The distribution of engineering output — across developers, across commits

Abstract

For every engineer and every commit merged in the selected window, we compute the Engineering-Terminal-Value (ETV). This page replaces aggregate totals with three complementary statistical views of how that quantity distributes across each population.

Reading engineers and commits side-by-side separates a people story (output tracks individual capability) from a commits story (output tracks the size of merged changes) — see Figure 5.

How ETV is measured →

Histogram $\hat{f} (x)$

The empirical density, binned linearly and clipped at the 99th percentile with an overflow bucket so tail outliers don't dominate the visual.

Lorenz curve $L (p)$

The cumulative share of total ETV produced by the bottom $p \cdot 100%$ of the sorted population; $L (p) = p$ is perfect equality, and the further $L$ sags below the diagonal, the more concentrated the distribution.

Gini coefficient $G$

$G = 1 - 2 \int_{0}^{1} L (p) d p$ , with $G \in [0, 1]$ , is a scalar summary (0 = uniform contribution, 1 = all mass in a single entity); reported alongside the top-20 % share and the $p_{90} / p_{50}$ ratio as redundant sanity checks.

The distribution of engineering output — across developers, across commits

Abstract

Reading engineers and commits side-by-side separates a people story (output tracks individual capability) from a commits story (output tracks the size of merged changes) — see Figure 5.

How ETV is measured →

Histogram $\hat{f} (x)$

The empirical density, binned linearly and clipped at the 99th percentile with an overflow bucket so tail outliers don't dominate the visual.

Lorenz curve $L (p)$

Gini coefficient $G$

1Developer distribution

How Performance is spread across active engineers in the selected window. A flat histogram means the team contributes evenly; a long right tail means a few top engineers carry most of the output. The Lorenz curve bends toward the bottom-right in proportion to that concentration.

Gini coefficient: 0.72; strong concentration
Top-20% share: 76%; top 129 of 643 engineers
p90 / p50: 10.0×; 16.2 ETV vs. 1.6 ETV

Figure 1 · Engineers per ETV bucket [ETV]

Bucket size

Read each bar as: “there are this many engineers whose total ETV in the selected window falls in this bucket.” Linear x-axis, clipped at p99 with an overflow bin on the far right so a couple of outliers don't dominate the view.

Figure 2 · Cumulative share of ETV across engineers

Read a point (x, y) as: “the bottom x % of engineers produced y % of the total ETV.” Dashed diagonal = perfect equality (every engineer ships the same). The further the curve sags below it, the more output is concentrated in the top few.

2Commit distribution

How Performance is spread across individual commits. Most commits are small; a few large ones carry disproportionate weight. The histogram uses a log x-axis because the distribution is heavy-tailed — on a linear axis everything collapses into the first bar.

Gini coefficient: 0.73; strong concentration
Top-20% share: 76%; top 6418 of 32089 commits
p90 / p50: 9.1×; 0.4 ETV vs. 0 ETV

Figure 3 · Commits per ETV bucket (log scale) [ETV]

Buckets

Read each bar as: “there are this many commits whose ETV falls in this bucket.” Buckets are spaced logarithmically because the distribution is heavy-tailed — on a linear axis everything collapses into the leftmost bar.

Show

Top 10 commits by ETV›

Figure 4 · Cumulative share of ETV across commits

Read a point (x, y) as: “the smallest x % of commits account for y % of the total ETV.” Diagonal = every commit contributes the same. The deeper the curve sags, the more a handful of large commits carry the window.

3Head-to-head

Is Performance a people story or a commits story? The same ETV, viewed two ways — once as how it piles up across engineers, once as how it piles up across individual commits.

Figure 5 · Engineer vs. commit Lorenz curves

Same Lorenz framing as Figures 2 and 4, overlaid. The x-axis is the bottom x % of whichever thing you're sorting (engineers for the teal curve, commits for the orange one); the y-axis is the share of total ETV they produce. Dashed diagonal = perfect equality. The deeper a curve sags, the more concentrated that dimension is.

Engineers · Gini 0.72Commits · Gini 0.73

Figure 6 · Volume vs. complexity per engineer

Each dot is one of 643 engineers active in the selected window: x = commits shipped, y = average ETV per commit. Dashed lines mark the median on each axis, splitting the plot into four quadrants. Dot opacity and size encode total ETV — top performers stand out as larger, fully saturated orange. Read the chart by where the strongest dots cluster: upper-left = complexity specialists (few commits, each high-impact), lower-right = volume merchants (many small commits), upper-right = balanced stars, lower-left = peripheral.

Histogram f^​(x)

Lorenz curve L(p)

Gini coefficient G

Histogram f^​(x)

Lorenz curve L(p)

Gini coefficient G

1Developer distribution

Figure 1 · Engineers per ETV bucket [ETV]

Figure 2 · Cumulative share of ETV across engineers

2Commit distribution

Figure 3 · Commits per ETV bucket (log scale) [ETV]

Figure 4 · Cumulative share of ETV across commits

3Head-to-head

Figure 5 · Engineer vs. commit Lorenz curves

Figure 6 · Volume vs. complexity per engineer

2Commit distribution

Figure 3 · Commits per ETV bucket (log scale) [ETV]

Figure 4 · Cumulative share of ETV across commits

1Developer distribution

Figure 1 · Engineers per ETV bucket [ETV]

Figure 2 · Cumulative share of ETV across engineers

3Head-to-head

Figure 5 · Engineer vs. commit Lorenz curves

Figure 6 · Volume vs. complexity per engineer

Histogram $\hat{f} (x)$

Lorenz curve $L (p)$

Gini coefficient $G$

Histogram $\hat{f} (x)$

Lorenz curve $L (p)$

Gini coefficient $G$