4 Misconceptions about (HR) Analytics
Did you know that most (HR) predictions are inaccurate? And do you know why you should never guarantee anonymity when you are dealing with HR analytics?
There are many misconceptions about HR analytics. In this blog I will address four of the most common ones. The first two are misconceptions about analytics in general and the last two are more specific to HR analytics.
1. Predictions are inaccurate
Human behavior is complex. Predictive analytics will never be able to accurately predict human behavior entirely. An algorithm that predicts performance with an accuracy of 30 to 40% is already very accurate. To put this into perspective: humans are much worse at predicting performance.
When assessing who your high and low performers are and who you should promote, HR should strive for accurate data. Especially when you use data to make important decisions about people’s lives.
HR data is dirty. It is often dispersed in multiple systems. It is not uncommon for people with the same function to have a wrong label. Most time analyses do not include all the relevant variables. The result is that we rely too much on measurable data and this might reflect badly on some people.
Psychological assessments require different levels of reliability for different instruments. Instruments that act as a tool for diagnosis of psychological disorders have to have excellent reliability. This is a requirement because, as you can imagine, the impact of such an assessment can be significant. Instruments with a much lower impact (e.g. a training inventory questionnaire) are allowed to be less reliable.
Related (free) resource ahead! Continue reading below ↓
51 HR Metrics cheat sheet
Data-driven HR starts by implementing relevant HR metrics. Download the FREE cheat sheet with 51 HR Metrics
A similar approach to data and predictive analytics would suit HR. It should assess their data in a comparable manner. It is important to always realize that data is imperfect and HR should actively question the reliability of their data and predictions.
2. Correlation is not causation
Correlation and causation are confusing the heck out of people.
The graph below shows that the cheese consumption per capital almost perfectly matches the number of people who died by becoming entangled in their bedsheets. Obviously, this is pure coincidence. These numbers do not mean that the one is caused by the other.
Made by Tyler Vigen (Source: http://www.tylervigen.com/spurious-correlations)
The West Wing is one of my favorite TV series. It is a political drama tv series set in the West Wing of the Oval Office. Martin Sheen (who plays U.S. President Josiah Bartlett) explains in this clip the difference between correlation and causation to his staff. What he describes is the so-called ‘post hoc ergo propter hoc’ fallacy.
It is important to keep in mind that most correlations are purely coincidental. This is also the case for people data. Oftentimes there is an (omitted) third factor that explains a correlation between two variables in people data. For example, the number of sandcastles build and the number of sold ice creams don’t have much in common, but they both have something to do with the weather.
Causation can only be found through longitudinal research. This is research that takes multiple measures over time. Only by proving a relationship between what happens now and what happened half a year ago can HR claim some proof of causation.
Additionally, good analytics research starts with a research question. This question is based on the issues that the company is dealing with and on literature. A systematic approach like this will result in the best and strongest statistical (predictive) models.
3. Guarantee confidentiality instead of anonymity
Privacy concerns everyone nowadays, and rightly so. The Snowden files and a recent EU-U.S. Data Privacy Shield to protect transatlantic data flows are two recent examples.
The outcry for privacy protection is also heard in the workplace. The line between personal and work-related data is becoming increasingly blurry. People use their personal email and web browser accounts at work, and they use their work accounts and laptops at home.
People data encompasses an increasing amount of (personal) data. Bloomberg captures every single keystroke of their employees, including the times they logged in. Humanyze’s social sensing platform uses sensors that capture a tone of voice, movement and even posture to analyze social interactions.
Other examples include email data, agenda appointments, contact lists for social network analyses, the times logged into systems and how people behave on social (work) networks.
Some countries regulate what data can be captured and where it can be stored. For example, European Union data protection laws dictate that employee data from EU-based companies are not allowed to leave the EU.
HR’s duty is to protect its employees. It has to find a balance between shielding employees from privacy violations on the one hand, and using people data to optimize work practices on the other hand. The argument that workplace data is only ‘used for good’ is hard to defend. People data is used for the company’s benefit. This is often connected to the employee’s benefit, but not always.
The best HR can do is to guarantee confidentiality. It cannot guarantee anonymity. This means that people data increasingly becomes an area of the Legal and Compliancy Department. HR should encourage this in order to create checks and balances within the company and to guarantee fair use of employee data.
Employees should always have an opportunity to opt-out of (some) tracking data. Also, legislation needs to deal with the growth in the storage and analysis of personal data in order to protect employees.
4. HR analytics is all about the ‘human’ factor
One of the most commonly heard concerns is that people do not feel comfortable with people analytics. Their main fear is that HR is losing the ‘human’ factor.
Even many HR business partners and managers do not like a data-approach to people. The general feeling is that you should see people as people. You should not reduce them to a single number. The fear is that analytics stifles innovation and reduces diversity: analytics forgets this ‘human’ factor.
According to Patrick Coolen (2016), you are forgetting the human factor when you do not apply analytics. Even though we are unaware of it most of the time, everyone is biased. These biases often results in bad decisions. Analytics help us escape our biases and imperfect decision-making. Daniel Kahneman, the only psychologist to ever receive a Noble price, explains this perfectly in his book Thinking Slow and Fast. Imagine a single dice with four green and two red sides. The dice is thrown twenty times: which series of colors is most likely to be rolled?
As mentioned before, the dice has four green and two red sides. Because the first option only has one green, it does not seem representative. This is why most people would choose the second option. However, the first option has four reds and only one green, while option 2 has four reds and two greens. Therefore, option 1 is more likely to happen.
Analytics is about running the numbers and checking your assumptions. It shows time and time again that the counter-intuitive answers are sometimes the right ones. This is something which analytics helps us see and realize, and why analytics has the potential to helps us be fairer to everyone.
Do you want to learn more about HR analytics? Check our review of 5 online HR analytics courses here.