There’s a right way and a wrong way to do everything and people analytics is no exception.
Sometimes the right way isn’t obvious and without proper guidance, you can easily veer into the erroneous territory. To ensure ROI and avoid costly mistakes, people analytics should be used wisely.
When done well it has the potential to provide valuable insights that, when acted on, can improve employee performance.
Avoid these common pitfalls and you will be well on your way to ensuring a favorable outcome and reducing the margin for error.
1. Don’t wing it. Be intentional.
Don’t start out trying to solve everything at once. Focus! Your strategy should be to start with one specific business problem.
You need to know ahead of time what central questions you want to be answered, keeping your specific business goals in mind. Think about your pain points. They are a good starting point in helping to formulate those questions.
Once those have been established you can determine the specific analyses to employ that are best suited to the inquiry. By taking a systemic approach you can avoid wasting time and money going down rabbit holes that do not ultimately benefit your business.
2. Don’t be inflexible.
Your process should always be iterative and agile. Always be ready to revise the strategy or methodology if the data output isn’t accurate or helpful. Don’t be so invested in a particular approach that you are unwilling or unable to adjust if it isn’t working or no longer suits your purpose. Remember it is better to have no insights than inaccurate insights.
3. Don’t confuse correlation with causation.
Just because you observe a significant relationship between two variables does not automatically mean it’s causal. A change in one may not necessarily cause the change in the other. For example, in investigating reasons for low employee engagement, analysis of the data may reveal that the lowest earners have the lowest engagement. You could say that there is a correlation between low engagement and low salary. You might be tempted to assume that low pay causes low engagement and paying them more might seem like an obvious solution.
However, this may be a premature guess. Further examination might reveal that another factor—job complexity—is the culprit, and they happen to be the lowest earners because job role complexity is linked to pay scale.
Avoid the low-hanging fruit. They can be deceptive. Staying objective will prevent you from assuming causation automatically because it aligns with your preconceptions or your hunches. Make sure you consider other possible factors that may explain the phenomenon and don’t be quick to rule anything out.
4. Don’t generalize.
A relationship between two variables in one area of the business will not necessarily hold true for another area. For example, a clothing retailer experienced an increase in revenue when employee engagement rose. However, one retail branch did not see an increase in revenue. Further digging unearthed the fact that the correlation/causation didn’t apply to this store because a competitor had opened a store on the same street.
As a result, it is clear that different measures would have to be employed to increase revenue at that store given the different circumstances.
5. Don’t ignore outliers.
In the example above, the store with new competition was the outlier. An outlier is a statistical anomaly or atypical specimen that statisticians sometimes ignore as anomalies may skew the results when trying to view the big picture.
In the case of people analytics, you should do the opposite. The presence of outliers may be an indication that you should take a closer look at the data since they can be caused by three things: inaccurate data, another variable that only affects certain data sets, or because there is a correlated relationship between two variables and not a causation relationship. In that case, it was the second reason.
6. Don’t neglect to clean your data.
Data that is merged from different sources and systems need to be “cleaned.” This entails several things:
- Make sure the data is up-to-date so it is always relevant.
- Make sure it is complete and there are no records missing.
- Removing duplications by ensuring that there are unique identifiers.
- Removing inconsistency errors caused by things like missing data, non-matching records, or different labels for the same item.
Dirty data is unreliable and invalid. The validity of your data will determine if you are in fact measuring what you want to measure. If we use the above retail store example, would an increase in revenue indicate that the sales associates are better at their job than another store or just that they have less competition?
Reliability is based on your ability to get consistent, stable results without bias. For example, the criteria used for each sales associate’s customer service performance rating shouldn’t be subject to each rater’s interpretation.
7. Don’t ignore or downplay privacy considerations.
Data gathered for people analytics can be of a sensitive nature. It may include private and personal information about your workforce which should be handled with care and discretion. As such this sensitive data should not be tracked or reported unless there is a solid business reason for it and access should be restricted to certain authorized personnel only. Which people can see specific data should be determined through careful selection.
GDPR compliance is also of particular importance since the gathering of employee data can come in direct conflict with privacy regulations. Employee vs. employer interests must be balanced in such a way that the principles of GDPR are upheld.
8. Don’t stop at reporting.
Gathering and analyzing the data is just the first part of the process of people analytics.
A report is useless if it doesn’t tell a story whose natural progression is to prompt an action. Data that isn’t transformed into actionable insight is the same as no data at all.9. Don’t blindly trust the data.
Always question the data. By asking questions you can differentiate between useful analyses and those that might be flawed. For example, questions about how the data is generated and what might be missing may help you to fully understand the strengths and weaknesses of the analyses. Healthy skepticism can identify potential blind spots and biases and help leaders think beyond their own limiting frames of reference.
Your Roadmap to People Analytics
See how you can get a head start in helping your workforce grow.
Read the eBook.