How data quality rules, issue detection, remediation, and measurement ensure the organisation's data is accurate, complete, consistent, and timely.
DQT-CON-01
Confidence in accuracy
How confident are you that the data and reports you rely on are accurate enough to base business decisions on, without needing to independently verify them first?
Maturity level descriptions
Accuracy is unknown and generally distrusted; the Data Consumer independently checks or reconstructs figures before using them. Every significant figure is cross-checked against another source or rebuilt manually before it is used in a decision or communication.
Confidence varies by dataset based on personal experience; some sources are informally trusted and others informally avoided, with no documented basis. The Data Consumer knows from experience which reports “are usually right”, but this is personal judgement with no organisational backing.
Accuracy rules are defined for priority data and the results are communicated to business users, so confidence rests on stated checks rather than experience. The Data Consumer can point to documented accuracy checks applied to the data they use and knows those checks are performed.
Accuracy is measured against defined thresholds, reported to business users on a regular basis, and breaches are visibly acted upon. The Data Consumer sees a regular accuracy measure for the data they depend on and observes remediation when it falls below threshold.
Accuracy is continuously validated, with issues detected and corrected before business users encounter them and confidence measured as an outcome. The Data Consumer uses data without independent verification because errors are intercepted upstream, and their confidence is itself measured and managed.
DQT-CON-02
Timeliness and currency
How reliably is the data you use current enough for the decisions you make with it, and how clearly do you know how current any given figure actually is?
Maturity level descriptions
Currency is unknown and unmanaged; the Data Consumer cannot tell whether a figure is from today, last week or last quarter. Reports carry no indication of when data was last updated, and refresh happens unpredictably or not at all.
Refresh happens broadly on a known rhythm understood informally, but is not stated, guaranteed or visible. The Data Consumer believes a report updates overnight, based on experience, but has no confirmation and no notification when it does not.
Refresh frequencies are defined and published for priority data assets, and currency is displayed to business users at the point of use. The Data Consumer sees a stated refresh frequency and a last-updated timestamp on the reports and datasets they use.
Timeliness is measured against the published commitments, reported, and failures to refresh are proactively communicated to affected users. The Data Consumer is notified when a refresh fails or is delayed, before they act on stale data, and timeliness performance is reported.
Currency requirements are set by business need and continuously met, with automated detection and remediation of staleness before it affects decisions. The Data Consumer’s currency requirements have been captured explicitly and are met continuously, with staleness intercepted automatically.
DQT-CON-03
Completeness
How confident are you that the datasets and reports you use contain all the records and fields they should, rather than silently omitting parts of the population you are analysing?
Maturity level descriptions
Completeness is unknown; the Data Consumer has no way of telling whether records or fields are missing from what they are given. Analyses proceed on the assumption that the data is complete, with no basis for that assumption and no check available.
Gaps are discovered incidentally, usually when a total looks wrong, and are handled case by case without record. The Data Consumer notices missing records only when a figure contradicts expectation, and the gap is explained informally if at all.
Completeness rules are defined for priority datasets and known gaps and exclusions are documented and communicated to business users. The Data Consumer can consult documentation stating what a dataset does and does not cover, including known exclusions.
Completeness is measured against defined thresholds, reported, and material gaps are flagged to affected users as they arise. The Data Consumer sees a completeness measure for key datasets and is alerted when a material gap appears rather than discovering it themselves.
Completeness is continuously monitored and enforced at source, with gaps prevented or corrected before data reaches business use. The Data Consumer works with data whose completeness is assured upstream, with automated controls preventing incomplete data from being published.
DQT-CON-04
Visibility of data quality status
When you open a dataset or report, how clearly can you see its current quality status — known issues, certification, or any warning that it should be used with caution?
Maturity level descriptions
No quality status is visible to business users; the Data Consumer cannot distinguish a trusted asset from a known-problematic one. Every asset looks equally authoritative regardless of its actual condition, with no indication of known problems.
Quality problems are communicated ad hoc by email or word of mouth when someone remembers, and are not attached to the asset itself. The Data Consumer may receive an email warning about a known issue, but nothing on the asset itself records it.
A defined mechanism marks quality status on business-facing assets — certification, known-issue notices or quality indicators — and business users are shown how to read it. The Data Consumer can see whether an asset is certified and whether any known issue is currently recorded against it.
Quality status is measured, kept current, and reported across the business-facing estate, with coverage of certification tracked. The Data Consumer finds current quality status on essentially all the assets they use, and the organisation measures how much of the estate carries it.
Quality status is generated automatically from live monitoring and surfaced in context, changing in real time as conditions change. The Data Consumer sees a live quality indicator that updates automatically when a check fails, without anyone manually posting a notice.
DQT-CON-05
Reporting a data quality issue
When you find something wrong in the data — a value that cannot be right, a missing record, a duplicate — how clear and easy is the path for reporting it?
Maturity level descriptions
No path exists for business users to report data quality problems; the Data Consumer corrects the figure locally or works around it. The Data Consumer manually fixes the number in their own spreadsheet and the underlying error is never reported.
Issues are reported informally to whoever seems relevant, with no record, no ticket and no expectation of a response. The Data Consumer mentions the problem to a colleague or analyst, and whether anything happens depends on that individual.
A defined, communicated reporting route exists (form, ticket category, in-tool feedback), and business users know how to use it. The Data Consumer reports issues through a named channel and the report is logged and routed to a defined responder.
Reporting is easy and in-context, reports are tracked with stated response expectations, and reporting volumes are monitored as a quality signal. The Data Consumer can raise an issue directly from the report or dataset, receives acknowledgement within a stated timeframe, and reporting patterns are analysed.
Most quality issues are detected by monitoring before business users encounter them; user reports are a supplementary signal feeding continuous improvement. The Data Consumer rarely needs to report an issue because it has already been detected and flagged, and reports that do arise feed control improvement.
DQT-CON-06
Resolution and closure of reported issues
Once you report a data quality issue, how effectively is it resolved, and how well are you kept informed of what happened?
Maturity level descriptions
Reported issues disappear; the Data Consumer receives no acknowledgement, no outcome and no evidence of change. Issues raised by business users are not tracked, and the same error recurs indefinitely.
Issues are sometimes fixed, depending on who receives them, with no consistent communication back to the person who raised them. The Data Consumer occasionally notices a problem has been corrected but is not told, and other reports go nowhere.
A defined resolution process exists with reports assigned to an owner, and the outcome is communicated back to the business user who raised it. The Data Consumer receives confirmation of what was found and what was done, closing the loop on each report.
Resolution is tracked against defined response and fix expectations, with performance reported and root cause recorded for significant issues. The Data Consumer’s issues are resolved within stated timeframes, and they can see that recurring problems have documented root causes.
Resolution feeds systematic prevention: root causes are fixed at source, recurrence is measured, and the Data Consumer sees issue rates fall over time. The Data Consumer observes that the same class of problem stops recurring, with recurrence measured and reported as an improvement outcome.
DQT-CON-07
Fitness for purpose
For the decisions you make, is there a shared and explicit understanding of how good the data needs to be — how much error, delay or incompleteness can be tolerated before the decision is materially affected?
Maturity level descriptions
No consideration is given to how good the data needs to be; the Data Consumer uses whatever is available without any tolerance being expressed. Data is used for decisions of very different consequence with no differentiation in the standard applied.
An informal sense exists that some decisions need better data than others, but nothing is documented or agreed with data providers. The Data Consumer applies their own judgement about when data is “good enough”, without discussing it with those who supply the data.
Quality requirements are documented for priority business uses, stating what standard the data must meet for those decisions. The Data Consumer has agreed documented quality expectations with data providers for their most significant uses of data.
Actual quality is measured against the documented business tolerances, with breaches reported to affected decision-makers. The Data Consumer is informed when data falls below the standard agreed for their decision, before that decision is made.
Tolerances are actively managed and continuously monitored, with data automatically withheld from or flagged for decisions it is not fit to support. The Data Consumer is prevented or warned automatically from using data that does not meet the standard required for a given decision.
DQT-CON-08
Quality of AI-assisted outputs
When an AI or GenAI tool produces an output you use for work — a summary, an answer, a generated analysis — how well can you judge its reliability and report it when it is wrong?
Maturity level descriptions
No basis exists for judging AI output reliability and no route exists for reporting errors; the Data Consumer accepts or discards output on instinct. AI-generated content is used or ignored based on whether it “looks right”, with no source shown and nowhere to report a wrong answer.
Business users are informally warned that AI output may be wrong, but no supporting information or reporting route is provided. The Data Consumer has been told to check AI output, without being given the means to check it or anywhere to report failures.
AI outputs used in business tools cite their sources, and a defined route exists for business users to report incorrect output. The Data Consumer can see which documents or data an AI answer drew on and can report an incorrect answer through a named channel.
AI output quality is measured, reported errors are tracked to resolution, and recurring failures are traced back to the underlying data. The Data Consumer’s reports of wrong AI output are tracked, and they can see that underlying source data problems are corrected as a result.
AI output reliability is continuously monitored, with degradation detected proactively and business-user feedback systematically improving both models and source data. The Data Consumer sees measurable improvement in AI output reliability over time, driven partly by feedback they and their colleagues provide.