Skip to main content
The Data tab of the Observability Dashboard gives an organization-wide view of data reliability: how often assets are fresh, how often volume behaves as expected, how often schemas stay stable, and which assets are currently unhealthy. Navigate to Analytics → Observability (Data) → Data to access it.
Observability is a separate signal from Quality: Quality scores correctness (DQ measures), Observability tracks reliability (did data arrive on time, in the expected volume, with the expected schema). An asset can be observability-healthy and still fail quality checks, or vice versa.

Layout

The tab is laid out top to bottom as: the three uptime tiles, the AI summary, the four count tiles, the Alert and Issue heatmaps side by side, the Recent schema changes table, and the Unhealthy assets table. Every tile and widget respects the filter bar and the date range.

KPI cards

Uptime tiles

Freshness Uptime, Volume Uptime, and Schema Uptime tiles each with a percentage and a comparison badge such as -2.12% vs 99.89% prior 90d

The three uptime tiles. Each shows the uptime for the date range and the change against the prior window of the same length.

Under each percentage a badge compares it with the same uptime measured over the window immediately before the date range, for example -2.12% vs 99.89% prior 90d. The label follows your range: with Last 90 days it reads prior 90d, and with a 14-day range it reads prior 14d. The badge shows the change in percentage points and the prior window’s own value.

AI Summary

AI summary card describing the last 90 days of freshness, volume, and schema reliability

AI summary for the Observability (Data) tab.

An AI-generated summary states the headline reliability figures for the current date range, compares them with the prior period (for example freshness alerts up from 1 to 41, uptime down from 99.89% to 97.77%), and names the widget to open first. In validation, this summary did not always refresh immediately when switching between the Catalog, Data, Pipeline, and Report tabs — it can briefly keep showing the previous tab’s text. If the summary’s wording doesn’t match the tab you’re looking at, give it a moment or treat the tiles and widgets below as authoritative.

Count tiles

Stale Assets, Volume Anomalies, Schema Changes, and Freshness Alerts tiles with comparison badges such as +217 vs 1 prior 90d

The four count tiles, each with a comparison against the prior window.

Each count tile carries a badge such as +217 vs 1 prior 90d: the change from the prior window of the same length, followed by the prior window’s own count. A downward arrow means fewer than before. For these counts a lower number is better, so a rise in anomalies, schema changes or freshness alerts is the signal to investigate.
All seven tiles animate with a count-up effect and can take several seconds to settle after the page loads or after any filter/date/Slice By change. Read a tile only after two consecutive looks agree — in validation, a tile read mid-animation showed values several multiples off from the settled number.

Widgets

Alert heatmap

Alert heatmap with Volume, Freshness, and Schema tabs, domains as rows, and a legend running from 0-20 green to 80+ red

Alert heatmap, Volume tab. Rows are domains and columns are Slice By buckets.

A heatmap of alert counts with Domain as rows and time buckets (controlled by Slice By) as columns, with sub-tabs for Volume, Freshness, and Schema. The card subtitle reads Domain by selected time slice; higher alert count is worse. The legend runs 0-20 green → 80+ red, because for alerts higher is worse — this is the inverse of the Quality Dashboard’s score heatmaps, where green means a high (good) score. Assets that are not assigned to a domain are grouped under Unmapped, and a pale cell means no alerts in that bucket. Use it to find the domains and weeks with the noisiest or least stable data.

Issue heatmap

Issue heatmap with Volume, Freshness, and Schema tabs and a legend running from 0-20 green to 80+ red

Issue heatmap, Volume tab. Same layout as the Alert heatmap, counting issues.

The same structure as the Alert heatmap — Domain rows, Slice-By time-bucket columns, Volume/Freshness/Schema sub-tabs, 0-20 green → 80+ red — but for issue counts instead of alert counts. Only domains and buckets that have issues are drawn, so this heatmap usually has fewer rows and columns than the Alert heatmap.

Recent schema changes

Recent schema changes table with columns Asset, Change Type, Changed Field, Impact, and Changed

Recent schema changes. The number in the title is the count of change events in scope.

A table of individual schema-change events at the column level (finer-grained than the Catalog Dashboard’s asset-level Schema Changes widget). Columns: Asset, Change Type (e.g. Column Added), Changed Field, Impact (e.g. Additive), Changed (timestamp). Click an asset name to open it, and use the magnifier icon to search. Review non-additive changes first, because they are the ones most likely to break downstream consumers.

Unhealthy assets

Unhealthy assets table with columns Asset, Data Updated, Freshness, Volume, Schema, Impacted Assets, Reports, and Dashboards

Unhealthy assets. The number in the title is the count of assets with at least one reliability problem.

A table of assets with at least one reliability problem. Columns: Asset, Data Updated (last update timestamp, — when the asset has no recorded update), Freshness (for example Stale or On time), Volume (for example Normal), Schema (for example Stable or Drifted), Impacted Assets, Reports, Dashboards (downstream BI impact counts via lineage). Use the three impact columns to prioritise: an unhealthy asset with downstream reports and dashboards matters more than one with none.

Filters

Filter bar on the Observability (Data) dashboard

Filter bar on the Observability (Data) dashboard

Same filter bar as the other Analytics dashboards: Domain, Application, Product, Tag, Source, Asset Type, Segment, Asset, Date Range, Slice By, Previous Period. See Catalog Dashboard for the full list of options; the mechanism is identical here. Previous Period sets the window behind every badge on this tab. With Match Date Range (default) the uptime and count tiles compare the whole range with the same-length window before it, which is why the badges read prior 90d or prior 14d. With Match Slice By the tiles narrow to the latest Slice By bucket and compare it with the bucket before it, and the label changes to match (for example prior week). See Previous Period and the KPI badges.
Filter selections persist across page reloads and across navigating away and back, for as long as your session is active. Click Reset to confirm you’re looking at the true unfiltered baseline before drawing conclusions from the numbers.

Slice By: time-bucket granularity for the heatmaps

Slice By offers the same six options as elsewhere (Day, Week, Month, Quarter, Half-year, Year; defaults to Week). On this dashboard it rebuckets the column headers of the Alert heatmap and Issue heatmap — it does not change any KPI tile. Worked example, no other filters applied: Only the heatmaps’ column buckets changed — every KPI tile stayed identical, the same pattern documented for the Quality Dashboard’s trend chart and heatmaps.

With filter vs. without filter: a worked example

Example captured during validation, Last 90 days, no Slice By change: Narrowing to a single asset type reduces every count tile (fewer matching assets/events) and, in this snapshot, also reduced the uptime percentages slightly — tables in this environment had a somewhat worse reliability record than the full asset population over the period. The same mechanism applies to every other filter field and to combinations of them.
The exact figures above are an illustrative example and will not match your environment. The relationship they demonstrate — filtering narrows the population, Slice By only rebuckets time — is what to rely on, not the specific numbers.

Catalog Dashboard

The Catalog tab of Analytics — asset discovery, classification coverage, and governance health

Quality Dashboard

DQ score, alerts, and issues — the correctness counterpart to this reliability dashboard