The operator asked not to be named. Every figure below comes from that operator's own pumper logs, shut-in records, and workover invoices, compared against the same wells before installation. Results on your leases will depend on well count, failure mix, and how alerts are staffed.
The problem the operator brought us
Forty rod-pumped wells across several Midland-area leases, run by three pumpers and no SCADA staff. The controllers on the units would shut a well in on a fault condition, and nobody knew whether that meant a failing gearbox or a well that had simply pumped off until someone drove out and looked. Wells sat down for shifts at a time for no mechanical reason at all.
The operator was not looking for analytics. They wanted to stop paying for trips to healthy wells and to stop being surprised by the failures that were real. That framing set the pilot's success metric: not model accuracy, but how many shut-ins turned out to be nothing, and how much warning the crew got on the ones that were something.
How the pilot ran
- Baseline — six months of the operator's own pumper logs, SCADA shut-in records, and workover invoices established a per-well downtime baseline before any hardware went out. No modeled baseline was used.
- Install — HL-P200 pressure sensors on the flowline and casing, three-axis vibration on the pumping unit gearbox, and an HL-M5 cellular modem with Starlink failover per pad. Install averaged under an hour per well with no lease power work.
- Learn — three weeks of continuous data per well to establish each unit's own normal (stroke signature, current draw, discharge pressure band) before alerting was enabled. Alerts route to the pumper's phone as assignable work items, not dashboard charts.
- Operate — months 4–6 ran under live alerting. Every alert was dispositioned by the pumper as confirmed fault, process change, or false positive, and that feedback fed back into the per-well thresholds.
What we learned
Most shut-ins were process, not mechanical. The majority of pre-pilot shut-ins the operator recorded as equipment faults turned out to be pump-off and fluid pound — process conditions the controller read as a failure. Separating those from real mechanical drift is where the 62% reduction in false shut-ins came from.
Vibration caught what pressure alone missed. Gearbox bearing wear and rod-string wear showed a detectable vibration trend weeks before any pressure or production signal moved. Pressure-only monitoring on the same wells would have flagged those units only after they had begun to lose production.
Backhaul was the hidden failure mode. Two pads sat in marginal cellular coverage where the prior telemetry dropped for hours at a time. Starlink failover on the HL-M5 kept those wells reporting, which is the difference between a monitoring program and a monitoring program with gaps.
Value showed up as avoided truck rolls. The operator's own accounting of the pilot put the largest single line of savings not in avoided workovers but in trips that were never made — pumpers stopped driving to wells that were not actually down.
What was deployed
Nothing custom — the same catalog hardware any operator can order, running the same models described on our predictive maintenance for oil & gas page.
- HL-P200 pressure sensors — flowline and casing pressure at high sample rates, the signal that separates a pump-off from a restriction.
- Three-axis vibration — gearbox and pumping-unit vibration signatures, the earliest indicator of bearing and rod-string wear.
- HL-M5 modem with Starlink failover — cellular backhaul that fails over automatically when the tower drops, so marginal-coverage pads keep reporting. See the full catalog.
Run the numbers on your own wells
Before you evaluate any monitoring vendor, put a dollar figure on what unplanned downtime already costs you. Our downtime cost calculator and breakeven calculator use your inputs, not ours — and the education guide walks through how those numbers flow into asset value.