Alarm ranking chart

Who it’s for

Whoever is asked why there are so many alarms. A maintenance lead deciding where next month’s effort goes, an integrator tuning rule chains that fire more often than anyone reads them, an operations manager who has been told the team is drowning in alerts and needs to say which ones. The question is never “what alarms do we have” — the alarm table answers that — it is “which few of them are most of the problem, and is fixing those enough”.

What it does

Eighty one alarms in a week across six types. High temperature fired forty times and Low battery twenty five; the other four together fired sixteen. The chart ranks them, draws the share accounted for as the ranking descends, and states the finding above it: 2 of 6 types cause 80% of the alarms. Two rule changes, not six.

• Ranks alarm types, devices or severities, so the same chart answers “which rules are noisy” and “which equipment is unwell”.
• Counts occurrences or total time spent active. Three alarms that stood open for a fortnight outrank two hundred that cleared themselves in a minute, and only one of those two lists is worth a maintenance visit.
• Puts the conclusion in words above the chart, as the sentence someone repeats in a meeting, with the chart as the evidence for it.
• Collects everything past the ranking cut into one bar rather than dropping it, so the cumulative line still reaches 100% and the shares stay honest.
• Withdraws the cumulative line and its target when fewer than three causes exist, because one cause always accounts for everything and saying so with a curve invites a conclusion the data cannot support.
• Ranks what is there. It does not tell you why a cause is top, or compare this period against the last.

How to set up

Data keys

Alarm widgets take no data keys. The alarms come from the widget’s alarm source — an entity, an alias or the whole tenant — and are narrowed by the alarm filter in the widget’s own settings, exactly as for any alarm table. The dashboard time window decides the period.

What it reads from each alarm:

• type — the alarm type, for ranking by type.
• severity — for ranking by severity, and for the bar color.
• originatorName — the device or asset, for ranking by device.
• startTs and clearTs — when it began and cleared, for total time spent active. An alarm still standing counts up to now rather than being skipped, which is what keeps a long-running fault visible in the duration ranking.

Ranking

• What the alarms are ranked by — alarm type by default. Type finds the noisy rules, device finds the unwell equipment, severity shows the mix.
• What is being counted — number of alarms by default. Occurrences find what fires most often; total time spent active finds what actually hurt.
• How many to rank — 8 by default, between 3 and 40. Everything past the cut collapses into one bar rather than disappearing.
• Collect the rest into one bar — on by default. Off charts only the leaders, and the cumulative line then stops short of 100%.
• How many alarms to pull — 1000 by default, between 50 and 20000. The ranking is computed from the alarms actually fetched, so a limit below the number in the period skews the shares towards whatever arrived first.
• Which way the bars run — horizontal by default, which gives each name a full line, as alarm types and device names usually need. Vertical is the classic shape for short names.

How to customize

• To change where the target sits — set Share to aim at (%). 80 is the convention and rarely worth changing; it also drives the sentence above the chart. Mark a share to aim at turns the dashed line off without changing the sentence, and Target line colors it.
• To drop the cumulative line — turn off Draw the cumulative share line. The line is what turns a ranking into an argument, so it is worth a reason. Cumulative line sets its color.
• To match a dashboard’s palette — set Bars. When ranking by severity, Color the bars by severity uses the platform’s own severity colors instead, which say more than one flat color.
• To hide the numbers on the bars — turn off Label each bar with its value. Value label font and Value label set how they look. Values that would overlap are dropped rather than printed over each other.
• To restyle the axes — set Axis font and Axis, which also style the no-data message.
• To drop the sentence — turn off State the conclusion above the chart. Conclusion font and Conclusion style it.

Tips

• Read the chart twice, once by occurrences and once by time spent active. The two rankings usually disagree, and the disagreement is the interesting part: what fires constantly and clears itself is a tuning problem, what fires rarely and stays open is a maintenance one.
• Rank by device after ranking by type. A type that dominates because one device produces all of it is a device problem wearing a rule’s name.
• Raise How many alarms to pull before trusting the shares on a busy site. A thousand alarms in a week is ordinary for a few hundred devices, and a cap below the real count quietly ranks only the most recent ones.
• Narrow the alarm filter to the severities you actually act on before concluding anything. A ranking dominated by warnings nobody reads says more about the filter than about the plant.
• Pair it with the alarm table on the same dashboard and wire a bar click to filter that table, so the chart chooses the target and the table shows the instances.