is.
If you run a refrigeration fleet for a food logistics operation and the problem we're describing is one you're dealing with, reach out. We'd rather have a direct conversation about your specific fleet than run you through a sales script.
