Explore Predictive Maintenance

Predictive maintenance platform that tells you which asset to fix first

Explore reads the sensors and BMS points you already have, forecasts equipment behavior before it fails, and turns the result into one ranked queue, not another alarm stream.

  • EU-hosted, GDPR-compliant
  • Grounded inference, nothing invented
  • No new hardware required
The Insights Hub in Explore: open items counted as new, in progress, handled and dismissed, a severity breakdown, and insights grouped by equipment with their severity, occurrences, latest date and linked work order.

How it works

From sensor data to a ranked fix list.

01 In

Sensor and BMS data

Points you already have, over BACnet, Modbus, OPC UA and oBIX. No new sensors, no new gateway hardware.

02 Analyse

Detection + forecasting

Seven anomaly-detection methods run continuously. Forecasting projects each point forward with a confidence bound.

03 Out

Ranked decision queue

Findings merge into one queue, ranked by cost and risk.

04 Fix

Work order

A finding that needs a technician becomes a work order, and the fix is checked against the data.

Why most predictive-maintenance projects stall

The model was never the hard part. Getting someone to act on its output is.

Most predictive-maintenance pilots don't fail because the anomaly detection is wrong. They fail because the output is another alert stream, and the team already ignores three of those. A vibration spike, a temperature drift, a runtime anomaly: each lands in its own dashboard, competing for attention with everything else that's already flashing red.

Explore starts from the actionability problem, not the detection problem. The same detection engine that powers BMS analytics runs seven anomaly-detection methods against every point continuously, then forecasts where each one is heading with a confidence bound, so a developing bearing fault or a refrigerant leak shows up days or weeks before it trips a threshold, not after.

The output isn't a feed. It's a ranked decision queue: one list, ordered by cost and risk, that tells a technician what to look at first, instead of a wall of equally-urgent alarms nobody has time to triage.

This page is the product view. The methodology behind it is in Predictive maintenance for buildings: a practical guide. How Explore stacks up against Clockworks, Facilio and CopperTree Analytics on this specific job is in the vendor comparison. For where predictive maintenance ends and condition-based maintenance begins, see the CBM guide.

What Explore actually does

Four capabilities, one ranked queue.

Detect

Seven-method anomaly detection

Every BMS point and connected sensor is checked continuously against seven detection methods, not one fixed threshold. Covers drift, stuck values, runtime anomalies, and pattern breaks a static alarm limit misses. Full mechanics are on the anomaly detection glossary entry.

The Anomalies view in Explore: open anomalies counted by severity and by kind, including statistical outlier, sensor calibration, energy spike, equipment health and forecast deviation.

Forecast

Forecasting with confidence bounds

Detection tells you something is already wrong. Forecasting tells you it's about to be. Explore projects equipment behavior forward with a confidence interval, so slow bearing wear or coil fouling surfaces while there's still time to schedule a fix instead of an emergency callout.

A chart in Explore showing actual sensor readings for the past week and the forecast for the next seven days with its likely range, with anomaly markers on the timeline.

Investigate

Root cause and evidence, written for you.

  • The bottom line and the likely cause, in plain language
  • Measured value against the usual level, and how long it has lasted
  • Evidence, occurrences and a full report one tab away
An AI investigation in Explore with tabs for what to do, evidence, occurrences and full report. It states the bottom line and the likely cause, then shows the measured value, the usual level, the difference and the duration of the fault.

Fix

From finding to finished job.

  • A finding that needs a technician becomes a work order
  • The visit and its outcome show on the insight itself
  • Explore checks the fix against the data afterwards

Explore Field Service Management →

CriticalHandledFault: 81% confident

VAV · 00-002 - Waiting Lounge · Demo Office London

Faulty temperature sensor on 00-002 VAV unit

First seen 28 Jul. Fixed 12 Nov. Verified 26 Nov.

FM partner work order
WO-1 · Demo FM ABVerified
  1. Work order created11 Nov, 22:01 London · covers 16 insights
  2. Visited · fix applied12 Nov
  3. Verified by dataDecided 26 Nov · Window 12 Nov to 26 Nov
Before the fix-8.2 °C
After the fix16.4 °C
Usual level16.35 °C

Explore measured the fix. The reading is back at its usual level and the fault has not returned.

Shared by Demo FM AB. Technician names are not shared.

Illustrative.

One ranked decision queueFindings are ranked by cost and risk and merged into a single queue, not scattered across a BMS alarm panel, a spreadsheet, and three inboxes. A technician opens one list and knows what to look at first.
Write access starts off, action on requestFindings can feed the Edge Agent to propose a maintenance action or schedule change. A person reviews it, or it executes automatically inside a scope you've explicitly granted. Nothing runs on your BMS without that scope.Automation →

How the three approaches compare

Raw alarms and a paper calendar vs. a ranked queue.

Predictive maintenance platform: raw BMS alarms, manual PM schedules, and Explore compared
Compared onRaw BMS alarmsManual PM scheduleExplore
TriggerFixed threshold breach onlyFixed calendar interval, regardless of conditionCondition and forecast, seven detection methods
Lead timeAfter the limit is already crossedArbitrary, unrelated to actual wearDays to weeks ahead, with a confidence bound
CoverageOnly points with a hardcoded limit setOnly equipment on the PM calendarEvery point already in your historian
PrioritisationNone, every alarm is equally urgentNone, fixed schedule regardless of riskRanked by cost and risk, one queue
False positivesHigh, floods the operatorN/A, but wastes visits on healthy equipmentFiltered by confidence before it reaches the queue
New hardwareUsually none, but limited to already-sensored pointsNoneNone, reads existing BACnet, Modbus, OPC UA, oBIX points
Path to actionManual triage by an operatorTechnician dispatched whether needed or notWrite access starts off; Edge Agent can propose or execute inside a granted scope

At a glance

The mechanics.

A predictive maintenance platform in Explore isn't a separate module bolted onto a BMS dashboard. It's the same sensor-intelligence layer used for energy and BMS analytics, applied to HVAC and plant health.

Questions

Predictive maintenance platform, answered.

A predictive maintenance platform for buildings reads BMS and sensor data, forecasts with a confidence bound when HVAC or plant is likely to fail, and ranks findings into one decision queue, so a fix can be scheduled before a breakdown. Explore's AI predictive maintenance does that from the points your BMS already has.

BMS alarms fire after a fixed threshold is crossed: a temperature limit, a runtime cap. Explore forecasts where a point is heading with a confidence bound, so a developing fault can surface days or weeks before it would ever trip that threshold, and every finding lands in one ranked queue instead of a raw alarm panel.

A calendar-based PM schedule services equipment on a fixed interval regardless of actual condition, sometimes too early, sometimes too late. Explore ranks findings by real condition and forecasted risk, so maintenance effort goes where it's actually needed first. When a finding needs a technician, the work order and its history live in Explore Field Service Management, or in the CMMS you already run.

Condition-based maintenance triggers a maintenance action when a measured condition crosses a defined limit. Predictive maintenance goes a step further and forecasts when that limit will be crossed, with a confidence bound, before it happens. Explore does both: continuous condition monitoring plus forward-looking forecasts, merged into one ranked queue.

Usually not. Explore reads the points your BMS and existing sensors already expose over BACnet, Modbus, OPC UA, and oBIX. New hardware is only needed where a piece of equipment genuinely isn't instrumented yet.

Write access starts off. Explore detects and forecasts, nothing more, until you say otherwise. Findings can feed the Edge Agent to propose a maintenance action or schedule change, which a person reviews, or which executes automatically inside a scope you've explicitly granted. It can't act outside that scope.

See what Explore finds in your own data.

Connect your BMS, point Explore at it, and get a ranked list of what actually needs attention, not another dashboard to check.