I feel that most companies start their People Analytics initiatives with something related to employee attrition (as know as in some companies employee turnover). It is not surprising since this study from 2017 reports that it costs employers 33% of a worker’s annual salary to hire a replacement if that worker leaves.
Employee attrition is not just a problem for the human resources department. When people are leaving companies, productivity might decrease, customer support can suffer, and quality can be in danger. Finance departments also have a hard time forecasting staff cost when attrition is high or unpredictable.
I send an email every Thursday (or so), once I publish my weekly article. I also share some good content. Mostly HR related.
Both in my HR reporting jobs and People Analytics consulting projects, attrition was the leitmotif in most of them. I am convinced that there no general solution to this problem. It depends on many variables, from social culture to the family situation of an individual. There are different attrition rates between Berlin and Munchen, and while in Germany, people tend to have a higher tenure than in Romania overall, there are different motivations that keep employees engaged in EMEA versus JAPAC or Americas.
Below is a short case study on attrition analytics. Some information is hypothetical since the client I worked with does not want to reveal it’s identity, and the data presented might be a little scrambled. I focus more on business and data engineering aspects since the actual analysis was performed in Excel. Do not expect any fancy Business Intelligence visuals, linear regression in R, or some decision tree algorithm in Python.
- An automotive company with over 100000 employees worldwide, spread across all continents, with back-office and support shared services, highly centralized operations;
- Their financial shared service center had offices in 3 countries, supporting the entire organization in a follow-the-sun model for transactional tasks and teams of financial analysts supporting other functions on-demand or project-based;
- Headcount of around 400 employees in Romania (supporting EMEA), India (supporting APAC), and the United States (supporting Americas).
In Romania, their attrition rate was almost double than in India or the US. In 10 years since it’s inception, the team almost changed the entire team members twice.
Romania is also a highly competitive environment for talent, especially in big cities where various multinational companies opened or moved shared services centers. From a social and cultural perspective, is very common for Romanian employees to leave a job for 15-20% salary increase in other company. At least compared to other European countries and the US.
As I do in any people analytics project, I tried to identify patterns. For this particular situation, I chose 3 categories to investigate:
- Tenure patterns – if any specific group is more likely to leave the company;
- Performance patterns – based on the annual appraisal reviews;
- Periodical patterns – if there is/are any particular periods of the year when people are leaving the company.
While tenure and performance patterns I could not identify since the distribution was pretty even, a periodical pattern has been easy to identify: 60% of the attrition was happening in 8 consecutive weeks of the year. Can you guess which weeks? The first of the fiscal year. Those who worked or know people working in the finance departments know how stressful the end of a fiscal year can be. There is pressure from everywhere and everyone to provide all financial statements, to make the press releases to shareholders and public, to calculate the dividends and so on. Onerous work done in very little time.
Moving forward, one other metric that was relevant in the attrition population was the overtime: those who are putting the more work, are the ones most probably to leave.
At this point, I could spot some patterns, but there was not enough to have a conclusion. I decided to search for data not only in the HR systems, so I got a report from the ticketing system to see I can get any insights that might conclude my ideas. As you may guess, the analysis showed a strong correlation between the employees with the highest number of tickets and the ones that left the company.
Another critical factor that I should mention is that this department was split into teams that usually included 2 managers, 10 junior analysts, and 10 senior analysts. The average period for someone to promote from junior to senior was 2.5 years.
Below are some of the actions the company took to reduce attrition. Each company may have a different list of actions, some feasible to others, some not. Solutions to this problem are something beyond just the data, and it can interfere with company policies and procedures.
Some of the actions the company implemented to reduce attrition:
- Moved the performance salary reviews in Q4 of the fiscal year, instead of Q2. The purpose of this measure was to increase the motivation of the employees;
- Created a clear path and a list of checkpoints and milestones for promotions from junior to senior;
- Decreased the time of promoting someone from 2.5 years to 2 years;
- Created a new role inside the small teams, Workload Manager, someone who was in charge of distributing equally the tickets and tasks to reduce the possibility of high load.
Ideally, one year after implementing these actions, it would be great to rerun the analysis to see how things improved. Unfortunately, my friend and I moved to other roles, and time was not our friend, but from what I heard from her, things improved significantly, including the results of their engagement pulse study inside the department was way better than the years before.
- The role of people analytics is to identify the causes of a problem;
- When doing people analytics on a small population, you don’t need fancy models or programming skills – I wrote more on this topic in another article: Let’s not overcomplicate People Analytics;
- People analytics is just a tool in the toolbox, is not the answer to all the problems of an organization;
- It’s always better to make decisions based on insights, even if the insights confirm your gut feeling;
- Look beyond the HR Data from HRIS -> data stored in ticketing software or other productivity tools might offer more than HR Data can.
If you want to receive more case studies like this, make sure you subscribe to my newsletter below: