Strategy

Data Readiness Assessment

Test the data behind one decision before you pay to build on it.

GIST stepGround
Download the PDF

Get the Data Readiness Assessment overview

A short PDF to share with your team. It covers what is in the package, how it works, and a worked example.

The problem

Data projects get approved before anyone looks at the data.

A use case is approved from a slide. A few weeks into the build, the team finds duplicate customer IDs, a source nobody has access to, and fields that are mostly blank. Half the budget is gone and the decision is no better.

What's included

One decision, checked against the real data.

01

Profiling and join checks

Column types, fill rates, ranges, and top values for every source in scope, run on real extracts. Join checks show how many keys in one system turn up in the systems it has to join to.

02

Readiness checks

Access, ownership, completeness, validity, uniqueness, timeliness, join coverage, and personal data, each scored against what the decision needs. Columns that look like personal data report counts and never show sample values.

03

Ranked gaps

Each gap is ranked by the effort to fix it and how much it affects the decision. Source owners review the list and correct it where they know better.

04

Decision Brief and verdict

A draft brief for the Decision Owner to sign, a plan for the first build increment, and a verdict of proceed, fix first, or stop. The verdict can gate a pipeline, and a JSON record tracks readiness as fixes land.

How it works

From a decision to a verdict.

  1. Frame the decision

    With the Decision Owner, agree what the decision is, who acts on it, how you will know it improved, and what that is worth.

  2. List the sources

    For each system, record the owner, the access route, the business key, the fields needed, how fresh the data must be, and what it joins to.

  3. Pull extracts

    A recent full extract or a large sample is enough. Direct database queries work too.

  4. Agree pass levels first

    Set thresholds with the Decision Owner before the run, so nobody argues about the results afterward.

  5. Run and review

    Go through the ranked gaps with the source owners.

  6. Decide

    The Decision Owner signs the brief, or the work waits for fixes or stops. Either way there is evidence behind the call.

Worked example

Choosing which customers get a retention offer each month.

The decision pulls from a CRM, an ERP, and a helpdesk. We seeded the sample data with the problems real projects hit.

Seeded in the data
  • Duplicate customer IDs in the CRM
  • Orders for customers the CRM has no record of
  • Invalid customer segment values
  • A helpdesk extract that is 40 days old
  • A source the team cannot access yet
Result

The run catches each problem, ranks the gaps, and recommends fixing the data first. It writes the readiness report, the draft Decision Brief, and a record of every check.

Where it fits

What it runs on, and where it stops.

Works with

  • CSV extracts from any source system
  • SQLite, or any database with a standard Python connection
  • Your own environment, wherever the data is allowed to sit

Assumptions and limits

  • Type and personal-data detection rely on patterns, so source owners review the profiles
  • Join checks confirm that keys match. Owners confirm that the joined records mean the same thing
  • Very large tables are profiled from a representative sample
Ask Dex

Ask about the Data Readiness Assessment

Dex answers from our own pages and links to where it found the answer.

Have a decision waiting on budget?

Name it and the systems behind it. We will tell you what a readiness check would look at first.