View mode Color

Demo, synthetic data

Usul Dynamics: People Analytics

Usul Dynamics is a fictional robotics company simulated from founding in June 2018 through today: ... synthetic employee records, zero real people. Every pattern in the data was planted on purpose, documented, and recovered by a validation gate before this page shipped (receipts at the bottom).

...Headcount

...YoY growth

...Voluntary turnover

...Regretted share

...Median tenure

...Open reqs

From

Department

Location

Level group

Headcount over time

Month-end totals stacked by department. Hover or tap a month for the breakdown.

Show

Hires up, exits down, per month

View SQL

Attrition by tenure

Annualized voluntary turnover: voluntary exits in each tenure range divided by person-years worked in that range.

View SQL

Where attrition concentrates

Annualized voluntary turnover by department and level. Darker means higher; hover or tap a cell for exits and person-years. Levels are the HRIS snapshot: at exit for leavers, current for actives.

View SQL

Metric definitions

Headcount: People actively employed on the date shown: hired on or before it, not yet exited.
YoY growth: Headcount today divided by headcount exactly one year earlier, minus one.
Annualized voluntary turnover: Voluntary exits in the trailing 12 months divided by average month-end headcount over those 12 months.
Involuntary turnover: The same calculation using involuntary exits (performance and restructuring).
Regretted share: Of voluntary exits in the window, the percent who carried a performance rating of 4 or 5.
Median tenure: The middle value of years-since-hire across everyone currently employed.
Open reqs: Requisitions with no hire yet at the data cutoff. Reqs carry a department and level, not a location.
Turnover by tenure: Voluntary exits while in a tenure range divided by the person-years worked in that range, annualized.
Person-years: Time worked summed across people, in years: two people for six months is one person-year.
Payroll cost: The sum of active employees' annual base salaries in that month, divided by 12: a monthly run rate.

How this was built

A seeded generator simulates the company month by month and writes plain CSVs (employees, comp events, requisitions). A DuckDB transform aggregates them into the single JSON file this page fetches. The build refuses to ship unless every planted effect is recovered within its documented tolerance, so the numbers above are not just plausible: they are checked.

Validation report: planted vs recovered

View SQL (full transform)

View

Department

Descriptive shares and annualized rates. A cell with fewer than 10 people is shown as n too small, never a zero or a bar.

Representation by level

In the gender view, the bar is women's share of women and men today and the tick marks the share 36 months ago: the gap between them is the three-year change. The ethnicity view stacks everyone to 100%. Each person is counted at their level at the time. Hover or tap for counts.

View SQL

Women's share over time

Women as a share of women and men in each level group, month by month. A line breaks where the group has fewer than 10 people. Use the department filter above to re-scope.

View SQL

How people enter, rise, and leave

Annualized rate per 100 person-years in each level group, women vs men: hires in, promotions out (one level up), and voluntary exits. Hover or tap a bar.

Hiring in

Promotion (one level up)

Voluntary exit

View SQL

Why the pooled promotion view looks even

Engineering and Sales only, fixed to the level each person was hired into: share promoted at least once within three years. Holding hire level constant removes the confound of women starting more junior. Levels with fewer than 10 of either gender are omitted. Always Engineering plus Sales, regardless of the department filter.

View SQL

Models run in the pipeline over every active employee as of ; nothing statistical runs in this page, so the module 1 filters do not apply here. The binary gap compares women and men (ethnicity is module 2's lens); nonbinary colleagues are included in the remediation scan but not the two-group gap.

Two honest numbers

Both are honest answers to different questions. The unadjusted gap asks: what does the typical woman earn here vs the typical man? It bundles pay setting together with who holds which level in which department. The adjusted gap asks: are two colleagues with the same level group, department, location, tenure, and performance paid alike? Closing the adjusted gap is a comp-policy fix; closing the unadjusted gap also takes the representation shifts in module 2.

View SQL

From raw to unexplained

Each bar adds one control to the regression, in the order shown; blue is gap a control explains away. Sequential attribution depends on that order (level absorbs shared variance first); what no control absorbs is the unexplained remainder, in red. Hover or tap a bar.

View SQL

Median pay by level

View SQL

The repair bill

View SQL

How these numbers were computed

Validation report: pay effects, planted vs recovered

What a hire costs, and how long it takes

Mean fully loaded cost per filled requisition: agency fee, ad spend, referral bonus, recruiter time at $85/h, interviewer panel time at $110/h. Signing bonuses are compensation, not recruiting cost. Bars in each source cluster are IC1-2, IC3-5, M, DIR+ left to right; a dashed box means fewer than 10 hires. Hover or tap for the component breakdown.

Mean cost per hire by source and level group

Median days to fill, by level

View SQL

Where recruiting money goes

Stage counts are modeled from standard pass-through rates (50 sourced : 12.5 screened : 5 onsite : 1.25 offers : 1 hire); every dollar is an actual requisition component, allocated to the stage the work happens in. Agency fees and ads buy sourcing; referral bonuses pay out at hire.

Show

View SQL

What one exit costs to replace

Three auditable components per level group: recruiting (the mean cost per hire above), vacancy loss (60% of daily comp for the median days the seat sits empty), and ramp loss (half productivity while the replacement ramps: 3 months for ICs, 5 for managers, 8 for directors and above). Comp is base salary; rules of thumb that also count lost knowledge and team drag run 0.5x to 2x salary, so these are conservative.

View SQL

Retain or replace: the trade, priced

Pick a segment, set a retention raise, and assume how much attrition it removes. The raise applies to everyone in the segment; the attrition reduction is your assumption, not a measured elasticity. Replacement cost per avoided exit comes from the model above, at the segment's average comp.

Department

Level group

Retention raise

Assumed attrition reduction

Breakeven: avoided replacement cost vs raise cost, by assumed reduction

View SQL

Model constants and what is excluded

What this means for your business

Same suite, your HRIS

Usul Dynamics is fictional; the method is not. Pointed at your real HRIS and ATS exports, the same pipeline ships a validated workforce mart and an interactive suite your leadership can actually explore, with every number backed by SQL you can audit.

Book an intro call See services