Soda Cloud dashboards
Open Soda Cloud and see what matters to you.
A dashboard view summarizes the data quality results Soda has collected across your estate. With the Soda Cloud dashboards, you can see the overall health of the data quality that matters to you, grouping metrics by your own attributes, building your own customized views, and accessing trends over time.
Soda Cloud ships with three views, each built around a different question:
Is quality improving?
Is the data quality program working?
What needs fixing today?
Pick the view closest to your job and start there. Every dashboard is customizable: every tile can be moved, resized, duplicated, or removed, and opening a tile lets you rewrite its title and description and change the measure, the grouping, and the visualization behind it.
Executive overview: the CDO cockpit
The executive view answers three questions a CDO gets asked in board and steering meetings:
Is quality getting better?
Is the program reaching more of the estate?
How much of the estate has an agreement behind it?
Use it when you need a number to report upwards, or when you want to see whether last quarter's investment moved anything. Filter the board to one domain (for example, a specific data source or dataset owner) and it becomes the same report for a single business unit.
Manager overview: the state of the data quality program
The manager view sits between the executive summary and the daily detail. It keeps the health and coverage headlines, then adds the operational picture underneath:
What is running?
What is failing?
What has not been brought into Soda yet?
DQ Health Score and Coverage carry over from the executive view. The rest shows the state of your data quality program and where to push it next: check outcomes and anomalies over time, what's onboarded and what isn't, as well as health per dataset.
This is the view for whoever owns the data quality program: a data quality lead, a governance lead, or an engineering manager whose team is responsible for the pipelines. It answers where to put the next unit of effort, which is usually either a data source with a lot of failing checks, or a pile of discovered datasets nobody has onboarded.
Steward overview: what to fix today
The steward view drops the estate-wide framing and shows the working queue:
Which datasets have failing checks?
How many rows passed and failed in each?
Where are anomalies happening?
Use the steward overview as the start of a shift. The datasets at the top of those bars are the ones worth opening first, and clicking through to the check takes you to the individual results.
Custom overview: a dashboard just for you
You can customize any of the default dashboard overviews or build one from scratch to see what you care about at a glance.
Learn more about how to customize your dashboard.
Definitions of tile metrics
Every tile that reports a score uses one of the definitions below. Hover the icon on a tile to see the same definition in the product.
DQ Health Score
The percentage of checks that passed, out of all checks evaluated in the tile's time range.
A tile's time range is part of the tile, so two tiles showing a health score over different ranges will show different numbers for the same datasets. The default range is the last 90 days.
The health score on a dashboard tile and the one on a dataset's Checks tab are not the same number.
The dashboard tile averages the score across the tile's time range, and leaves out checks that have never been evaluated.
The Checks tab shows the score for the current state of the contract, and counts a never-evaluated check as not passing.
A dataset with two unevaluated checks can read 84% on a dashboard and 73% on its Checks tab at the same moment. Both are correct for what they measure.
Coverage
The share of datasets that have data quality in place. A dataset counts as covered once it has at least one check, metric monitor, published contract, or data standard.
Has Contracts
Datasets count as "With contract" once their contract is published. A draft doesn't count as a dataset with contract.
Has Monitors
Monitored datasets have at least one metric monitor configured. The rest have none, so Soda isn't tracking their metrics for anomalies.
Onboarding Status
Onboarded datasets have been added to Soda, so Soda can scan and check them.
Discovered datasets were found in the data source but aren't onboarded yet, so nothing runs on them.
Dashboards outside Soda Cloud
If your organization already reports from Power BI, Tableau or Looker, you can pull Soda's results into those tools instead of building the view in Soda Cloud.
Two routes, depending on what you need:
The Soda REST API returns aggregate results: check outcomes, dataset metadata, and incidents. Use it for the same numbers the dashboard tiles show, expressed in your own reporting model.
The Diagnostics Warehouse holds the rows behind a failing check. It runs in your own environment, so point a BI tool at it directly when you need record-level detail. Nothing at that level is available through the API, because failed rows are never stored in Soda Cloud.
Limitations & edge cases
Dashboards are updated hourly.
Dashboards report on your data, not on your users. Active datasets and active checks are available as metrics, but user activity is not.
Public beta
During public beta, we are working to improve functionalities. For the time being, dashboard views have the following limitations to keep in mind:
Dashboard views are stored locally in your browser, not in Soda Cloud. A view you build is only available in the browser you built it in, and clearing your browser data removes it.
Dashboard views cannot be shared with other users.
Dashboards can show historical data up to 90 days. However, during public beta, the timeline will be limited to the date when dashboards were implemented in your organization. This means that, if the feature was introduced 25 days ago, dashboards will show data going back up to 25 days.
Soon, dashboard views will be shareable and persist in your Soda Cloud account.
Record-level detail
Dashboard tiles aggregate results. They do not show the rows behind a failing check, because failed rows are written to your Diagnostics Warehouse, which runs inside your own environment and is never stored in Soda Cloud.
To analyze failed rows, query the Diagnostics Warehouse directly or point a BI tool at it.
Last updated
Was this helpful?