Last checked: 7 November 2026
In short: Data quality is how accurate, complete, current and consistent your information is. AI learns from data and acts on it, so poor data quality leads straight to poor predictions, wrong answers and bad decisions.
Table of Contents
How it works
- Set standards. A business decides what good looks like for each data set, such as no missing postcodes or no duplicate customers.
- Measure. Checks, often automated, count errors, gaps, duplicates and out-of-date records against those standards.
- Fix at the source. Teams correct the records and change the forms or processes that let errors in, then keep checking.
Where you’ll meet it
Duplicate contacts in HubSpot or Salesforce, blank cells in Microsoft Excel and returned post to old addresses are everyday signs of data quality problems.
What it means for your business
- AI magnifies errors. A model trained on wrong records repeats the mistakes across thousands of decisions.
- Name an owner. Give each important data set, such as customers or products, one person responsible for keeping it right.
- Check before you build. Before any AI project, measure the error rate in the data it will use. Fixing data is often the biggest part of the work.
Watch out for
- Bad data from outside. Supplier feeds, bought lists and customer uploads can carry errors into your systems. Check them on arrival.
- Slow decay. People move, change jobs and switch suppliers, so records go out of date every month.
Example
Unity Software. In May 2022 the games software company said a fault in its Audience Pinpointer ad-targeting tool had cut the accuracy of its models. Chief executive John Riccitiello told investors on the earnings call that Unity had also taken in bad data from a large customer, which damaged part of its training data. Unity expected the problems to cost about $110 million in 2022 revenue.
Data quality vs data governance
| Data quality | Data governance | |
|---|---|---|
| Question | Is this data right | Who decides how data is handled |
| Covers | Accuracy, gaps, duplicates | Rules, roles, access, retention |
| How they relate | One result governance aims for | The system that keeps quality up |
Frequently asked questions
How clean does data need to be before using AI?
Clean enough for the decision it supports. Measure errors in the fields the AI will use, fix the worst ones and test results on a small sample first.
Can AI help improve data quality?
Yes. AI tools can spot duplicates, odd values and missing fields far faster than people. A person should still approve changes to important records.
Who should be responsible for data quality?
The team that creates the data, with one named owner per data set. IT can provide checking tools, but the people entering the records are best placed to keep them right.