If you have ever wondered how you might be able to get a sense of the data skills within your team/s the Data Practitioners Network at Hounslow Council have created a new tool to map the data skills of all colleagues. In addition to showing us who has what skills, our skills matrix is a data informed method for grouping all data practitioners based on data skill familiarity.

As a reminder, within the Data Practitioners Network at London Borough of Hounslow we believe: “All colleagues are Data Practitioners”. 

To build our skills matrix, we wanted to make sure it was tool and supplier agnostic. Instead of asking how familiar colleagues were with specific tools, we focused on the data lifecycle. We asked about topics like “Creation or Collection of Data” or “Processing and Preparation of Data.” New tools emerge all the time, but the stages of data work can always be mapped to the data lifecycle.   

This also makes our skills matrix much more encompassing, welcoming everyone from data curious colleagues (who may be new to data tools) all the way through to those that have been working within stages of the data lifecycle for decades. 

For each skill, colleagues rated their familiarity on a scale:

1. Curious About It
2. Learning the Basics
3. Comfortable Using It
4. Can Teach Someone About This
5. Expert and Innovator

In addition to mapping individual skills, we use this scale to find groupings of people with similar skills in our network. 

We did this by Clustering our responses with an AI model. 

Our narrow AI model looked at everyone’s self-assessment skill scores (anonymised) and found similarities in the scores and found three distinct groups (or clusters) in our network. 

1. The Data Operations Group: This group are particularly familiar with data handling. They prepare data and build data workflows across the organisation. 
2. The Data Curious & Emerging Practitioners Group: This group represents our learners. They are comfortable with data preparation and collection but are generally less familiar with the data lifecycle. They’re curious about all things data, including AI, and are on a journey to build their expertise. 
3. The Data Insights & Analytics Group: These are the people who turn raw data into narratives and actionable strategies. They are particularly familiar with understanding what happened through the data (descriptive analysis), visualising data to make it clear, and recommending future actions (prescriptive analysis). 

Finding the groups was step 1, but we wanted to also understand what specific skills are the most important for telling these groups apart? 

We used another narrow AI tool to build flowcharts that ask a series of skill-based questions (e.g., “Are they familiar with archiving of data?”) to figure out which group a person belongs to.

This tool uses a method called Feature Importance, which tells us which questions (skills) were the most helpful in defining groups. The higher a skill’s importance score, the more it helped our AI tool distinguish between the different groups.

Looking at our results, here are the top questions we could be asking that would suggest which group an individual might belong to (I was particularly surprised that Data Visualisation was so low down!):

Top 10 Most Important Skills for Cluster Differentiation (Random Forest):

1. Prescriptive Analysis of Data
2. Programming Languages
3. Archiving and Retention of Data
4. Machine Learning
5. Predictive Analysis of Data
6. Disposal or Deletion of Data
7. Building Large Language Models (AI)
8. Descriptive Analysis of Data
9. Using Large Language Models (AI)
10. Data Visualisation

This is a new way we have developed to understand which groups and skills exist within our Data Practitioners Network.

With these groups in mind we can now offer targeted workshops. We will start to label our workshops to showcase whether they are targeted for the “Data Curious” or for the “Data Operations” group.

By focusing on the data lifecycle rather than specific tools, our skills map will remain relevant even as technologies evolve. We will also use this analysis to continue to build a better understanding of our collective strengths, and will rerun this analysis to work out the evolving groups as our network continues to grow.