top of page

Machine learning visualisations within Power BI

Microsoft released the key influencers visualisation for Microsoft Power BI desktop in February 2019. This visual uses machine learning functionalities to provide users with the ability to quickly gain insights from their data through the identification of key drivers by listing components and factors that contribute to the value of a specific metric.

The value of this feature and how the key influencers visual delivers insight can be demonstrated using our HR Dashboard. For this example scenario, an HR employee is trying to understand why some new employees leave the organisation within 90 days. These leavers (<90 days in service) are defined as 'Bad Hires' within the 'Leave Type' field.

For more information about the HR dashboard and data set used click here.

In this example a key influencers visual was created to analyse the 'Leave Type' metric by adding this field to the 'Analyze' field well in the Visualizations panel. Next, within the visualisation itself the category 'Bad Hire (<90 days in service)' is selected:

Key Influencer Microsoft Power BI Machine Learning Artificial Intelligence AI visualisation
Fields selected within the Power BI key influencer visualisation

Lastly, the 'Explain by' fields are populated by factors that could potentially explain 'Bad Hires'. The following fields were added: Reason for Term (reason why a former employee left the firm), Employee Source (from where an employee was referred), Sex, Department, Age Group and State.

Once applied, the key influencer visual does its magic and the results of the analysis can be inspected:

Power Bi insight from data through AI influencers
Fields with an impact on the metric 'Bad Hire'

The visual identifies a correlation between 'Bad Hires' and 'Company Culture'. Moreover, a link between 'Bad Hires' and the 'IT department' was found. Based on this insight, an HR employee can take action by talking with IT employees to confirm whether cultural problems exist, thereby establishing a causal relationship between the culture and bad hires within the IT department. In a next step, actions can be taken to prevent future 'Bad Hires' from occurring.

This analysis also showed that Sex, Age Group and State did not significantly explain 'Bad Hires' and are therefore not considered as influencers.

Through this key driver analysis using machine learning functionalities, user can quickly identify correlations within their data without the need for programming skills. A causal relationship can be confirmed through a root cause analysis (in the scenario above, this was done through understanding cultural problems within the IT department) and actions can be taken to resolve business or IT critical problems. Moreover, through periodic reporting an end-user can follow up on whether the taken actions effectively resolved the problems.

Hope this is useful! Any questions or comments please feel free to leave below, or alternatively you can contact us at


bottom of page