3 Data Quality Checks for BI Reports – The 3 C’s

Data Quality
Estimated reading time: 3 minutes

When building a Business Intelligence (BI) report, it is easy to spend more time doing data preparation and ensuring the data is high quality rather than making the report itself. Even if a report is the most user-friendly [1] and follows design and data visualisation best practice techniques to the letter [2], users can lose all trust if the data is wrong or misleading. In an ideal world, complete, consistent, and correct data will be provided. The development would be focused on report design and metric calculations. In reality, notable time and effort must be put in to ensure that data quality is as high as possible. Here are three vital but straightforward data quality checks to get you started! They are easy to remember as the 3 C’s…

1. Completeness

Is the data complete? The chances are the answer to this is, in most cases, is no. An entirely complete data source is ideal for a BI report. However, data collection techniques are often complicated and missing data may not be a  clear or simple issue to resolve. On the other hand, a crucial data quality check whilst building a report is simply exploring the data. Group it by the various aspects and aggregations and start identifying where/if there are missing data areas. It is key to assess the possible impact of missing data as data transformations may be required to account for missing values. For example, adding extra columns to flag rows or adjusting calculations.

2. Consistency with BI Reports

Is the data consistent within the BI report? Examples of consistency could be naming conventions of columns or categories within a dimension, metric definitions, and unique identifiers. This is more relevant when joining data from multiple sources/systems, ensuring that all systems’ relevant data align. It is worth cleaning the data if there are inconsistencies and working with key stakeholders to agree on the correct naming conventions and metric definitions. This can help produce a data dictionary that acts as a central, verified source of business metrics and definitions.

Love this content?

Sign up to our mailing list!

3. Correctness of BI Reports

Is the data correct? This can be difficult to answer because it suggests you know what the values should be, which may not always be the case. However, it is vital to check that the BI report’s data matches the source system, e.g. SQL tables and row counts. A BI report can only present the data that is available, and the values must align. A method of checking is to build a data audit within the report. Visualising the key metrics and highlighting inconsistencies, variances, or metric thresholds. Using colour, size, or icons, you can draw attention to potentially incorrect figures.

Having said that, if the data source itself is incorrect, the report draws attention to a broader issue (such as a user error creating duplicate records in a system). Still, the report remains accurate, reflecting the data source provided. When exploring data, it is always sensible to check any outliers with the relevant stakeholders to uncover data source issues. If not, it simply adds context to any unexpected trends outlined in the report – a win either way.

Concluding the '3 C's' of Data Quality

To summarise, this post’s purpose has been to provide three simple but essential data quality checks when building BI reports. The three C’s:

  • Completeness – ensure any vital transformations are included to account for missing data.
  • Consistency– take time to align naming conventions and metric definitions, particularly when joining data from multiple sources.
  • Correctness – check the data in the report matches the source system and, where possible, investigate any outliers.

 

Thank you for reading!

References

3 tips for successful benchmarking
Data
Jess Parker

3 Tips for Successful Benchmarking

Estimated reading time: 6 minutes What is benchmarking? Oberlo defines benchmarking as a “process where you measure your company’s success against other similar companies to

Read More »
5 Alternatives to Pie Charts
Data
Jess Parker

5 Alternatives to Pie Charts

Estimated reading time: 6 minutes Pie Charts Pie charts are a popular visualisation choice used to represent a part-to-whole relationship. In other words, the ‘parts’

Read More »
Skip to content