Fifteen rules
0:07
Fifteen rules. Severity and status, kept apart.
0:12
AlertsAlerts lists every rule watching the business and its data at Acme Home Appliances, a demo company, how severe each is, and how often it actually goes off.
0:24
Firing nowThree rules are firing now, and two of them are critical.
0:30
ArmedFourteen are armed. One is still a draft.
0:35
NoiseOne fires often without being severe, which usually means nobody reads it any more.
0:42
Never firedAnd one is armed and has never fired, so nobody knows whether it works.
0:49
Two columnsStatus says whether a rule is firing. Severity says how bad it would be. They are drawn from different colour families, so nobody reads high severity as on fire.
1:01
Four severitiesCritical means somebody is woken up. High is looked at the same day, medium this week, and low is logged for the record.
1:11
Six kindsWatches for names the kind of rule: a threshold, an anomaly, a ranking, freshness, schema or volume.
1:19
How oftenFired counts the last thirty days. A high number on a low-severity rule is noise somebody should turn off.
Firing now
1:28
Firing now. One business rule, two data rules.
1:33
A business ruleMonth-to-date revenue pacing fires when any region is more than ten percent below the plan due by now. The West is.
1:43
In words firstThe panel leads with the sentence: firing now and unresolved, fired nine times in thirty days.
1:50
This morningIt last fired under an hour ago.
1:55
What it watchesIt watches the West region and notifies by chat and email. Its owner is the VP of strategy.
2:05
A data ruleChannel stock freshness fires when the stock snapshot has not landed inside its 24-hour window. It pages the on-call engineer.
2:14
FreshnessFreshness means data did not arrive when it should. It is watching the staging table, stg_channel_stock.
2:23
Schema driftSchema drift fires on any new, missing or type-changed column. Warranty claim amounts started arriving as text on 3 August, and nothing downstream failed.
2:37
Active, and busyRequired column null rate is not firing now, but it fired seven times in thirty days at high severity. That is not noise: required columns keep arriving more than one percent empty.
The states that mislead
2:51
The states that mislead. Never fired, a draft, snoozed and noisy.
2:56
Critical, never firedPrimary key uniqueness is critical and has never fired. That is either a rule doing its job silently, or one that has never been proven to work.
3:08
NeverLast fired reads never, not a long time ago. They are different facts, and they are not stored the same way.
3:19
A draftMarketing spend pacing was written and never armed. It watches nothing, and the person who wrote it likely believes otherwise.
3:30
SnoozedWarranty rejection spike is snoozed. It still evaluates; nobody is told when it triggers.
3:38
NoisyChannel stock aged past ninety days fired six times in thirty days at medium severity. A rule that fires this often has usually stopped being read.
3:50
Retire itSo the panel offers to retire it.
3:57
Low, on purposeScheme payout above slab is low severity and fires by design: any payout over ₹2 lakh to one distributor goes to Finance for review. Three in thirty days.
4:11
A blind spotChannel stock row-count drop watches the same data for volume. It last fired ten days ago and has been quiet since: a feed that sends nothing has no rows to compare. That is the freshness rule's job.
Finding rules
4:27
Finding rules. By owner, by target, by kind.
4:34
By ownerThe data platform team owns five rules, all of them about the data itself: freshness, schema, volume and keys.
4:45
By targetThree rules watch all 470 distributors: a quiet distributor, a drop out of the top twenty, and ageing stock.
4:58
Most oftenSorted by how often they fire, the two nine-time rules lead: revenue pacing and channel stock freshness.
5:06
ThenThen required column nulls at seven, and ageing stock at six.
Acknowledge, not fix
5:14
Acknowledge, not fix. Stopping the noise is not the same as solving it.
5:19
AcknowledgeThe West regional head has seen the pacing alert and has a plan. Acknowledge it.
5:28
Seen, not solvedAcknowledged means seen, not fixed. The rule stays armed; the West is still behind; only the noise stops.
5:37
Two firingTwo rules are firing now.
5:43
ActiveRevenue pacing is back to active: armed, watching, not currently shouting.
5:53
On the recordThe audit trail records who acknowledged it, with the reminder that the rule is still armed.
A rule that clears itself
6:01
A rule that clears itself. Fix the data, and the freshness alert stands down.
6:08
The causeThe freshness alert is not something to acknowledge. Its cause is an expired key on the channel stock feed, in Antvia Data. Repair it.
6:21
Six stagesThe run lands fresh stock and runs the quality gates.
6:32
One firingBack in Alerts, one rule is firing: schema drift, which needs somebody to look at the warranty portal.
6:41
Stood downThe freshness rule is active again. Nobody acknowledged or edited it. The table it watches is current, so it has nothing to say.
6:51
Two kinds of fixBusiness alerts are acknowledged by people. Data alerts clear when the data is fixed. Keeping the two apart is what lets both be trusted.