Insights Best Practices

Habits that keep Insights trustworthy as a reporting surface, so the numbers you share mean the same thing next month.

Fix the report date first, then vary the timeframe

The report date decides every total on the page; the timeframe only decides what those totals are compared against. Setting the date first and then moving through Week, Month, and Quarter gives you three readings of the same moment, which is how you tell a one-week blip from a quarter-long slide. Changing both at once leaves you comparing two different snapshots and reading the difference as a trend.

Report the trend, not the total

Totals depend on how long a deployment has been running, so "4.2 billion records scanned" says more about tenure than about health. The percentage beside the total is the part that carries information, because it is measured against a defined earlier point. When you paste a number into a status update, paste the movement and name the timeframe with it.

Give the Quality Score something to average

The score covers only containers that have been scanned and scored, weighted by importance. On a deployment that has profiled widely but scanned narrowly, the score describes a small corner of your estate with full confidence, which is worse than describing all of it roughly. Before treating the score as a program metric, check that Containers Scanned is close to Containers, and scan the gap.

Set weights before the score becomes a target

Because the average is weighted, the score answers "how healthy is the data we care about" only if the weights say what you care about. Reviewing weights on your top containers once, early, is cheaper than explaining later why a staging table moved the number as much as a revenue table.

Do not reconcile Data Volume against Data Under Management

They are built from different populations on purpose, so making them agree is not a goal. Use Data Under Management for scale and licensing questions, and the Data Volume chart for change over time on the containers you monitor continuously. If the gap between them bothers you, the fix is enabling volumetric tracking on more containers, not reinterpreting the numbers.

Treat Not Asserted as a coverage gap

A check that has never been asserted is intent without evidence. Watching that count fall after each scan is a cleaner progress signal than watching Active Checks rise, because authoring more checks moves the second number without testing anything.

Export when a number needs to survive the conversation

The page always reflects a live snapshot, and refreshing it changes what a colleague sees when they open the same link. The PDF export carries the report date, the timeframe, the generation time, and the datastores in scope, which makes it the right artifact for an audit trail, a monthly review, or anything you will be asked about again in three months.

Refresh before you doubt the data

If an operation finished minutes ago and its results are missing, the page is probably serving a saved snapshot rather than reporting a problem. Refresh recalculates immediately from what has been recorded. It does not start a Profile or a Scan, so it is always safe and never the cause of new load on a source.