The question we hear most often from reliability engineering teams in the first evaluation conversation is a variation on the same thing: "What does the first 90 days actually look like?" It is the right question. Utility procurement cycles are long, OT change management is deliberate, and field crews are busy. Teams evaluating a new monitoring platform want to know what they are committing to in terms of time, internal resources, and what they should realistically expect to see before they make a production decision.
This post is a direct answer to that question, based on how our pilot programs have actually gone. We are writing it because the sales cycle version of this answer — "quick to deploy, immediate value" — does not serve reliability engineers who need to plan their team's time and set realistic expectations with management.
What a Pilot Covers
A Fieldiq pilot program is scoped to a defined subset of the distribution fleet — typically 50–200 assets across one to three feeders, selected to include at least one feeder with known historical failures, at least one with known high-load or high-growth characteristics, and an asset mix that represents the fleet types the utility most needs monitoring coverage on. The goal is to generate enough anomaly detections within 90–120 days to validate the model's performance on the utility's specific fleet before a fleet-wide deployment decision is made.
The pilot scope is also scoped to a specific set of integration touchpoints — typically dashboard-only in the first 60 days, with SCADA/historian write-back integrated in the second phase if the pilot is progressing well. This keeps the initial deployment footprint small and the OT change management burden minimal while the anomaly model is being calibrated.
Timeline: What Happens When
Weeks 1–2: Asset Inventory and Sensor Siting
Before sensors are installed, we need a GIS export of the pilot feeder assets — transformer locations, ratings, installation years, and substation connectivity. This typically comes from the utility's ADMS or GIS system as a shapefile or CSV. The purpose is to pre-configure asset records in the platform so that when telemetry arrives, it is immediately mapped to an asset identity rather than appearing as an unlabeled data stream.
Sensor siting decisions are made during the week 1–2 period. For distribution transformers, the primary siting question is accelerometer placement — the sensor location on the tank wall that captures the strongest diagnostic signal for the target fault modes. For reclosers, the question is whether we are reading from existing controller event logs via DNP3 or supplementing with a retrofit temperature sensor.
Weeks 3–6: Sensor Installation and Initial Telemetry
Physical sensor installation is performed by the utility's field crews or by Fieldiq field engineering staff. Installation time per transformer is typically 45–90 minutes for the sensor and cellular gateway — no outage required for an external tank-mounted installation. Within 24–48 hours of installation, the platform shows live telemetry from that asset. The first 3–4 weeks of telemetry are the baseline calibration period. No anomaly flags are generated during this period — the model does not have a calibrated baseline to flag against yet.
Weeks 6–10: Model Calibration Completion and First Flags
After approximately 4–6 weeks of continuous telemetry, the per-asset models reach their calibration stability threshold. The first anomaly flags typically appear in weeks 6–10. In most pilot programs, 3–8% of the monitored fleet shows elevated anomaly scores within the first 60 days of active monitoring. Some of these are genuine incipient condition issues. Some are false positives that require investigation and baseline model adjustment.
We are not suggesting that every flag in weeks 6–10 warrants a truck roll. The appropriate response to an early-phase flag is a desk review: pull the DGA history for that transformer if available, check the outage management system for any recent events on that secondary circuit, and apply the reliability engineer's local knowledge about that asset's history. The flag is a starting point for investigation, not a dispatch trigger.
Weeks 10–16: Model Refinement and Integration Scoping
The second half of a standard 90-day pilot involves model refinement based on the initial flag outcomes. When a flagged asset is inspected and the inspection confirms a genuine condition issue, that outcome validates the model and increases the anomaly score calibration confidence. During this period, we scope the SCADA and historian integration pathway with the utility's SCADA and IT/OT teams. The connectivity architecture, firewall rules, and PI tag hierarchy are defined so that the integration can be executed without delay if the pilot proceeds to production.
Instrumentation: What You Actually Need
For distribution transformers, the minimum sensor configuration for meaningful condition monitoring is: one tri-axial accelerometer mounted on the tank wall in the recommended position, one surface-mounted temperature sensor, and a cellular gateway for uplink. Fieldiq does not manufacture sensors — the gateway is sensor-agnostic, compatible with standard MEMS accelerometers and thermocouple or RTD temperature sensors from multiple hardware vendors. The utility can purchase sensors through existing hardware procurement channels rather than through a proprietary hardware contract.
For reclosers with modern electronic controllers, the existing DNP3 controller data stream provides load current and operations counter data sufficient for the wear modeling. A retrofit temperature logger on the mechanism housing adds significant diagnostic value but is optional in the first phase.
Setting Realistic Expectations with Management
The expectation-setting conversation that matters most is not about the technology — it is about what constitutes a successful pilot outcome and what the evaluation criteria are before a fleet-wide deployment decision is made.
A well-designed pilot has a defined success threshold: something like "at least two confirmed anomaly detections with inspection confirmation in the 90-day window, false positive rate below 15% on inspected flags, and at least one SCADA/historian integration pathway validated." These criteria should be agreed before the pilot starts, not negotiated after it ends.
We have seen pilots succeed on these criteria and proceed to production. We have also seen pilots where the fleet subset selected did not include any assets approaching failure in the 90-day window — a circumstance that produces valid baseline calibration data and platform validation, but does not generate the confirmed detection outcome that many management teams want before committing to fleet-wide deployment. A fleet that is generally young and healthy is harder to demonstrate detection value on quickly than a fleet with known older assets in stressed loading conditions. This is an honest operational reality of the detection timing problem, and it is worth surfacing before the pilot scope is defined.
