Employee Attrition

Task Overview

• Research question
• Executive summary
• Methodology + sample
• Key findings: insights and trends on attrition
• Predictive analytics: a Simple Attrition Model (SAM)
• Questions and discussion

Research question

Research question: What can we learn about employee attrition at Multiplex in Q2 2017?

Why are employees leaving?

Who is leaving, what are their characteristics?

When, and after how long are employees leaving?

Executive summary

1. Problem: The overall trend in attrition is that talent is leaving the company voluntarily. If this is a long-term trend, it is worrying for the overall competitiveness and thus health of the company.

   Insight 1: Employees generally resign voluntarily.

2. Trends and insights: Talent is generally on the upper end of performance and potential. The job profile (job title, department, location) and timing (season, employment duration) play a role in attrition.

   Insight 2: A relatively high proportion of talent is leaving.

   Insight 3: The job profile (job family and title) plays a role in attrition, location plays a role as well.

   Insight 4: Timing matters when employees leave.

   Insight 5: There is a strong relationship between job family and employment duration.

3. Predictive analytics: We can predict which employees are likely to leave by building a Simple Attrition Model (SAM) and applying it to all employees.

   We can make this task more feasible by prioritising the most important employees.

   I suggest an easy measure to identify the most important employees.

4. Recommendations: HR can optimise attrition by estimating all employees chance to leave by approaching talent proactively. We can predict which employees are likely to leave. A custom-built attrition model helps HR to employees at risk of attrition and prioritise these according to their performance, potential, profile, and timing.

Methodology + dataset

Preparation

1. Consolidate data into single spread sheet in Excel.
   
2. Clean the data (data types, typos, formatting, etc.).
   
3. Create a reference table for the new value variable employees.
   
4. Calculate additional categorical and quantitative variables (e.g. unique identifier~ index, employment length, value scorecards).

Analysis

1. The calculations of tables and and pivot charts and the early analysis are conducted in MS Excel.
   
2. The visualisation is conducted in Excel and R.
   
3. The presentation is in the reproducible R markdown format (PPT, PDF and HTML).

Dataset

Who?
Terminated employees at Multiplex.

How many?
37 employees (N=37).

When?
April - June 2017, Q2.

Where?
Multiple locations across the UK.

Which profiles?
Various roles and job families.

1) Talent is leaving the company voluntarily.

1. The overall trend in attrition is that talent is leaving the company voluntarily. If this is a long-term trend, it is worrying for the overall competitiveness and thus health of the company.

Why are employees leaving Multiplex?

Insight 1: Employees generally resign voluntarily.

• 1 a: approximately 9 out of 10 resigned (i.e. left voluntarily).

1) Talent is leaving the company voluntarily (continued).

Why does this matter for the business?

Insight 2: A relatively high proportion of talent is leaving.

• 2 a: High performing talent is leaving the company.
  19% were high performers, 46% were moderate performers, and 33% were low performers.

• 2 b: High potential talent is leaving the company.
  Employees with high and moderate potential are leaving the company. 27% had high potentials, 55% had medium potential, and 16 % had low potentials.

2) Trends and insights: job profile and timing play a role in terminations.

Who is leaving? What is their profile?

Insight 3: The job profile (job family and title) plays a role in attrition, location plays a role as well.

• 3 a: Attrition varies highly by job title, high diversity. 19 Job titles in total.

• 3 b: There is a strong relationship between job families and attrition. 9 job families in total, but strong dominance in some job families. Few groups that matter the most. 17 from Operations, nearly half, 6 from Commercial, second biggest group, and 6 from Safety Health Environment.

• 3 c: There is a relationship between Job sites and attrition.
  In total, there are 16 construction sites and offices. Top 5 locations E, F, H, A G account for more than half (20) of attrition.

2) Trends and insights: job profile and timing play a role in terminations (continued).

When, and after how long are employees leaving?

Insight 4: Timing matters when employees leave.

• 4 a: Terminations peak in early in Q2 and then decrease from April to June.

• 4 b: Employment duration varies heavily. There are two peaks moments of one and two years.

2) Trends and insights: Relationships between variables

Is there a relationship between job family and employment duration?

Insight 5: There is a strong relationship between job family and employment duration.

• 5: Job families differ extremely in their average employment duration.
  The average employee stays nearly 1 year and 9 months with the company (644 days).
  Longest duration: Operations Members stay 2 years and 1 month (771 days).
  Shortest duration: Commercial Members stay on average 1 year and nearly 1 month (386 days).

3) Predictive analytics: We can predict which employees are likely to leave by building a Simple Attrition Model (SAM) and applying it to all employees.

Sub-conclusion: We have learned which factors play a role in attrition. We can predict which employees are likely to leave in the future.

The Simple Attrition Model (SAM) has the following input:

• Performance
  
• Potential
  
• job family
  
• Employment duration

The output:
Termination score [1 to 100]

80 - 100 Very likely
60 - 80 likely
40 - 60 indifferent
20 - 40 unlikely
0 - 20 very unlikely

Problem: Approaching all employees is not practical, as the total population is too big with 3-4 thousand.

Question: How do we identify the most important employees?

Given measures address particular aspects on an employee, but they do not evaluate and summarise the total value or importance.

Solution: We look at the top 1/3 of people first.

How? We create a single measure that is a combination of both performance and potential. It ranks employees into 9 levels, from 9 (‘star’) to 5 (‘normal’) to 1 (‘avoid’).

Concept schema

• Insight: Higher value talent is leaving.

17% were in the high value star cluster,
47% were in the normal value core cluster 47%, and
36% were in the low value avoid cluster.

Recommendation: The Simple Attrition Model helps HR to identify employees at risk of leaving.

How to optimise attrition

• Increase or incentivise attrition in the ‘avoid’ 1 to 3 range,
• spend some efforts on managing the core ‘normal’ range 4 to 6, and
• proactively reduce attrition in the high value range 7 to 9.

Identify whom to contact with the Simple Attrition model (SAM)

When to approach depends on job family

Questions and discussion

Why are employees resigning? Where do they go?

Why are employees leaving before the summer? Are these trends also valid for the rest of the year?

What are the reasons for the long employment duration in Operations? And the short employment duration in Commercial?