Uncovering the Drivers Behind Employee Attrition
A portfolio analytics project using a public employee dataset to investigate attrition patterns across 1,470 records and demonstrate Tableau, exploratory analysis, segmentation, dashboard design, and workforce-planning recommendations.
Interactive dashboard. Use filters to explore the data.
Employee attrition is one of the most expensive problems a workforce-planning team can face, and one of the hardest to see coming, because the warning signs stay buried until headcount has already dropped. This project takes a public dataset of 1,470 employee records and asks the question such a team would need answered: not just how many people are leaving, but who, and why.
Read as flat exports, the records surface only a headline: overall attrition of 16.1%. Beneath that number lies wide variance across departments, age groups, and job roles. R&D alone accounts for 56.12% of all attrition, a concentration that stays invisible until someone segments the data.
The core challenge here isn't a lack of data; it's a lack of structure. Employee records, satisfaction ratings, compensation, and demographic fields all arrive in one flat file with no unified analytical view, the state most workforce data is in before anyone can act on it.
The data existed. The visibility didn't.
Analytical Gaps in the Raw Data
No ready-made attrition view
The raw dataset is a single flat table: attrition status sits beside 34 other fields with no roll-up or segmentation, so the drivers stay buried until they are modelled
No early-warning signal
Attrition is recorded only after the fact. The dataset flags who has already left, with no leading or predictive signal until one is engineered from the satisfaction and tenure fields
Hidden departmental disparities
R&D's 56% share of total attrition is invisible in the headline rate. It only surfaces once the data is broken out by department
Satisfaction fields left uncorrelated
The dataset carries job and environment satisfaction fields, but nothing links them to attrition outcomes until the relationship is modelled
Demographic blind spots
Age, education, and tenure patterns sit in the fields but stay invisible until the data is segmented along them
16.1%
overall attrition rate across the dataset's workforce
I started with a dataset of 1,470 employee records spanning demographics, job roles, compensation, satisfaction ratings, and attrition status. The first step was exploratory data analysis: profiling distributions, identifying nulls, and understanding the shape of the workforce.
Three patterns emerged quickly. First, the 25 to 34 age group accounted for a disproportionate share of exits. Early career employees were leaving faster than any other cohort. Second, employees who rated their job satisfaction as 1 or 2 (out of 4) had attrition rates nearly triple those who rated 3 or 4. Third, the R&D department wasn't just the largest department; it was haemorrhaging talent at a rate that dwarfed Sales and HR combined.
These weren't just correlations. When I cross-referenced satisfaction with department and tenure, a clear narrative formed: dissatisfied early-tenure R&D employees were the highest-risk segment. This was the insight that would shape the entire dashboard design.
Discovery Process
Data Profiling
Audited 1,470 records across 35 fields, identified key dimensions for attrition segmentation and cleaned inconsistencies
Cohort Analysis
Segmented attrition by age group, department, education field, and job role to isolate high-risk populations
Satisfaction Correlation
Mapped job and environment satisfaction ratings against attrition outcomes, revealing a clear inverse relationship
Cross-Dimensional Analysis
Combined department, tenure, and satisfaction to identify the highest-risk employee segment: early-tenure R&D staff
Key Insight
Dissatisfied early tenure R&D employees were the single highest risk attrition segment, a finding invisible in the aggregate figures.
The goal was not to build a reporting dashboard. Reporting tells you what happened. I wanted to build a decision support tool, one that helps HR leaders ask better questions and act on the answers.
I structured the design around three principles: the questions a workforce-planning team would bring to this data, not the ones a static report answers.
Design Principles
Segment Before You Summarise
Aggregate metrics hide the story. The dashboard leads with department-level and cohort-level breakdowns so users see where attrition is concentrated before looking at totals.
Correlation Over Causation, But Make It Actionable
Satisfaction and attrition are correlated, not causal. The dashboard frames this correlation in terms of actionable segments: which groups are both dissatisfied and leaving, so HR can intervene where it matters.
Self-Service First
Every chart is filter-linked. Clicking a department, age group, or education field updates the entire view, so an HR business partner could explore any segment of the workforce without waiting on an analyst.
I designed and built the dashboard in Tableau, choosing it for its interactive filtering capabilities and Tableau Public publishing for open, no-login sharing. The layout was structured to guide the user's eye from high-level KPIs down to segmented breakdowns.
Every visual maps to a specific question a workforce-planning team would bring to this data. No chart exists without a corresponding decision it's meant to support.
Dashboard Components
KPI Summary Bar
Total employees (1,470), attrition count (237), attrition rate (16.1%), and average age (37). The four numbers a workforce-planning team would check first
Department Attrition Breakdown
Donut chart showing R&D (56.12%), Sales (38.82%), and HR (5.06%), immediately surfacing where attrition is concentrated
Age Group Distribution
Bar chart segmented by age bands revealing the 25 to 34 cohort as the highest attrition group across all departments
Satisfaction vs. Attrition Matrix
Heat map correlating job satisfaction levels (1 to 4) with attrition rates. Satisfaction level 1 shows nearly 3x the exit rate
Education Field Analysis
Stacked bar showing attrition distribution by education background. Life Sciences and Medical degrees lead exits
Interactive Cross-Filters
Click any segment to filter the entire dashboard. Department, age group, gender, and education are all interlinked
The dashboard reframes how attrition could be approached: instead of periodic reports that surface problems too late, it offers a live view of exactly where retention efforts would matter most.
The R&D finding alone would justify a closer look at budget allocation: targeted engagement and mentorship for early career employees, and more frequent satisfaction check-ins for the high risk segments the dashboard surfaces. The analysis recommends these as interventions to consider, not as changes already made.
Impact Metrics
1,470
Employee records analysed across 35 data fields
56.12%
R&D's share of total attrition, surfaced as the critical intervention point
3×
Higher attrition rate for employees with satisfaction level 1 vs. level 4
25 to 34 Age Band
Identified as highest-risk cohort for early attrition across all departments
A 16.1% overall attrition rate sounds manageable. But when R&D carries 56% of exits and your highest value talent pipeline sits in the 25 to 34 age band, the aggregate is hiding a crisis. Always segment before you summarise.
The best dashboards don't provide answers. They help users ask better questions. Interactive filters and cross dimensional views are built to turn passive report consumers into active analysts.
By the time someone leaves, the decision was made months ago. Satisfaction data, when properly visualised and correlated with outcomes, becomes the closest thing HR has to a predictive signal.