Optimize Department Performance Through Team Clustering

A New Approach to Department Improvement that saves time and effort

When working with multiple teams, I often face the challenge of having enough time to optimize only one team at a time. This one-by-one approach is slow and limits the overall impact, as each improvement benefits only a single team. 

While we can prioritize which team to work with first by identifying the lowest performers, a more efficient strategy is to identify improvements that could benefit multiple teams simultaneously.

• Are there improvements that could benefit all teams?

• Are there groups of teams that would benefit from the same improvements?

In this article, I explain at a high level how to find impactful initiatives that will benefit multiple teams at once, instead of working with one team at a time. 

I found groups of teams that benefit from the same initiative

What Did I Find in My Department?

After analyzing multiple variables representing team performance across seven teams in my department, I identified two main findings:

1. I found groups of teams where the same Improvements will increase performance.

I looked into how to make groups of teams that need improvement in the same areas. In other words, can we create initiatives to boost team performance in similar ways, saving time and effort?

Using the correlation between team survey answers, I identified four clusters of teams, allowing me to plan initiatives based on these groupings.

For instance, team 4 and team 3 both have low scores in observability and monitoring among other areas. From this understanding, we are going to plan the same workshop and action steps with them.

I clustered the teams into 4 groups and I will plan initiatives according to those, making our initiatives more impactful and efficient.

This grouping approach is highly useful for those responsible for team performance, such as Engineering Managers or Agile Coaches, as it enables us to optimize our initiatives

2. Is there a pattern across all teams?

Continuing with my analysis, I wanted to know if there is a specific area where the entire department could benefit from the same initiative. Examples might include areas such as code maintainability or customer feedback collection.

My analysis began with assumptions based on observations and past experiences. From a leadership perspective, there was a belief that certain areas might need attention across all or most teams.

We made an assessment of 17 areas using a survey. This bar chart shows the aggregated answers from our survey. I used the average value (red line) to assess if there is a strong pattern across all the teams.

Even though some average values are lower than others, when doing statistical analysis, I conclude that there isn’t any question in our questionnaire that significantly under-performs across all teams in the department.

This means there is no need for a department-wide action that will benefit all our teams, and acting at a department level without this insights would be a waste of time.

How did I find these patterns?

My analysis data came from a survey that we conducted across all teams in our department, which consisted of 17 questions related to various aspects of team performance.

From these responses, I first established a baseline using either the average or median, depending on the presence of outliers, to create a stable reference point.

How do I identify Improvements for the Entire Department?

To find possible department-wide improvements, I began with hypothesis testing to pinpoint questions with significantly lower scores. This helped see if the differences were just normal variations or if they show an important trend that needed attention.

My hypothesis test results showed that no question in our survey significantly underperforms across all teams.

How did I find Clusters of Teams?

To group teams that might benefit from similar initiatives, I used a correlation-based approach. 

The idea is that if two teams' responses are positively correlated, they likely perform well in the same areas and struggle in similar areas, indicating they could benefit from the same initiatives.

I applied Correspondence Analysis to transform the correlation data into Euclidean distances, allowing me to map each team onto a two-dimensional space to easily identify patterns and clusters.

The correlation shows if two teams underform in the same areas.

What I learned

If you work with many teams and want to optimize your initiatives with them, it is very useful to find groups of teams. This way you save time and effort. You can use simple statistics to both find groups of teams or even find what needs improvement across all your teams.