Guidance

Data quality issues framework

Published 11 November 2025

Who this framework is for

This framework is for anyone with an interest in, or responsibility for, the quality of data within a business area or for a specific data asset. It defines a data quality issue, and explains how to identify issues when they occur and assign a priority to the issue.

If you’re a data practitioner in a public sector organisation, you should use this framework to help create a data quality action plan (DQAP). It will be especially useful for steps 3 to 5 of a DQAP:

• assess current data quality
• prioritise improvements and set goals
• identify root cause and take action

The framework includes a working example based on the preceding sections, and a glossary of some of the terms used in this guidance, such as data item.

What data quality rules and issues are

To maintain high-quality data in your organisation, for each of your data assets you need to:

• understand what the purpose of the data is – including its importance and impact
• create data quality rules
• set targets and performance bands against each data quality rule

Data quality rules are checks that help make sure your data meets a required standard. They define what ‘good’ data looks like and help you spot errors or inconsistencies.

For example, you might set:

• a completeness rule that states a field cannot be left blank (ie no null values)
• an accuracy rule that states a value in a field must be within a valid range – such as dates not being set in the future

You can assign different levels of priority or criticality to each rule, and it may or may not be acceptable for a data asset to contain errors.

Once you’ve defined the rules, you can then measure the performance of the data asset. If its performance is below the target, this is a data quality issue.

Determine the importance of purpose

Each data quality rule should document a purpose – like in Table 1 which refers to a data set that was used to book resources for training courses.

Table 1: resources for training courses

Purpose	Part of data	Requirement	Relevant aspect	Measurement	Metric
To allocate the right resources for training	Trainers’ details	Trainers’ names should be complete	Completeness	Trainer name should not be null	Percentage of records with names filled in

Based on its importance to your organisation or the wider public sector, allocate the purpose one of the following importance categories:

1. Low
2. Medium
3. High
4. Critical

Low

If the purpose fails or is disrupted, it might:

• incur minor unnecessary losses of costs, time or other resources 
• cause minor inconvenience to a person in the exercise of their rights and freedoms

Medium

If the purpose fails or is disrupted, there is a high risk of:

• incurring excessive unnecessary losses of costs, time or other resources 
• causing unnecessary inconvenience to a person in the exercise of their rights and freedoms

High

If the purpose fails or is disrupted, there is a very strong risk that:

• you will not achieve your goals without significant unnecessary losses of costs, time or other resources
• your organisation will be unable to meet regulatory requirements
• a person will be able to exercise their rights and freedoms, and do so without unnecessary significant inconvenience

Critical

If the purpose fails or is disrupted, there is a major risk that:

• there will be a serious negative impact on you achieving your goals
• your organisation or the government will break the law
• there will be loss of life or personal injury
• there will be loss or destruction of property (including financial loss)

Determine the impact of non-conforming data

Consider how severe the impact of a data item in the data asset not meeting the rule condition would be on the ability to fulfil the purpose.

Allocate the data an impact category and associated score, referring to Table 2.

Table 2: categories of impact on the purpose

Category	Score	Impact on the purpose if the data set value, or the data set itself, fails to meet the rule condition
Low	1	Fulfilling the purpose may be affected in a non-significant manner
Medium	2	Fulfilling the purpose may be affected in a non-trivial manner
High	3	Fulfilling the purpose may be seriously compromised
Critical	4	It may not be possible to fulfil the purpose

Set a data quality target

Multiply your importance and impact scores to derive a combined importance-impact score. Use this score to decide how much of the data set it’s tolerable to have not passing the rule. You can then allocate ranges to the combined score.

Refer to Table 3 for suggested rule target ranges. As an example, if your importance-impact score was 4, you would set a target between 90% and <97.5% – such as 95%.

Table 3: suggested rule target ranges

Importance × impact score range	Suggested data quality target range
1-3	75% to <90%
4-6	90% to <97.5%
8-9	97.5% to <99.9%
12-16	99.9% to 100%

Set below-target performance bands

If the data fails to meet the target, use performance bands to grade the extent to which it failed to do so.

By assigning a score to the band the data falls into, we can derive a third factor to use in combination with the importance and impact scores to assign a priority to the issue.

The performance bands and their scores are:

• 3 – Low
• 2 – Middle
• 1 – High

If it’s not acceptable to have errors in the data asset

If your target is 100% there is no need to set Low, Middle and High bands. Instead, grade the performance as Critical and give it a score of 4.

If it’s acceptable to have errors in your data asset

Based on the data quality target you’ve set, work out your Low, Middle and High performance bands, starting with the Middle band.

Refer to Table 4 for suggested performance bands based on common data quality targets.

The Middle band represents the point at which the number of errors in your data is double the target you’ve set for your rule. For example, if your quality target is for 75% of the data to pass a rule, it means it’s acceptable for up to 25% of data fields to fail to meet it. In this case, your Middle performance band would start at 50% (75% minus 25%) invalid data fields.

Performance that falls into the Low band is therefore anything from 0% to the start of the Middle band.

The High band start point is halfway between the Middle band start point and the target.

Table 4: suggested Low, Middle and High performance bands

Data quality target	Low band	Middle band	High band
75%	<50%	50% to 62.5%	62.5% to 75%
90%	<80%	80% to 85%	85% to 90%
95%	<90%	90% to <92.5%	92.5% to <95%
97.5%	<95%	95% to 96.75%	96.75% to 97.5%
99.9%	<99.8%	99.8% to 99.85%	99.85% to 99.9%

Assign a priority to a data quality issue

You should prioritise each data quality issue to determine:

• the urgency of addressing it
• the resources to allocate to investigating and potentially fixing it

Work out which band the performance lies in (Low, Middle or High) and multiply the number allocated to this band by the importance-impact score. 

Table 5 gives examples of which priority to assign to the issue based on the outcome of this calculation.

Table 5: issue priority ranges

Importance × impact × performance band	Issue priority
1-7	Low
8-16	Medium
17-32	High
33-64	Critical

Working example

Based on the training resources data in Table 1, imagine you set an importance score of 2 to the purpose of this data. You might give a score of 2 because failure of the purpose (allocating the right resources for training) might cause unnecessary extra time. Or, if the course went ahead without the right resources, it would be less useful to attendees, causing them unnecessary inconvenience.

Then imagine you set an impact score of 2 (Table 2). You might give a score of 2 because if the trainer’s name is not populated, this might indicate that a trainer has been allocated without their name being recorded. This would make it hard to know whether the trainer has the right skills and experience, which would have at least a non-trivial impact on the purpose. This makes a score of 2 feel appropriate.

The combined importance × impact score would therefore be 2 × 2 = 4.

Based on the suggested rule target ranges in Table 3, you would then set a data quality target within the range 90% to <97.5% – for example, 95%.

Next, you would assess how well your data is performing. Refer to the DQAP implementation guide to support this step. In this example, the data item has been assessed against a completeness rule. A data steward determined that 94.2% of the data was complete against the data quality target of 95%. 

The Low, Middle and High performance bands were calculated based on Table 4. This places the data item in the High performance band, with a performance score of 1.

Table 6: example performance bands

Current performance	Target	Low band	Middle band	High band
94.2%	95%	<90%	90% to <92.5%	92.5% to <95%

Next, multiply the importance score by the impact score by the performance score to identify the issue priority for the data item: 

 2 × 2 × 1 = 4

Using Table 5 as a guide, a score of 4 indicates a Low priority data quality issue. This feels intuitively right because a score of 94.2% is only slightly below the 95% target, so the data item is performing well and close to meeting the required standard. This suggests that it could be more effective to focus resources on other data items with greater importance and impact, or lower performance bands.

Allocate time and resources

Based on the issue priority, use Table 7 to decide when to treat the issue (how urgent it is) and how much money and resources you should allocate.

Table 7: issue timescales and resource allocation

Issue priority	Timescales	Money	Remedial resource allocation
Critical	Immediate or very short term	Allocate necessary funds preferentially over standard operations, planned investments or lower priority issues	Allocate effort time preferentially over standard operations, committed business projects or lower priority issues
High	Short term	Allocate necessary funds based on cost versus benefit preferentially over lower priority issues	Allocate effort time preferentially over lower priority issues
Medium	Medium term	Allocate necessary funds based on cost versus benefit preferentially over lower priority issues	Allocate effort time preferentially over lower priority issues
Low	Medium to long term	Allocate necessary funds based on cost versus benefit if no higher priority issues exist	Allocate effort time if no higher priority issues exist

Consider what these timescales mean to your organisation in practical terms. It might help to establish a standard for what period covers short term, medium term and so on. This will ensure you treat timescales consistently throughout your organisation.

As the basis of this, you might consider quantifying timescales in terms of measurement cycles. Each timescale would reflect how many times (cycles) you measure the data asset against the rule before you investigate and resolve the issue.

The rationale for this is that, assuming you’ve set the measurement intervals with reference to some combination of the rate at which the data changes and the frequency with which it’s used, this calibrates issue management against these factors.

Table 8 suggests an example for using measurement intervals to determine issue investigation and resolution timescales.

Table 8: resolution timescales

Issue priority	Investigation and remediation timescale
Critical	Immediate
High	Within 1 measurement
Medium	Within 3 measurements
Low	Within 6 measurements

You can use the Actions section of the DQAP template to record your targets for resolution and review dates.

Glossary

Data quality rule

One of the specific conditions a data asset must meet to be fit to support an intended purpose, defined in one dimension of data quality.

Data item

An element of a data asset to which a rule can be applied – this might be a field in a record, or an entire record.

Non-conforming data

Data that fails to meet the conditions of a rule.

Purpose

A purpose that the use of the data seeks to achieve.

Importance of purpose

The importance of the purpose that the data asset supports – both to your organisation and the rest of the public sector.

Impact on purpose

If a rule fails to pass, this represents the actual or potential effect it has on the purpose.

Target

How much of a data asset needs to pass a rule to be considered of sufficient quality to support the purpose.

Data quality performance

The actual performance of a data asset when measured against the rule – usually measured as the number of records passed divided by the number of records tested.

Performance band

Bands of performance below the target, categorised as either Low, Middle, High or Critical.

Data quality issue

A problem with the fitness for purpose of a data asset, identified by its failure to hit a rule target.

Issue priority

The degree of preference you should give to investigating and resolving the issue in terms of timescales and allocating resources.