Insights
A chart shows a line. A manager wants an answer: what did that cost, how long did it run, did it stay within the limit, is it getting worse, and what should we expect next month. Insights answers those from the data the console already stores — no export, no spreadsheet, no second system — and says how much of the window it had data for, every time.
What a meter consumed per hour, day, week or month, and what that cost: every hour's consumption is priced by the tariff window it fell in — peak, off-peak, weekend — in the tariff's own time zone. Hours where the meter went backwards (a reset, a replacement) are left out of the sums and counted separately, never credited as consumption.
How long a machine ran and how often it started, from nothing more than a threshold on a signal it already reports — current, power, speed, pressure. Runtime, starts, the longest run and the duty cycle per period. Time counts only where data exists, and a start is counted only when two consecutive minutes disagree, so a sensor going quiet neither adds hours nor invents a start.
Proof that a reading stayed inside its limits: how much of the window had data, how much of that was within bounds, and every excursion with its start, end, duration and worst value. Judged on each minute's extremes rather than its average — a reading that touched the limit for twenty seconds left the band, and the report says so.
Every sensor under an asset compared with its own recent past: the last days' mean and peaks against the weeks before, measured in units of how much that sensor normally varies from day to day. The pump whose vibration is creeping up while still far below any alarm is ranked first — the failure that a threshold cannot see.
For one sensor, every sibling under the same asset scored by how closely it moves with it, at the lag where they agree best — so "the temperature follows the current by four minutes" is a result, not a hunch. The sample count is shown beside every score, because a perfect correlation over a dozen minutes is a coincidence.
A trend fitted through the daily means, optionally corrected by the weekday pattern, projected ahead — and the day it crosses a limit, if it does. Slope and fit quality are shown so the projection reads as "rising 0.4 a day, Mondays run high", which is the form a capacity decision is made in.
The same quantity across many assets, divided by what makes them comparable — floor area, units produced, beds — and ranked against the fleet's median. Assets without a normaliser are listed with their raw value but take no part in the ranking, because a comparison that mixes "per square metre" with "per building" compares nothing.
What a tracker did: distance covered, time moving and stopped, the trips between stops, and how long it spent inside each named site. Pair a meter on the same asset — fuel, battery — and the result includes consumption per kilometre.
A line chart can draw the asset's own log lines — errors, warnings, whatever you name — as ticks along its timeline. Hover a point and the readout lists what the machine said in that minute, so "the pressure dipped at 14:07" and "pump restarted at 14:07" finally meet on one axis.
Three kinds of definition live beside the questions: a tariff (currency, time zone, the default price and any number of priced windows by weekday and hour), a set of sites (named geofences), and asset attributes (what each asset is divided by to compare it). Saved once per account, reused by every panel that needs them, and sent along with each question — so the answers never depend on anything the query service has to remember.
Every answer is computed from the same per-minute aggregates and per-day summaries the charts read — nothing new is collected, nothing is estimated. Every result carries its data coverage, resets are counted and never credited, a forecast refuses to fit on fewer than seven days, and a correlation shows its sample size. A figure that would have to guess says so instead.
Insights sits in the console's Explore group, beside Log Explorer and Reports. See the Explore group
