Who does the work?

Describe a team with the four cards below. The page simulates thousands of teams like it and shows how unevenly the work splits between them. Every number is model output, not data about a real company.

Start from

1 · The team

How many people?

Which top group do you want to measure?

Top 1%Top 50%

The headline says how much of the work this group does.

2 · The work

What kind of work is it?

Everyone similarHuge gaps

3 · Hiring

How picky was hiring?

AnyoneTop 0.1%

How well can hiring spot real talent?

Mostly guessingPerfectly

How much do drive and luck differ between people?

Not at allA lot

Things no interview can see. Picky hiring doesn't shrink this.

4 · Projects and AI

Do some projects pay off far more than others?

All pay the sameWildly different

Even equally good people land on bets worth very different amounts.

Does AI lift the weakest people?

NoA lot

Does AI widen the gap at the top?

NoA lot

Best vs median

Half the work

80% of the work

Top person

Gini0 is a perfectly even split, 1 is one person doing everything

How fast the work piles up

Read across: the best X% of people do Y% of the work. The further the orange line sits above the blue one, the more uneven the team.

Typical team10th to 90th percentile teamEven split

One typical team, best to worst

Each bar is one person's share of all the work, ranked from best to worst.

The top group you're measuringEveryone else

How the model works

Each person has a hidden ability drawn from a bell curve. Hiring sees that ability plus noise (set by how well hiring spots talent) and takes the top slice of the market (set by how picky hiring was).
Has a ceiling turns ability straight into output, so nobody can do ten times the median. Skills multiply multiplies several abilities together, which gives a long right tail. Winner takes most maps ability onto a Pareto curve.
Drive and luck, and project payoff, each multiply a person's output by a random factor that hiring cannot see.
The two AI settings act last: widening the gap raises output to a power, then lifting the weakest adds a baseline for everyone equal to a share of the median person's output.
Results are the median over many simulated teams, with the 10th to 90th percentile range when at least 20 teams fit the compute budget. Very large companies run as one or two simulated teams.

The settings are modelling choices, not measurements. What the research does say: output across 198 samples and 633,263 people looked Paretian rather than bell-shaped (O'Boyle & Aguinis 2012); a later look at 229 samples found most fit a power law that bends down at the very top (Joo, Aguinis & Bradley 2017, summarized); and Price's Law says about √N people do half the work. A pure power law gives the same split whether you hire from everyone or only the top 1%. Selection only flattens the split when output has a ceiling.