18 Diagnostic People Analytics
Sometimes you will want to know not just what happened (descriptive analytics) but why it happened. When you want to explain why specific events or patterns occurred, diagnostic analytics will be your best friend. Diagnostic analytics focuses on finding potential underlying reasons or factors related to particular results or trends. Diagnostic analytics can mostly be conducted using the same basic statistical analyses discussed previously, but requires slightly more sophisticated approaches to how and in what ways you explore the various descriptive analytics in order to address the question, “Why did it happen?”
Consider a situation where descriptive analytics identified that a company’s employee turnover rate has increased dramatically. The natural next question is, “Why did it increase?” Your focus has now shifted from what happened to why did it happen and you will need to use diagnostic analytics to search for potential underlying causes.
The analytics required for diagnostic analytics usually aren’t “harder” than those used in descriptive analytics – in fact, you can use all of the skills you already learned from your descriptive analytics skill building. The real effort of diagnostic analytics comes from the fact that it isn’t clear where to start. You no longer have a simple question like “what is our attrition rate?” which sort of implies what analysis is needed in the question itself. Instead, you now need to consider all possible factors that could potentially be related to attrition. This might involve many different factors and almost always requires compiling a lot of information on a range of different possible topics. For example, you might need to gather data on employee demographics, performance evaluations, pay, engagement survey results, comments from exit interviews and more. Only once you have the data can you start to examine the links between factors and identify potential causes of the high turnover rate by using diagnostic analytics approaches. This is where diagnostic gets both fun and tricky. There are just so many things the company could find:
- they might discover a link between low-performance ratings and higher turnover, indicating a correlation between poor performance and attrition.
- they might see that there is an impact of salary on retention, showing that employees receiving below-average salaries compared to market rates are more likely to leave their jobs.
- they might identify low engagement ratings among employees who ultimately quit, demonstrating a connection between engagement levels and turnover.
- they might analyze exit interview feedback revealing recurrent themes influencing employees’ decisions to leave, such as limited opportunities for advancement, ineffective leadership, and work-life balance issues.
It’s hard to know in advance just which factors may present themselves in diagnostic analytics, the company may uncover various factors contributing to the high turnover rate all of which could provide different and valuable insights into the dynamics of employee attrition. It’s also entirely possible to run multiple analyses and finding that none of them seem to indicate data informed insights into the issue. It’s important to go into the process of diagnostic analytics with hopeful, but realistic expectations. It’s like mining for gold, you will likely work very hard and do a lot of analyses that turn out to be nothing more than dust and dirt, but it’s all in the grand pursuit of a few shiny nuggets of pure gold. When data can help diagnose a problem, you are onto something truly transformational. Because strong diagnostic analyses can enable an organization to implement targeted interventions and address underlying issues in targeted ways, they are key to identifying data-informed solutions to issues and opportunities.
Beware Analysis Paralysis!
It’s important to navigate the diagnostic analytics process strategically. While the urge to analyze every possible variable might be tempting, it can lead to analysis paralysis – a state of overthinking that hinders progress. It can also lead to situations where a project is never completed, or the amount of effort and time spent was too much in comparison to the benefit of the eventual findings. This can be particularly challenging because people analytics professionals tend to be naturally curious and they want to assess ALL the things. So, how do we fight analysis paralysis? Here’s a key tip: First, remember that people analytics is about supporting decision making aligned to the needs of the workforce, then prioritize only those analyses that can lead to an actionable decision.
When there are too many possible variables to assess – don’t try to assess them all. Focus on the ones that can align to targeted interventions or that can be modified to effect a positive result. Some findings just aren’t actionable. So if you are limited on time, spend it analyzing things that can be acted upon. For example, let’s say your organization is having a hard finding qualified candidates to hire and they’ve asked you to use your data skills to diagnose what is going on. There are literally hundreds of different things you could analyze on this topic, but not all are actionable. It is likely that current low unemployment rates could be a main reason why it is difficult to hire new employees, but it is unlikely that you will be able to change much about the broader labor economy. So, while this finding may be factual, important, and could even be the main cause, it’s not one that the organization would be able to do much about. I would recommend instead focusing your analyses on the variables and relationships that you can address. For example, maybe your diagnoses indicate that your organization is paying salaries that are only 80% of what the competitor is paying people for the same positions. So, the few people who are looking for jobs are choosing to take the higher paid positions elsewhere. Salaries may be something the organization could choose to take action on and with your information, leaders in the organization could now consider this potential as a potential solution and area for decision making. When you keep in mind that people analytics is all about using data to inform decisions and you focus your diagnoses on topics that are actionable, you can reduce some of the analysis paralysis that can occur.
Diagnostic Analytics Skills
While most diagnostic analyses can be accomplished using the same analytics techniques described in the last section, it is important to expand your statistics knowledge a bit more if you want to accurately use data to indicate that the relationships you are finding are statistically significant and truly indicate a pattern, not just resulting from error or chance. To do this, you’ll need to add to your analytics skills by also learning about:
- Statistical significance testing of your analyses (e.g., is your correlation statistically significant?) to see if the observed relationships are statistically significant or simply due to chance.
- Statistical tests to compare groups (e.g., T-tests, Chi-square, ANOVA)
- Statistical analyses to assess linear relationships when you assume one thing leads to or happens before the other (e.g., regression)
- Research methods & hypothesis testing. If you are implying that one thing caused another thing to happen, you’ll want to ensure you have a method for applying research methods and hypothesis testing (more on this in the research methods section)
- Measurements of reliability & validity of the data being used (more on this in the data management and research methods section)
Common techniques for diagnostic analytics are:
- Exploratory Data Analysis: Start by collecting or creating relevant data from various sources and then use all those wonderful descriptive analytics skills plus your descriptive ones. Don’t limit yourself to only answering pre-determined questions, explore and play in the data looking for patterns, groupings, or differences.
- Drill-Down Analysis: Drill-down analysis involves delving into data at a granular level to extract detailed insights concealed within aggregated results. For instance, let’s say an organization wants to understand performance reviews more comprehensively. To execute a drill-down analysis, they would start with aggregated performance review ratings and segment the data by job roles, departments, locations, or any other groupings available in the data and then analyze patterns within or between all possible segments of interest looking for interesting differences or findings.
- Root Cause Analysis: Root cause analysis is a methodical approach used to uncover reasons behind specific outcomes. For instance, imagine an HR team notices a concerning drop in employee engagement scores. To conduct a root cause analysis, the team would gather data related to all possible things that they image might be at the ‘root cause’ of the issue and then systematically assess each one to see whether it seems to be related to employee engagement, helping to rule out some possibilities and pinpoint the issue(s) that may be responsible for the drop with an end goal of developing a targeted solution.