AvailableIncluded in beta access

Micro-course · Intermediate

Control data quality with SQL

Turn data-quality suspicions into repeatable, prioritised checks.

60 to 90 minCommon SQL · PostgreSQL6 lessons · 9 activities available
Start the course

Free 30-day beta access · no payment · progress saved

The business problem

Duplicates, missing values and orphaned relationships are discovered too late in reporting.

What you will be able to do

Write a suite of checks and produce an actionable anomaly report.

  • Measure completeness and uniqueness
  • Detect orphaned records
  • Document a rule
  • Prioritise a correction

Scope of this micro-course

This micro-course targets one focused skill. It does not replace a complete path or server administration training.

Prerequisites

  • Filters, aggregations and joins

Project and practice

9 verified exercises are available. A business project is included.

Focused path

Your programme, in order.

Content remains centralised: completing an activity here also updates the path that contains it.

Micro-lessons

  1. 01
    Data quality is measured by rule

    Quality data is not just a table without NULL. It must respect the rules necessary for its use: unique identifiers, mandatory fields, authorized values ​​and consistency between sources.

    Open lesson
  2. 02
    A reusable method

    A good query is built in short steps. Run the source first, add a transformation, observe the result, then only the next clause.

    Open lesson
  3. 03
    Go from a correct query to a professional result

    In a professional environment, this technique is used in particular to secure an import, validate an indicator before publication or monitor consistency between two systems. However, a query is not reliable only because it executes: you must link each clause to a business rule, specify what a row represents and verify that the period, exclusions and missing values ​​are processed voluntarily.

    Open lesson
  4. 04
    The problem to solve before syntax

    A status column contains `paid`, `PAID`, a trailing space, and an unknown value. An impossible date or a poorly formed email can be silently ignored by overly optimistic analysis.

    Open lesson
  5. 05
    Build a controllable method

    Keep the raw value, create a separate normalized value, and classify each anomaly by rule. A quality report should measure coverage, uniqueness, validity and consistency across columns.

    Open lesson
  6. 06
    Check before interpreting

    Compare before/after normalization, do not automatically replace an ambiguous value and publish the number of returned records excluded from each KPI.

    Open lesson

Practical exercises

  1. 01
    Detect duplicate emails

    Transform a uniqueness constraint into measurable control.

    Do exercise
  2. 02
    Measure phone completeness

    Produce a quality rate from explicit volumes.

    Do exercise
  3. 03
    Reconcile orders and payments

    Compare two independent totals in a single control line.

    Do exercise
  4. 04
    Identify missing deliveries

    Distinguish a missing date from an empty string or status.

    Do exercise
  5. 05
    Analyze a time series with missing months

    Make absent periods visible before calculating evolution and moving average.

    Do exercise
  6. 06
    Deduplicate with a deterministic rule

    Choose a canonical email line despite identical dates and multiple sources.

    Do exercise
  7. 07
    Audit a KPI before publication

    Detect why a join with tags inflates revenue, then rebuild the KPI.

    Do exercise
  8. 08
    Clean an imperfect dataset before analysis

    Profile anomalies, define validity rules and calculate a trackable KPI.

    Do exercise
  9. 09
    Conduct an autonomous seller analysis

    Choose your own method to compare net income and reimbursements.

    Do exercise

Back to the full catalogue