Back to All Projects

📊 HR Analytics: Employee Attrition Intelligence

A 4‑page prescriptive retention framework built with SQL, Python, Power BI & DAX

1,470

Employees Analyzed

16.1%

Overall Attrition

10.4%

High‑Risk Flagged

$1.2M+

Annual Savings*

Launch Live Dashboard Explore GitHub Repo Download EDA Report (PDF)

📸 Dashboard Gallery (Click any image to zoom)

Page 1: Executive Overview – Workforce Snapshot

Page 2: Drivers – Root Cause Analysis

Page 3: Analysis – Risk Prediction & AI Key Influencers

Page 4: Strategy – Prioritized Recommendations & ROI

The Insight That Changed Everything

Employees who rated their work‑life balance as "Good" had the highest attrition count (127) — not "Poor" or "Average."

This revealed that attrition isn't always dissatisfaction. Sometimes it's hidden structural pressure — workload creep, stalled careers, unmet expectations — masked by a score that looks fine on paper.

🔍 Key Data‑Driven Insights

Overtime = 3× higher attrition – 31% vs 9% baseline
41% of exits happen in first 2 years – an onboarding failure
Sales Representatives: 39.8% attrition – highest of any role
Job Level 1 drives 60% of attrition – entry‑level most vulnerable
Promotion gap >4 years doubles risk – median gap 3 years for leavers vs 2 for stayers
AI Key Influencers: No overtime → 1.29× retention; Tenure >2y → 1.25×; Income >$13,758 → 1.17×

⚙️ Code & Technical Implementation

-- Tenure cohort creation
SELECT 
    EmployeeID,
    CASE 
        WHEN YearsAtCompany <= 2 THEN '0-2'
        WHEN YearsAtCompany <= 6 THEN '3-6'
        WHEN YearsAtCompany <= 10 THEN '7-10'
        WHEN YearsAtCompany <= 15 THEN '10-15'
        ELSE '15+'
    END AS TenureGroup
FROM Employees;

-- Aggregated view for dashboards
SELECT 
    Department,
    COUNT(*) AS TotalEmployees,
    SUM(CASE WHEN Attrition = 'Yes' THEN 1 ELSE 0 END) AS AttritionCount,
    AVG(MonthlyIncome) AS AvgMonthlyIncome
FROM EmployeeMaster
GROUP BY Department;

# Overtime vs Satisfaction heatmap – critical insight
pivot = pd.pivot_table(df, values='Attrition_Flag',
                       index='JobSatisfaction',
                       columns='OverTime')
sns.heatmap(pivot, annot=True, cmap='Reds')
plt.title("High Risk: Overtime vs Low Satisfaction")

# Correlation analysis
corr = df[['AttritionFlag', 'OverTime', 'JobSatisfaction', 
           'YearsAtCompany', 'YearsSinceLastPromotion']].corr()
sns.heatmap(corr, annot=True, cmap='coolwarm')

// Risk Score – composite index for proactive flagging
Risk Score = 
VAR OvertimeRisk = IF(Employees[OverTime] = "Yes", 0.3, 0)
VAR PromotionRisk = IF(Employees[YearsSinceLastPromotion] > 4, 0.25, 0)
VAR TenureRisk = IF(Employees[YearsAtCompany] <= 2, 0.2, 0)
VAR SatisfactionRisk = IF(Employees[JobSatisfaction] <= 2, 0.15, 0)
RETURN
    OvertimeRisk + PromotionRisk + TenureRisk + SatisfactionRisk

// Retention Multiplier – quantifies impact of a factor
Retention Multiplier = 
DIVIDE(
    CALCULATE([Retention Rate], Employees[OverTime] = "No"),
    [Retention Rate]
)

🎯 Strategic Recommendations

Audit overtime‑heavy roles – reduce 31% attrition by balancing workload or adding comp
Redesign onboarding for first 2 years – implement 30‑60‑90 day checkpoints and mentorship
Transparent promotion cycles – flag employees at year 3 to avoid >4 year gaps

Projected impact: 20–25% reduction in high‑risk attrition → $1M+ annual savings (based on 1.5× salary replacement cost).

📚 Full Project Resources