When this is the right conversation
Teams reconcile spreadsheets every month, duplicate records distort reporting, or nobody agrees on what counts as an active customer.
What the work involves
- Define critical elements and the decisions that depend on them; agree meanings before choosing tests.
- Specify validity, completeness, uniqueness, consistency and timeliness checks with fit-for-purpose thresholds.
- Measure an approved sample or approved environment, recording scope and known blind spots.
- Assign root-cause fixes, exceptions and repeatable monitoring rather than repeatedly cleaning the same export.
What we need from your team
A business definition owner, a technical implementer and an approved environment or redacted sample. Production changes require separate approval.
What useful completion looks like
Each reported quality measure has a defined numerator, denominator, scope, refresh schedule and accountable response when it fails.
A quality score does not establish accuracy for every use or certify the underlying data. Automated remediation is never assumed.
Before we begin
We agree the business objective, named delivery roles, access arrangements, fee, timetable and acceptance criteria in a written scope. Dependencies and out-of-scope work stay visible. We do not assume permission to change production systems or transfer sensitive information.
See the engagement processRelated field notes
A quality score needs a denominator
A field can be filled in—and still be wrong.
Consider a required account-owner field. A completeness check asks whether a value exists. A validity check asks whether it matches an allowed identifier. Neither establishes that the person is the correct owner. Define separate rules for separate questions, identify the population being checked and connect failures to a responsible person.
Illustrative scenario, not a client case study or a claim of past results.
The working evidence
- Rule specificationBusiness question, field, period, population and pass condition.
- Measure definitionNumerator, denominator, exclusions and approved threshold.
- Exception recordWhy an exception exists, who approved it and when it expires.
- Remediation evidenceCorrection, root cause, retest result and operating handoff.
Who does what
The domain owner approves what is fit for the business purpose. A steward coordinates issues and definitions. A custodian implements authorised checks. A measurement owner keeps the population and method comparable over time.
Take one failed example through detection, triage, correction and retest. Show what the rule proves—and what it does not. A format test is not evidence that a contact is genuine or that a record is accurate.
No universal quality percentage is a substitute for the intended use. A rate can rise because the denominator or exclusions changed; preserve those changes before claiming improvement.
Make the first workshop useful.
Bring roles, questions and approved examples—not an unrestricted data export.
- Which decision is affected by the suspected quality problem?
- What is the rule, denominator, time period and acceptable exception process?
- Who investigates a failed check, and what evidence closes the issue?