PlatformModel
Check the values before they reach a report
Data Checks adds rules to the values a query returns. A row that breaks them is flagged or excluded, and each run gives a quality score.
- amount
- between 0 and 100,000 · excludes
- customer
- not null · flags
- status
- open, closed, cancelled · flags
- Quality
- 98.4%
- In quarantine
- 12 rows
Excluded rows are kept for 90 days.
Illustrative example
Simple rules, on the data that matters
Four kinds of rule
A range, a non-null value, allowed values, or a single expression.
Flag or exclude
Each rule flags the row or excludes it. Excluded rows go to a quarantine table, kept for 90 days.
A score on every run
Each run gives a quality score; the people who own the rules are told when it changes.
Bronze, Silver, Gold
A query is labelled by how it is built, and Athena can propose checks for you to accept.
How it works
- 01
Pick the query
Start from a query in the Lab, for example the orders one.
- 02
Add the rules
For each column choose the rule and whether the row is only flagged or excluded: for example an amount that is not negative.
- 03
Read the score
On each run you see the score and the rows in quarantine, and decide what to fix.
In detail
- Rules
- Range, not null, allowed values, expression.
- Effect
- The row is flagged or excluded.
- Quarantine
- Excluded rows are kept in a table for 90 days.
- Quality
- One score per run; the people who own the rules are told when it changes.
- Labels
- Bronze, Silver or Gold, by how the query is built.
- Live Query
- A query can become incremental, with its table kept up to date by a flow.
Frequently asked questions
A rule asks whether a value is possible; an alarm asks whether a situation needs attention. They are two different tools, each with its own page.
They go to a quarantine table and stay there for 90 days.
It can propose them: you accept them.
Bring us a question you can't answer today.
We start from a real question your business has and walk the path from source to dashboard with your systems, not a demo dataset.