Employee attrition analysis and retention plan from the IBM HR data set

Corvane Pharma Services runs sales, research support and HR for a mid-size drug maker in the US. It has 1,470 employees. Last year about 16% left, and in the sales function the rate was higher. The CHRO estimates each departure costs between a third and a full year of salary once recruiting, onboarding and lost sales are counted, but nobody has added it up.

The CHRO has asked you, as an HR analytics intern, to answer three questions in six weeks: - Who is leaving, and what seems to drive it? - What does it cost the company each year? - What should HR do about it, how much will that cost, and how will we know it worked?

You will use the IBM HR Analytics Employee Attrition and Performance data set from Kaggle. It has 1,470 rows and about 35 columns, including age, department, job role, monthly income, overtime, job satisfaction, years at company, years since last promotion and attrition. It was created by IBM data scientists as a fictional data set, so say so in your report, and treat it as Corvane's HR extract for this project.

You will audit the data, explore attrition patterns, build an interpretable model, put a dollar figure on attrition by segment, design a retention plan with interventions and their expected effect, and check the plan for fairness and privacy.

Done means a CHRO can read one page and know which three groups to focus on, what it costs to keep doing nothing, and what the proposed plan will cost and save. The analysis must never be used to target or penalise individual employees.