I know what you’re thinking, “how can more number-crunching and data-mining help to personalise you employee engagement strategy?” Well hear me out…
We’re running a cluster analysis with an increasing number of our clients following their employee survey. You might also know the cluster analysis process as segmentation or profiling – essentially these are one and the same thing.
Using cluster analysis
Here’s what the cluster analysis process involves and what it gives you:
- Your engagement survey data is entered into statistical analysis (data-mining) software before sorting the employees into separate clusters
- Each separate cluster is made up of groups of employees with shared characteristics
- In a medium-sized company, running this type of analysis would typically give you three clusters – individuals with high engagement, moderate engagement and low engagement
- In doing this, we can determine what most engages (and disengages) a particular cluster of employees
- Analysing at a cluster rather than team or organisational level offers a far quicker, more intuitive and accurate way to understand and improve engagement.
A case study example
With one of our clients – a financial services organisation – we identified three clusters that were named: ‘Role model leaders’, ‘Sceptical but supported’ and ‘Disengaged and unheard’. The chosen labels for clusters usually depend on what the other variables are found within each profile.
Here’s a summary of the profile of each cluster for this organisation:
‘Role model leaders’ were leaders themselves, or leaders of leaders, and were highly engaged. They had a high performance rating and a high manager rating too (they rated their manager’s behaviours highly).
The ‘sceptical but supported’ group were moderately engaged but their performance ratings were lower. Although they had high manager ratings, they were typically distrustful of senior management.
‘Disengaged and unheard’ made up some 20% of the overall population. This cluster included those with the lowest engagement scores, the lowest manager ratings and the lowest performance scores. The fact that our analysis uncovered this pretty large, disengaged population was a big red flag for our client.
Having defined the clusters, we next used a key driver analysis based on our ‘Think, Feel and Do’ engagement model and questions. The analysis showed us which ‘think’ survey question scores – employees’ perceptions of their organisation – are having the biggest impact on their engagement index question scores. The most notable finding here was that the key drivers of engagement were different for each cluster.
Our analysis clearly demonstrated that what most significantly drives engagement depends greatly on how favourably their employees ‘think’ about the organisation, and ultimately, how engaged they currently are. Armed with this knowledge, our client was then able to be really specific about how they address and approach employee engagement initiatives within each of these groups, leveraging their ‘role model leaders’ to engage others in the organisation.
Taking data-driven action
Action planning following an engagement survey is a common area of struggle for companies. In larger organisations where you may be rolling out actions at a global level, these often fail to have the local impact desired. And, when taking a more localised team or regional approach to action plans, these can easily get lost in the sea of initiatives before falling off the radar altogether.
The ability to use your survey data to target action plans at a ‘cluster level’ with your employees helps bring into sharp focus both the size the problem and the prize. It’s essentially about taking a ‘valuing differences’ approach to your employee engagement strategy, which promises to be much more effective in affecting change.
After all, wouldn’t you want to know that engagement initiatives you’re investing so much time and resources into were the right ones, and that they had the very best chance of success? Surely, adapting your approach in this way is a no-brainer.