QuantiriEvery statistic below was measured by code before any language model ran. Nothing here is generated by AI.
The order isn't arbitrary — findings are sequenced by how much they'd change a decision, not by p-value.
The narrative explains the computed evidence and is constrained to it. A number it didn't measure can't appear.
1,470 employees · 9 job roles · 3 departments
Attrition is a role problem: Sales Representatives leave at 39.8%, Research Directors at 2.5%
Every company tells itself that people leave for personal reasons, unpredictably, one resignation at a time. This dataset says otherwise: who quits is written into the org chart, with 33 of 83 Sales Representatives gone (39.8%) against 2 of 80 Research Directors (2.5%), and overtime nearly tripling the odds of a departure. Attrition here is not weather — it is geography, and the map has exactly three hot zones: low pay, low level, and long hours.
attrition × job role
The revolving door has a name tag — 33 of 83 Sales Representatives left (39.8%), 15.9 times the Research Director rate.
outcomeincome × job role
Your job title is your paycheck — role alone explains 81.6% of income variance, a $14,556 monthly gap between Manager and Sales Representative.
categoricalattrition × overtime
Overtime is the quiet exit sign — 127 of 416 overtime employees left (30.5%) versus 110 of 1,054 without it (10.4%).
outcomeExecutive summary
People don't leave this company at random — new, junior, underpaid employees working overtime do
Five independent cluster analyses corroborate one another: attrition concentrates where pay, level, and tenure are lowest and hours are longest, peaking at 39.8% for Sales Representatives against a 16.1% baseline. Everything else in the org chart — pay, seniority, even department walls — moves in lockstep with the same hierarchy.
Key findings
attrition × job role
0.98monthly income × job role
1.00attrition × overtime
0.7710 signals analysed ↓
Four in ten Sales Representatives walk out; one in forty Research Directors does
Walk the sales floor and you are looking at this company's revolving door. 33 of 83 Sales Representatives left (39.8%) — nearly two and a half times the 16.1% company baseline — while only 2 of 80 Research Directors (2.5%) did the same. The gap between the two ends of the org chart spans 37.3 percentage points, and the chi-square test leaves no ambiguity (χ² = 86.2, p ≈ 0).
So what
A Sales Representative is 15.9 times more likely to quit than a Research Director (39.759 ÷ 2.5 = 15.9). Company-wide retention programs are aimed at the wrong target — the churn lives in three roles, with Laboratory Technicians (62 of 259, 23.9%) and Human Resources (12 of 52, 23.1%) right behind sales.
Risk multiple, worst vs safest role
15.9×
39.8%
Sales Representative attrition (33 of 83)
2.47× baseline
2.5%
Research Director attrition (2 of 80)
0.16× baseline
16.1%
Baseline across 1,470 employees
237 total leavers
HIGHEST
Sales Representative
39.8%
33 of 83 left · risk ratio 2.47
LOWEST
Research Director
2.5%
2 of 80 left · risk ratio 0.16
Attrition rate by job role
%barchart · JobRole × Attrition · OutcomeSegmentationEvidence
Attrition rate by job role
Sales Representatives leave at 39.8% — nearly 16 times the Research Director rate of 2.5%.
Where the leavers work
Attrition rate · %Sales Representative
39.8%
33 of 83 employees left
2.47× the 16.1% baseline
Laboratory Technician
23.9%
62 of 259 employees left
1.48× the 16.1% baseline
Manager
4.9%
5 of 102 employees left
0.30× the 16.1% baseline
Research Director
2.5%
2 of 80 employees left
0.16× the 16.1% baseline
Connected signals
Job title alone explains 81.6% of who earns what
Forget negotiation, tenure, or performance — in this company, your paycheck was decided the day your job title was. Managers average $17,182 a month while Sales Representatives average $2,626, a gap of $14,556 (17,181.68 − 2,626.00 = 14,555.68). ANOVA attributes 81.6% of all income variance to role alone (η² = 0.816, Cohen's d = 7.19, p ≈ 0).
So what
A Manager earns 6.5 times a Sales Representative's monthly income (gap ratio 6.54), and the role paid least is the same role quitting fastest — the pay map and the attrition map are the same drawing. Note the cliff: the top two roles both clear $16,000 while the third-highest, Healthcare Representative, drops to $7,529 — a $8,505 fall (16,033.55 − 7,528.76 = 8,504.79) between second and third place.
Pay multiple, Manager vs Sales Representative
6.54×
$17,182
Manager mean monthly income
n = 102
$2,626
Sales Representative mean
n = 83
$4,919
Company-wide median
mean $6,503 — right-skewed (skewness 1.37)
HIGHEST
Manager
$17,182/mo
median $17,455 · n = 102
LOWEST
Sales Representative
$2,626/mo
median $2,579 · n = 83
Mean monthly income by job role
$barchart · JobRole × MonthlyIncome · CategoricalSegmentationEvidence
Mean monthly income by job role
Two roles clear $16,000 a month; the other seven never break $7,600.
Where incomes fall across all 1,470 employees
25th percentile · $2,911
2,911
Median · $4,919
4,919
Mean · $6,503
6,503
75th percentile · $8,379
8,379
The mean sits $1,584 above the median because two senior roles stretch the ceiling to $19,999 while half the company earns under $4,919.
Overtime nearly triples the odds an employee walks out
Overtime looks free on a payroll report; the exit interviews say otherwise. Among employees logging overtime, 127 of 416 left (30.5%) — among those who did not, 110 of 1,054 left (10.4%). One yes/no field opens a 20.1-point gap (χ² = 87.6, Cramér's V = 0.244, p ≈ 0), the single strongest two-group split in the dataset.
So what
An overtime employee is 2.9 times more likely to leave (30.5288 ÷ 10.4364 = 2.93), and the 28% of the workforce on overtime (416 ÷ 1,470 = 28.3%) supplies 53.6% of all departures (127 ÷ 237 = 53.6%). Cut overtime for one high-churn role and you attack two risk factors with one policy.
Risk multiple with overtime
2.93×
30.5%
Attrition with overtime (127 of 416)
1.89× baseline
10.4%
Attrition without overtime (110 of 1,054)
0.65× baseline
28.3%
Workforce on overtime
416 of 1,470 employees
HIGHEST
OverTime: Yes
30.5%
127 of 416 left · risk ratio 1.89
LOWEST
OverTime: No
10.4%
110 of 1,054 left · risk ratio 0.65
barchart · OverTime × Attrition · OutcomeSegmentationEvidence
Attrition rate by overtime status
30.5% of overtime employees left versus 10.4% of everyone else — a 2.93× risk multiple.
2.93×
attrition risk for overtime employees
127 of 416 overtime employees left (30.5%) versus 110 of 1,054 without overtime (10.4%).
30.5288 ÷ 10.4364 = 2.93Per 100 overtime employees
0.305288 × 100
30.5 leavers
10.4 without overtime
Across all 416 overtime employees
416 × 0.305288
127 leavers
53.6% of the company's 237 total departures
The overtime attrition multiple
Connected signals
The seniority ladder is a 20.5-year climb, and it bends for no one
There are no shortcuts up this org chart — every rung is bought with years. Level 5 employees average 26.4 years of total working experience against 5.9 years at Level 1, a 20.5-year staircase (26.377 − 5.891 = 20.486) that ascends in strict order through every level between. Job level alone explains 63.5% of experience variance (η² = 0.635, Cohen's d = 4.35, p ≈ 0).
So what
Top-level employees carry 4.5 times the career mileage of entry level (26.377 ÷ 5.891 = 4.48) — seniority here is purchased with time, not talent-spotted early. The signal index shows income follows job level even more tightly (η² = 0.925), so the ladder sets both your title and your paycheck.
Experience multiple, Level 5 vs Level 1
4.48×
26.4 yrs
Level 5 mean experience
n = 69 · minimum is 21 years
5.9 yrs
Level 1 mean experience
n = 543 · the largest cohort
63.5%
Variance explained by level
η² = 0.635
HIGHEST
Job Level 5
26.4 yrs
n = 69 · min 21, max 40
LOWEST
Job Level 1
5.9 yrs
n = 543 · min 0, max 20
Mean total working years by job level
yrsbarchart · JobLevel × TotalWorkingYears · CategoricalSegmentationEvidence
Total working years by job level
Experience climbs monotonically with level — from 5.9 years at Level 1 to 26.4 at Level 5.
Career mileage across the whole company
25th percentile · 6 yrs
6
Median · 10 yrs
10
75th percentile · 15 yrs
15
Three quarters of the workforce has 15 or fewer years of experience, yet Level 4–5 jobs average 25+ — the top of the ladder is demographically scarce by construction.
Only two performance ratings exist, and one is worth 7.8 extra raise points
This company grades on a scale of two. Every one of 1,470 employees holds a rating of either 3 or 4 — and the 226 rated 4 (226 ÷ 1,470 = 15.4% of staff) received salary hikes averaging 21.8% against 14.0% for everyone else. The 7.8-point difference (21.850 − 14.003 = 7.846) is enormous relative to the noise: η² = 0.732, Cohen's d = 3.38.
So what
The rating scale is a binary gate, and crossing it raises your salary bump by more than half (7.846 ÷ 14.003 = 56.0%). With rating-4 hikes spanning only 20–25% and rating-3 hikes 11–19%, the ranges never even overlap — your rating fully determines your raise bracket.
Share of workforce rated 4
15.4%
21.8%
Mean hike, rating 4
range 20–25% · n = 226
14.0%
Mean hike, rating 3
range 11–19% · n = 1,244
84.6%
Employees rated 3
per signal index — ratings 1 and 2 are absent
HIGHEST
Performance rating 4
21.8% hike
n = 226 · min 20%, max 25%
LOWEST
Performance rating 3
14.0% hike
n = 1,244 · min 11%, max 19%
barchart · PerformanceRating × PercentSalaryHike · CategoricalSegmentationEvidence
Mean salary hike by performance rating
A rating of 4 is worth a 21.8% average hike versus 14.0% for a rating of 3.
Single employees leave at 2.5 times the rate of divorced colleagues
The quietest attrition predictor in this dataset sits in the personal-details field. 120 of 470 single employees left (25.5%), against 84 of 673 married (12.5%) and just 33 of 327 divorced (10.1%) — a clean gradient the chi-square test scores at χ² = 46.2 (Cramér's V = 0.177, p ≈ 0). Single employees run 1.58 times the company baseline; divorced employees run 0.63 times it.
So what
A single employee is 2.5 times more likely to leave than a divorced one (25.532 ÷ 10.092 = 2.53). The signal index offers a mechanism worth investigating: marital status is strongly entangled with stock option level (Cramér's V = 0.583), and employees with zero stock options leave at 24.4% versus 7.6% at option level 2.
Risk multiple, single vs divorced
2.53×
25.5%
Single (120 of 470 left)
1.58× baseline
12.5%
Married (84 of 673 left)
0.77× baseline
10.1%
Divorced (33 of 327 left)
0.63× baseline
HIGHEST
Single
25.5%
120 of 470 left · risk ratio 1.58
LOWEST
Divorced
10.1%
33 of 327 left · risk ratio 0.63
barchart · MaritalStatus × Attrition · OutcomeSegmentationEvidence
Attrition rate by marital status
Attrition steps down cleanly: single 25.5%, married 12.5%, divorced 10.1%.
Connected signals
You can read an employee's career length off their paycheck
Somewhere in this company's pay bands, a clock is ticking. Monthly income and total working years move together with a Pearson correlation of 0.773 (r² = 0.597, p ≈ 0), from an anchor point of $1,009 alongside 1 year of experience up to $19,999 alongside 34 years. The relationship is genuinely straight — a nonlinear fit improves r² by just 0.0004 (0.5978 vs 0.5974).
So what
Every extra $1,000 of monthly income maps to about 1.28 additional years of career experience (0.001277 × 1,000 = 1.28). Income alone accounts for 59.7% of the variance in career length — pay here is seniority converted to currency, echoing the income-tracks-experience signals throughout the index (income × job level runs at η² = 0.925).
Years of experience per $1,000 of monthly income
+1.28 yrs
0.773
Pearson r
Spearman 0.710 — rank-robust
59.7%
Variance in career length explained
r² = 0.597
$1,009 → $19,999
Observed income span
anchored at 1 and 34 working years
scatterchart · MonthlyIncome × TotalWorkingYears · CorrelationEvidence
Monthly income vs total working years
The cloud climbs steadily: r = 0.773, with income explaining 59.7% of career-length variance.
Tell me your job title and I will tell you your department
Some org charts are suggestions; this one is a wall. Department and job role are associated at Cramér's V = 0.939 on a 0-to-1 scale (χ² = 2,594.4 across 3 departments and 9 roles, p ≈ 0) — job titles live inside departments and almost never travel. A second signal in the same cluster shows the walls extend backwards in time: department and education field associate at V = 0.590.
So what
At V = 0.939, knowing one of 9 job titles all but names one of 3 departments — careers here are corridors, not networks. The education-field echo (V = 0.590) means the sorting happened before hiring: what you studied largely decided which corridor you entered.
0.939
Department ↔ JobRole (Cramér's V)
near the 1.0 maximum
0.590
Department ↔ EducationField (Cramér's V)
signal 9731b43ea59b5c73215c
barchart · Department associations · CategoricalDependencyEvidence
Association strength by dimension pair
Department and job role associate at Cramér's V = 0.939 — education field follows at 0.590.
Level 1 loses one employee in four; Level 4 loses one in twenty
The bottom rung of the ladder is where the grip fails. 143 of 543 Level 1 employees left (26.3%) — the single largest bloc of departures in the company — while Level 4 lost 5 of 106 (4.7%) and Level 5 lost 5 of 69 (7.2%). The gradient is decisive: χ² = 72.5, Cramér's V = 0.222, p ≈ 0.
So what
An entry-level employee is 5.6 times more likely to leave than a Level 4 colleague (26.335 ÷ 4.717 = 5.58), and Level 1's 143 leavers alone represent 60.3% of all 237 departures (143 ÷ 237 = 60.3%). Fix Level 1 and you fix most of the attrition problem outright.
Share of all leavers who are Level 1
60.3%
26.3%
Level 1 (143 of 543 left)
1.63× baseline
9.7%
Level 2 (52 of 534 left)
0.60× baseline
4.7%
Level 4 (5 of 106 left)
0.29× baseline
HIGHEST
Job Level 1
26.3%
143 of 543 left · risk ratio 1.63
LOWEST
Job Level 4
4.7%
5 of 106 left · risk ratio 0.29
barchart · JobLevel × Attrition · OutcomeSegmentationEvidence
Attrition rate by job level
Level 1 attrition runs at 26.3% — every level above it sits at 14.7% or lower.
Connected signals
The exit door is closest to the entrance: 29.8% of the newest quartile leaves
Attrition here is front-loaded — the danger zone is the first stretch of tenure. In the newest quartile of employees by years at company, 102 of 342 left (29.8%); in the longest-tenured quartile, 46 of 448 left (10.3%). The decline is monotonic across all four quartiles (χ² = 66.4, Cramér's V = 0.213, p ≈ 0).
So what
A newest-quartile employee is 2.9 times more likely to leave than a longest-tenured one (29.825 ÷ 10.268 = 2.90). Every year survived lowers the risk — retention effort spent in an employee's early tenure buys far more than the same effort spent later.
Risk multiple, newest vs longest-tenured quartile
2.90×
29.8%
Q1 tenure (102 of 342 left)
1.85× baseline
16.4%
Q2 tenure (39 of 238 left)
1.02× baseline — exactly average
10.3%
Q4 tenure (46 of 448 left)
0.64× baseline
HIGHEST
Newest quartile (Q1)
29.8%
102 of 342 left · risk ratio 1.85
LOWEST
Longest tenure (Q4)
10.3%
46 of 448 left · risk ratio 0.64
barchart · YearsAtCompany × Attrition · OutcomeSegmentationEvidence
Attrition rate by tenure quartile
The newest quartile leaves at 29.8% — nearly triple the longest-tenured quartile's 10.3%.
Connected signals
Closing summary
People don't leave this company at random — new, junior, underpaid employees working overtime do
Five independent cluster analyses corroborate one another: attrition concentrates where pay, level, and tenure are lowest and hours are longest, peaking at 39.8% for Sales Representatives against a 16.1% baseline. Everything else in the org chart — pay, seniority, even department walls — moves in lockstep with the same hierarchy.
16.1%
Baseline attrition (237 of 1,470)
39.8%
Sales Representative attrition
2.93×
Attrition risk multiple with overtime
81.6%
Income variance explained by job role
Data integrity
All nine analysis families ran (temporal, structural-quality, and entity-recurrence emitters found nothing to report on this dataset), producing 186 signals — 13 featured in full across 10 clusters and 173 catalogued in the signal index.
Ready to analyse your own data?
Quantiri turns any CSV into a finished intelligence report — ranked findings, interactive charts, plain-language narrative. Join the early-access list to run it on your own data.