Data Readiness Maturity Model

Data readiness maturity describes how reliably an organisation can maintain usable, understood and appropriately controlled data over time. It differs from the readiness of one AI use case at one decision point. A carefully bounded pilot may be possible in a relatively immature data environment, while a mature data function can still lack the source rights or historical evidence needed for a new task.

This guide presents an original capability model for data owners and transformation leaders. It is not a certification scale, a regulatory requirement or a measure of general organisational AI maturity. Use it to plan sustained capability improvement alongside enterprise data and AI readiness assessment, which answers the more immediate question of what a defined activity can responsibly do next.

Define the data domain and the observation period

A maturity assessment needs a boundary: the data domain, entities, systems, users and period being considered. “Our data is mature” is too broad. A water utility may manage meter-reference identities well while having incomplete histories for contractor notes or customer correspondence. Do not average unrelated domains into a reassuring corporate label before understanding their different responsibilities.

The hypothetical example here concerns meter asset-reference data used to support internal maintenance planning. It covers meter identifiers, installation and replacement relationships, reference locations and source ownership. It does not assess billing decisions, water-supply control, individual customer profiling or the safety of field work. Those activities have different prerequisites and consequences.

Choose an observation period that includes meaningful operating events, such as source updates, meter replacements, access changes and detected defects. A single clean extract can demonstrate a useful snapshot but cannot prove that the organisation repeatedly handles changes well. If few changes occurred, describe that evidence limit instead of awarding a high maturity level simply because no failures were observed.

Assess capabilities, not the number of purchased tools

A catalogue, quality platform or data-governance committee can support capability, but its existence is not evidence that the relevant work happens consistently. Ask whether users can identify the authoritative meter record, explain an identifier change, obtain permitted access and trace a correction into dependent views. These are observable outcomes, not a software inventory.

The model considers six capability areas: ownership and decision rights; identity and business meaning; provenance and change traceability; quality measurement and remediation; access and permitted use; and maintenance of accepted evidence under change. A domain can have different levels in each area. Preserve that profile rather than forcing a single number that conceals the weakest relevant prerequisite.

The UK Government Data Quality Framework emphasises fitness for purpose, measurement and improvement. The maturity levels below are this guide's original practical design, not that framework's official rating system or a Saudi compliance scale. Their usefulness depends on observable evidence and a clearly defined domain.

Use five evidence-based levels

Proposed data-capability maturity levels

Level	Meaning	Evidence that supports the level
0: Unestablished	The assessed capability has no demonstrated arrangement within scope	Investigation confirms missing ownership, definitions or process; unknown remains a separate state
1: Locally owned	Named people handle the work, but execution depends on individual knowledge	Examples of decisions and corrections with identifiable owners and material informal steps
2: Repeatable	Documented rules and roles are applied through a defined process	Multiple comparable instances, preserved outputs and explicit exception handling
3: Measured	The process is evaluated against defined outcomes and corrected when it falls short	Comparable measures, reviewed causes and accepted follow-up across the observation period
4: Sustained through change	The capability remains effective through relevant changes and dependencies	Evidence of source, role or process changes handled with traceable impact review and retained boundaries

Unknown is not level zero. Zero is an assessed absence; unknown means insufficient evidence to make the judgement. A level should also carry a confidence explanation and the period covered. The scale is ordinal: four is a later evidence state than two, not twice as much capability. Avoid arithmetic that implies otherwise.

Translate each level into domain-specific observations

For meter identity and meaning, locally owned might mean an experienced steward can explain identifiers but the rule is not consistently documented. Repeatable requires a documented distinction between an asset identifier, manufacturer serial number and installation relationship, applied to actual replacement cases. Measured adds evidence that violations are detected and resolved through an agreed method. Sustained through change requires showing that a source or lifecycle change did not silently break those meanings.

For access and permitted use, a named administrator is not enough. The evidence should show who decides access, what purpose is approved, how changes are handled and whether dependent views preserve the intended boundary. A mature arrangement does not mean everyone can obtain data quickly; it means the organisation can make and maintain appropriate, explainable access decisions.

For quality measurement, a monthly percentage alone does not establish the measured level. The population, rules and method must be defined, the results reviewed, and material problems acted on. If the metric changes every month without reconciliation, it cannot support a reliable capability trend. Representative data-quality measurement explains how to distinguish measured evidence from convenient samples.

Worked longitudinal record: meter replacements

The following sequence is hypothetical. In January, a utility's reference team can resolve replacement questions through one experienced steward, but the relationship between retired and installed meters is held partly in email. In June, an approved replacement rule and versioned mapping process are used across several reviewed cases. In September, the team has comparable exception measures and evidence that agreed corrective actions were accepted.

Illustrative evidence accumulated over three review points

Capability	January evidence	June evidence	September evidence
Replacement meaning	Steward explains cases from personal knowledge	Approved lifecycle rule applied to preserved examples	Rule exceptions measured and owner decisions recorded
Traceability	Current table lacks a consistent link to prior records	Versioned mapping links old and new asset references	Dependent planning views are included in impact review
Access and use	Existing access is visible but its decision basis is unclear	Named owners review the intended internal use	Evidence remains incomplete for a new contractor audience

The first capability may progress from locally owned to repeatable and then measured if the evidence satisfies the agreed anchors. The access capability does not automatically advance because identity handling improved. The new contractor audience remains an unresolved boundary for any use case involving those users. A maturity profile should preserve that asymmetry rather than average it away.

September's traceability evidence is promising, but it does not by itself establish sustained performance through every relevant change. The next review might examine an actual source migration or a controlled change exercise. State whether the evidence came from ordinary operations or a rehearsal, because those observations support different confidence levels.

Keep a capability profile with evidence and next proof

Illustrative current profile for the scoped meter-reference domain

Area	Proposed level	Evidence limit	Next proof to obtain
Ownership and decision rights	2: Repeatable	Cover arrangements during owner absence are untested	Demonstrated handover with the same decision quality
Identity and meaning	3: Measured	Assessment covers the included replacement types	Evidence for a newly introduced lifecycle exception
Provenance and change traceability	2: Repeatable	Some downstream views have only one reviewed change	Repeated impact reconciliation across relevant changes
Quality measurement and remediation	3: Measured	Comparable measures exist only for the stated period	Continued cause review and accepted closure
Access and permitted use	1: Locally owned	New contractor audience has unresolved evidence	Documented and demonstrated purpose/access review
Evidence maintenance under change	2: Repeatable	Limited source-change experience	Accepted evidence refresh after a material change

This is a profile, not a score of 13 out of 24. Adding ordinal levels would obscure their meaning and the different evidence limits. If leadership requires a summary, describe the strongest sustained capabilities, the consequential gaps and the investment needed. Preserve the underlying profile so a sponsor can interpret it for a particular AI proposal.

Measure progress without confusing volume and capability

Useful measures can include the proportion of in-scope sources with accepted ownership, the age of unresolved meaning questions, the recurrence of a verified defect and the time required to trace a change into dependent views. Each needs a denominator, observation rule and purpose. Choose measures that indicate whether the capability works, not merely whether the team produces more documents.

A rising number of detected issues may indicate better inspection rather than deteriorating data. A falling count may indicate reduced coverage rather than improvement. Interpret the measure alongside changes in population and method. If new meter types enter the domain, explain whether the earlier baseline remains comparable and which new conditions have not yet been assessed.

Do not make the number of closed tickets the sole success measure. A team can repeatedly close the same export correction without removing its source cause. Data remediation backlog design connects closure to accepted evidence and recurrence. Maturity improves when the organisation can maintain the necessary capability, not when a temporary project generates a large activity count.

Choose the next capability investment from actual demand

One useful service measure is the time required to answer a properly scoped evidence request. Record when the request was complete, when an authoritative answer was accepted and what waiting time reflected missing decisions rather than technical extraction. Faster delivery is beneficial only if meaning, permission and traceability are preserved. A quick spreadsheet assembled without those conditions should not be counted as evidence of a more mature data service.

There is no requirement to reach level four everywhere before useful AI work begins. Prioritise capabilities needed by the planned use cases and their consequences. A restricted internal reference assistant may need strong source meaning and access evidence within a narrow domain. A broader prediction task may additionally need historical availability and mature outcomes that this domain profile has not assessed.

Define the next investment as a capability outcome. “Buy a catalogue” is an implementation choice. “Enable an authorised reviewer to trace a meter replacement from source decision through the planning view, with preserved meaning and exceptions” is a capability outcome. The latter can guide tool selection, process design and acceptance without assuming that one product automatically supplies the capability.

Use a small number of focused improvement objectives with owners and evidence dates. Avoid a sprawling maturity programme whose activities cannot be connected to business decisions. AI readiness versus AI maturity explains the broader distinction; this model remains focused on the data-management capabilities that can support repeated, well-bounded readiness decisions.

Review change resilience rather than only steady-state quality

A domain can appear well managed while its principal steward and source system remain unchanged. Test what happens when an owner leaves, an identifier format changes, a source is replaced or a new audience requests access. These changes reveal whether knowledge and control are embedded in the process or depend on one person's memory.

Use actual change evidence where available and proportionate controlled exercises where necessary. An exercise might ask a second authorised steward to interpret a replacement case from the documented record, or trace a proposed new field into dependent views. Record the expected result and observed gap. Do not claim that a rehearsal proves every operational change will succeed.

Catalogue and lineage evidence supports this review by connecting source versions, transformations and consumers. A diagram that has not been maintained through changes is a description, not evidence of resilient capability. The maturity finding should identify how the organisation knows that its traceability still reflects the current domain.

Maintain a reusable capability evidence record

For each capability, keep a record that a later assessor can inspect without reconstructing the entire programme. It should explain the claim, observation period, relevant events, evidence selection, exceptions and reason for the proposed level. Link to accepted source artifacts rather than copying an uncontrolled set of operational records into a maturity presentation.

Filled capability evidence record for replacement meaning

Field	Illustrative entry
Claim assessed	Replacement relationships are interpreted consistently under the approved lifecycle rule
Domain boundary	Included meter types and internal maintenance-reference uses
Observation basis	Preserved cases from the stated review period, with selection and exclusions documented
Evidence of repetition	Different authorised staff apply the same rule to comparable replacement events
Evidence of measurement	Defined exceptions are counted, reviewed and linked to accepted actions
Known limit	A newly introduced lifecycle exception has not yet been assessed
Next review	Inspect the new exception and its effect on dependent planning views

Record whether the evidence selection was representative, deliberately challenging or based on all relevant events in a small population. A maturity conclusion can draw on several methods, but each method supports a different claim. Repeated success on hand-picked examples is useful demonstration evidence; it is weaker evidence about ordinary process consistency if the selection excludes difficult events.

Include counterevidence. An unresolved replacement, a missed notification or an owner handover that failed can reveal the limit of an otherwise well-documented capability. Do not remove those observations because they lower the level. Explain whether they are isolated, recurring or outside the stated boundary, and identify the evidence needed to decide. A credible profile can acknowledge weaknesses without losing its usefulness for investment planning.

Interpret apparent maturity reversals carefully

A lower rating at a later review does not always mean that staff performed worse. The domain may have expanded, the evidence standard may have become clearer, or new inspection may have exposed an old weakness. Separate genuine capability deterioration from a changed measurement basis and improved knowledge. Preserve both the earlier conclusion and the reason the newer evidence supports a different judgement.

For example, adding a contractor audience can reveal that access decisions previously depended on informal internal relationships. The access row in the illustrative profile concerns the capability to assess and maintain such purpose and audience changes. It does not say that every existing internal permission is invalid. The evidence supports a named local owner, while a consistently applied process for the new request remains unproven.

A source migration can similarly invalidate a formerly useful traceability map without erasing all capability. Ask whether the organisation detects the mismatch, assigns the impact review and restores accepted evidence. The ability to recognise and manage a disruption is itself relevant to maturity. Treating every disruption as proof of total immaturity would discourage transparent reporting and hide the difference between controlled recovery and unmanaged loss of meaning.

Set a review rule for provisional ratings. If the available period is short or important changes have not occurred, describe the level as supported within that limited evidence, with the next observation required. Do not manufacture certainty by demanding a fixed number of months for every domain. The relevant question is whether the observed events and methods are sufficient for the capability claim being made.

Separate maturity improvement from permission for a new use

A stronger capability profile makes future assessments more efficient, but it does not pre-approve every new use. The meter-reference domain may be suitable for internal maintenance information while a proposal involving customer data, external contractors or automated operational decisions requires different evidence. The maturity result should not be presented as a reusable blanket authorisation.

Likewise, a successful narrow readiness decision does not prove a high domain maturity level. A team may assemble a carefully controlled one-off extract through substantial manual effort. That can support a limited experiment while remaining difficult to reproduce or maintain. Record both conclusions without treating them as contradictory: the bounded activity is supportable, and the broader capability still needs work.

When presenting results to leadership, state the domain, observation period, evidence profile, unresolved boundaries and next capability objectives. To discuss data readiness maturity, prepare examples of source changes, repeated defects, ownership decisions and accepted corrections. Those examples reveal sustained capability more clearly than a self-rated questionnaire about whether the organisation values data.