Artificial Intelligence

Data quality issues as an operational risk

Katya Kamlovskaya

Katya Kamlovskaya

— 1 min read

When assessing risks stemming from AI use in your organisation, remember that one of the major operational risks is data quality issues. Why?

Poor data quality can cause incorrect decisions, inefficient operations, compliance issues, or reduced productivity. This is why it is important to assess the likelihood and impact of data quality issues as a risk.

How to approach it?

Assess risk likelihood

  • What data sources are we using? Are they reliable, consistent, reputable, known for accuracy?
  • What data collection methods were used?
  • Wha data governance practices are in place?
  • Are the existing data quality checks and validation processes effective?
  • Are there mechanisms in place to identify and correct errors, inconsistencies, or outliers?
  • Are these checks frequent enough, thorough enough?
  • What efforts are put into data cleaning - sufficient processes tools? Are the activities timely?
  • Are there any external factors that could impact data quality?

Assess impact of data quality issues

  • How can inaccurate data impact model accuracy?
  • How critical are the model's predictions?
  • What are the consequences of errors?
  • Can data quality issues can operational disruptions? (for example, incorrect/incomplete customer information can lead to delays in order processing; inaccurate inventory data - to stockouts or overstocking etc.)
  • Can you estimate the potential financial losses associated with data quality issues? (loss of revenue, legal and regulatory penalties, missed sales opportunities, costs of repeat data cleaning)
  • Can data quality issues cause reputational damage? (think customer trust, brand image, stakeholder relationships).

Create a risk matrix to visualise risk likelihood and impact

  • Place risk likelihood on one axis and risk impact on the other
  • Categorise data quality issues based on their position in the matrix
  • Risks to prioritise are those in the high-likelihood, high-impact quadrant

Now that the data quality risks are prioritised, this is where the fun begins: it is time to mitigate them!

One more thing to remember: in addition to the most obvious strategies like data governance and data quality management, organisational culture and data literacy training are a crucial aspect. If you promote data literacy within your organisation, it will help to ensure that employees understand the importance of data quality and how to handle data responsibly.

Read more