PRISM learns the normal operating signature of every pump, well, plant and lift station from the SCADA you already collect. It then flags the ones drifting toward failure, with months of lead time to plan the work. Your crews get a work order, not another alarm.
The ranked list as an operator sees it. One explained list replaces the flood of threshold alarms, and every flag is triaged before a person is paged. Site and utility names anonymized.
SCADA fires after a threshold is crossed, not before. By the time the high-wet-well alarm sounds, the pump is down and the call-out clock is running. The drift that led there was readable for months.
Stations throw constant alarms on conditions that are perfectly normal for that site. Operators learn to tune them out, and then get surprised by the failure buried in the noise.
A failed pump pulls in after-hours call-outs, overtime, vendor premiums, bypass pumping, cascade damage to drop pipe and controls, boil-water notices and clearance sampling. The same fix in daylight costs a fraction.
With limited data on buried and aging assets, replacement runs on age and paving schedules. Good equipment gets pulled while a quietly failing station waits its turn.
PRISM learns what normal looks like for each station, and recognizes the drift away from it while there is still time to act. Every flag arrives as a decision your crews can run with.
| Station | Spent | Avoided |
|---|---|---|
| Transfer station, worn marine hose | $225 | $29,740 |
| Residential station, pump overload | $1,474 | $28,491 |
| Subdivision station, both pumps clogging | $2,981 | $26,984 |
| Residential station, greased-over floats | $28 | $3,992 |
| All nine interventions | $8,749 | $138,101 |
An eight-week validation on a lift-station network. The parts get bought either way. What advance notice recovers is the emergency premium, the cascade damage, the overflow and the overtime, which is most of the bill. See all nine, line by line
A regional extract from the reference deployment: every failure in the window, tested against the model that was running the whole time. Twenty-six of twenty-eight were flagged. Two produced no crossing at all. Those two carry full cost in the analysis and earn credit nowhere.
One thing to expect, and to say out loud. Acting on a warning means touching assets that had not yet failed, so total maintenance events go up while emergency events go down. Any pilot review that measures success by counting work orders will read as a failure. The measure that holds up is emergency and after-hours labour hours against the same interval a year earlier.
No new hardware, no field installation and no system replacement. PRISM reads your existing SCADA and historian.
Read-only access to SCADA, historian and work-order history. Imperfect, mixed data is expected, and we clean it on our side.
Baseline the normal operating signature of every pump, well, plant and lift station, including site-specific naming.
Rank by risk, name the probable failure mode, quantify the time to act, and triage before anyone is paged.
Into the CMMS, the dispatch queue or the on-call phone. Whatever your crews already open every morning.
Typical time from first call to a live model: four to eight weeks. The engineering hours are ours, not yours.
This is the view a maintenance planner opens. Not a health score to interpret, but every station laid out against the clock, so the work sorts itself.
Each bar is one asset's predicted days-to-failure window. Where it sits is when to act. How dark it is is how certain and how severe. Sorted soonest first, so a planner works top to bottom.
PRISM reads the SCADA and historian telemetry your utility already collects, under a read-only credential, and delivers the decision into the system your crews already open every morning.
One-directional ingest in, decisions out. Nothing in the product can change a setpoint, close a valve or start a machine.
See which stations are drifting before the call-out, and schedule the fix on your own terms and in daylight.
Arrive with the right part and a plan, rather than diagnosing under emergency conditions.
Prioritize condition assessment and rehab on the assets that actually carry risk, not the ones that are simply old.
Justify capital with evidence, and hold an auditable, time-stamped asset-health record you can put in front of a regulator.
The honest answer is that it comes out of your own telemetry, and we will not quote you a savings figure until it does. But the validated deployment gives a defensible per-asset anchor, and you can scale from it here.
The reference number is $1,616 of realized savings per monitored asset per year, reconciled against an operator's own cost records across 391 assets over nine months. Not modeled. Not projected. Reconciled.
It was earned at an intervention rate of 83.6%, meaning the share of flagged failures actually acted on. That one number moves the answer more than anything else. So it is a control you set, not an assumption we make for you.
A first-order projection scaled from the validated deployment. Your real number comes out of your own telemetry, not this slider.
Basis. Anchored to $1,616 of realized savings per monitored asset per year, measured across 391 assets over nine months at an 83.6% intervention rate, and scaled linearly from there. Event frequency from 28 failures across those assets in a 228-day window, annualized. System-class weighting is modeled, not measured. Excludes asset-life extension, capital deferral, energy and regulatory upside.
No. PRISM starts with what you already have: SCADA, historian records, pump attributes and work-order history. Mixed and imperfect data is expected. A station that drops off the network for a week and then sends a week of readings in two minutes is the normal case, not a blocker. The model improves as more of your operating history is read.
No. PRISM is software only. It reads the telemetry you already collect, so there is nothing to install in the field, nothing to maintain on a pole, and no capital cost to deploy. The reference deployment covers 391 assets with zero field hardware installed, and we were never on site.
SCADA alarms fire after a single threshold is crossed, which is to say once the failure is underway. PRISM models the joint signature across correlated channels and catches the drift before any threshold is reached. It then names the probable failure mode and the time you have to act. It also reduces alert volume rather than adding to it, because every flag is triaged before it reaches a person.
Because it is tuned against your cost of intervention, per asset class. For remote stations where a truck roll is expensive, we set the model for precision. It stays quiet unless it is confident. For plants where a crew is already on site and the consequence of a miss is a compliance event, we set it for sensitivity. And the output is a ranked daily list with reasoning attached, not a feed. That is a configuration decision we make with you at scoping, not a default you inherit.
No, and you should expect the opposite. Acting on a warning means touching assets that had not yet failed, so total maintenance events rise while emergency events fall. If your pilot review counts work orders, it will read as a failure. Measure emergency and after-hours labor hours against the same interval in the prior year. That is where the money moves.
A short scoping call, a model built on your telemetry in four to eight weeks, and a review of the predictions against your real failure history. Evaluation is at no cost, with no exclusivity and no obligation, and there is no invoice until the first report is delivered. Deployment requires essentially zero engineering hours from your team: name who receives the daily list, confirm the severity routing, and countersign.
We build a model on your telemetry, then review the predictions against your real failure history, event by event.
Deployment data. Results, event register, lead times, intervention rate and accuracy are from a Firstlook production water deployment covering 391 assets over nine months, together with a 28-event regional extract over a 228-day observation window, scored against the customer's closed work orders.
Cost basis. The cost stack is built from a loaded straight-time field labor rate with an after-hours multiplier, plus documented mobilization, cascade damage, notification, sampling and lost-revenue lines. Forward economics, including the calculator on this page, are modeled from that basis rather than measured. Full working available on request under NDA, and set out on the proof page.
All performance figures on this page are measured on Firstlook production water deployments and reconciled to the operator's own cost records. The 391-asset nine-month deployment and the eight-week lift-station validation are separate engagements and their figures are not combined. Emergency-versus-planned repair multiples and system-class weightings are modeled estimates. Full methodology and the underlying cost model are on the proof page, and available in full on request under NDA.