Insights Examples

Reports people actually ask for, and the settings that produce them. Each scenario starts from a goal and ends with the configuration that answers it.

A monthly quality report for leadership

Context. A data steward presents the state of the estate at the end of every month and needs a figure that holds still after the meeting.

What happens. Clear the filters so the report covers everything in scope, set the report date to the last day of the month, and set the timeframe to Month. The Score Summary gives the headline number and the percentage beside it gives the month's movement. Score Progression, grouped by week, shows whether the month ended on a rise or a fall. Export the report to close it out.

Why it works. The PDF carries the report date, the timeframe, the generation time, and the datastores included, so the number in the deck can always be traced back to the exact context that produced it. Reopening the page later will show different values; the export will not.

One business domain's trend over the quarter

Context. A domain owner is accountable for the finance data and wants to know whether its quality is improving, without the rest of the estate diluting the picture.

What happens. Apply the domain's tag in the Explore filter bar, set the timeframe to Quarter, and leave Score Progression grouped by month. Read the overall line first, then use Compare Quality Dimension Trends to plot Accuracy and Conformity against each other and find which one drove the movement.

Why it works. The tag filter matches both datastore-level and container-level tags, so a domain spread across several datastores reports as one unit. Three months grouped by month is enough resolution to separate a trend from a bad week. Selecting dimensions replaces the total line rather than overlaying it, which is why the order matters. Read the total, then decompose it.

Whether last night's scan found anything

Context. In a daily triage routine, the scans run overnight and someone checks in the morning whether they surfaced anything new.

What happens. Set the report date to today and the timeframe to Week. On Scanning Activity, switch the bars to Anomalies Identified and read the last bar. If it is higher than the days before it, go to the Anomaly Summary card and open the Active list.

Why it works. Anomalies are counted on the date of the scan that produced them, so the last bar lines up with last night's run rather than with when someone got around to reviewing it. The drill-through opens the Anomalies tab filtered to Active, which is the queue to work from, though it lists every active anomaly rather than only the ones behind that bar.

Which rules are failing most often

Context. A team has more failing checks than it can work through and needs to decide where remediation effort goes first.

What happens. Read Failed Check Distribution and raise it to the top 10 rule types. Hover the widest segments to see each one's share of the anomalies, its count, and its importance. Compare against Check Distribution just above it. A rule type that is a small share of your checks but a large share of your anomalies is the one to look at first.

Why it works. Reading the two distributions together separates "we have many of these checks, so they fail often" from "these checks fail at a disproportionate rate". Only the second is a signal about the data.

How much data is under management and how fast it grows

Context. A platform owner reports on scale each quarter, for capacity planning and for licensing conversations.

What happens. Set the timeframe to Year and read Data Under Management for the current scale. For the shape of the growth rather than the endpoints, read the Data Volume chart grouped by month.

Why it works. Data Under Management counts every container's most recent record count, whichever operation produced it, so it reflects the whole estate. The Data Volume chart covers only containers under volumetric tracking, which makes it the better shape but the worse total. Using each for what it is good at avoids reconciling two numbers that were never meant to match.

Whether coverage is keeping up with onboarding

Context. A team has been connecting datastores quickly and wants to know whether quality coverage is growing with them or falling behind.

What happens. In Data Overview, compare Containers against Containers Scanned in the Anomalies & Scanning section. Then read the Not Asserted count on the Check Summary card, and the Coverage dimension on the Score Summary.

Why it works. The three numbers describe the same gap at different stages. Containers that were never scanned contribute nothing to the Quality Score, checks that were never asserted are intent without evidence, and Coverage is the dimension that prices both. Onboarding that outpaces scanning shows up here before it shows up as a bad score, because unmeasured data does not lower the score, it just leaves it describing less of your estate.