80% of JavaScript code comes from 20% of its features?

I analyzed several popular JavaScript repositories to see how much of the language is actually used in real projects.

Turns out the 80/20 rule is real — about 20–25% of language features cover 80–90% of the code.

I believe the data and interpretation are roughly correct — but I’m always open to feedback or alternative views.

Aggregate Per 1K LOC

Summary

Aggregate Analysis Summary

Dataset Overview:

• Total repositories analyzed: 55
• Total features in JavaScript catalog: 121
• Analysis covers: totals, presence, perKLOC metrics

Total Occurrences - Aggregate Results

• 80% of totals achieved with 17 features (14.0% of catalog)
• 90% of totals achieved with 30 features (24.8% of catalog)
• 95% of totals achieved with 42 features (34.7% of catalog)

File Presence - Aggregate Results

• 80% of presence achieved with 33 features (27.3% of catalog)
• 90% of presence achieved with 46 features (38.0% of catalog)
• 95% of presence achieved with 60 features (49.6% of catalog)

Per 1000 Lines of Code - Aggregate Results

• 80% of perKLOC achieved with 21 features (17.4% of catalog)
• 90% of perKLOC achieved with 34 features (28.1% of catalog)
• 95% of perKLOC achieved with 46 features (38.0% of catalog)

---

Top 7 Features

Rank	Feature	Count	%	Cumulative %
1	binaryOp_/	1298069	20.08%	20.08%
2	labels	781944	12.09%	32.17%
3	assignOp_=	711165	11.00%	43.17%
4	functions	392279	6.07%	49.24%
5	return	265406	4.10%	53.34%
6	if	255167	3.95%	57.29%
7	binaryOp_+	253173	3.92%	61.21%

Pareto Analysis

80% of usage comes from 17 features (14.0% of catalog):
binaryOp_/, labels, assignOp_=, functions, return, if, binaryOp_+, binaryOp_*, binaryOp_|, arrowFunctions, binaryOp_===, binaryOp_-, binaryOp_>>, logicalAND, binaryOp_<, ternary, awaitExpressions

90% of usage comes from 30 features (24.8% of catalog):
binaryOp_/, labels, assignOp_=, functions, return, if, binaryOp_+, binaryOp_*, binaryOp_|, arrowFunctions, binaryOp_===, binaryOp_-, binaryOp_>>, logicalAND, binaryOp_<, ternary, awaitExpressions, logicalOR, binaryOp_>, binaryOp_!==, exportDecls_named, importDecls, switchCases, jsxElements, updateOp_++, templateLiterals, forClassic, asyncFunctions, break, spreadElement

95% of usage comes from 42 features (34.7% of catalog):
binaryOp_/, labels, assignOp_=, functions, return, if, binaryOp_+, binaryOp_*, binaryOp_|, arrowFunctions, binaryOp_===, binaryOp_-, binaryOp_>>, logicalAND, binaryOp_<, ternary, awaitExpressions, logicalOR, binaryOp_>, binaryOp_!==, exportDecls_named, importDecls, switchCases, jsxElements, updateOp_++, templateLiterals, forClassic, asyncFunctions, break, spreadElement, classFields, classMethods, binaryOp_==, binaryOp_>>>, interpolatedTemplates, throw, classDecls, binaryOp_&, assignOp_+=, binaryOp_<<, binaryOp_%, updateOp_--

Catalog Usage Summary

Total features in JavaScript catalog: 121

Usage Threshold	Features Required	Catalog Percentage
80%	17	14.0%
90%	30	24.8%
95%	42	34.7%

---

---

Metrics explained

• Totals How often each feature appears overall across all repositories. Shows which constructs dominate codebases globally (like if, =, or function).

• Presence In how many different files a feature is found. Reflects how widespread a construct is — some appear everywhere, others only in a few modules.

• Per KLOC Density per 1000 lines of code. Normalizes for project size, letting you compare big and small repos fairly.

Each of these views shows a clear Pareto pattern:

A small core of features (around 20–30 out of 121) accounts for most of the language usage.

You can see the results and charts here:

Live reports

GitHub repo

Any comments or insights are welcome

Comments

[Gohar]

Yes, that’s true. Around 80% of JavaScript code usually relies on just 20% of its core features, showing how a small set of tools covers most common programming needs.

[Peter Vivo]

On my coding style I near alway use const instead function. Also avoide using this expect the web component.

[Francesc López]

Thanks, Peter — that’s a great point!
I’ve just added the top-N most used features per repo, and it’s really interesting to see how it matches what you said: Express barely uses arrow functions, while Fastify is around 0.75, clearly preferring them.
It made me realize that coding styles can actually be inferred from these metrics — like your example of preferring const functions and avoiding this except in web components.
Really appreciate the insight!