Summary. The 80/20 rule is not a law of nature but an observed pattern: a small share of the contributors to an outcome accounts for most of it. In 1897 Vilfredo Pareto published not an “80/20” ratio but a slope parameter (α) for the distribution of income; for England in 1843 its value is 1.50. By CEOtudent’s calculation, that slope means the top 20% in the tax records received about 58.5% of income; an exact 80/20 would require α of about 1.16, and none of the 17 values Pareto did not flag as uncertain is that low. Joseph Juran gave the “Pareto principle” its name and later wrote himself that the name was a mistake. In current data the pattern is strong but the ratio varies: the richest 10% in the world own 75% of wealth (World Inequality Report 2026), and for consumer packaged goods brands the top 20% of buyers make 73% of sales on average. CEO move: allocate your attention like capital, to the measured vital few. Student move: treat 80/20 not as an assumption but as a hypothesis you retest every quarter.
This article covers the concept itself and how to measure it. Which tasks to automate first is covered in the 80/20 of knowledge-work automation. Which metric to use for measuring contribution is the subject of the guide how to measure knowledge work output, and the habit of looking at outside data instead of assumptions is discussed in base rates and the outside view.
What does the 80/20 Pareto principle actually say?
Joseph M. Juran, the quality-management pioneer who gave the principle its name, defines the concept without a ratio: in any population which contributes to a common effect, a relative few of the contributors account for the bulk of the effect. Juran called this the “vital few and trivial many”; he later replaced “trivial many” with “useful many.” The figure “80/20” does not appear in Juran’s 1974 text itself (it appears only in the key-takeaways box the Juran Institute added later).
Three readings of the concept are often mixed up: the descriptive reading (many distributions are skewed; the data support this), the managerial reading (limited attention should go to the few items that produce most of the outcome; measurement comes first) and the mistaken reading (the ratio is exactly 80 to 20 everywhere, so there is no need to measure). Through the CEOtudent lens the value lies in the second reading: the CEO allocates limited resources to the highest-return area, and the student measures that area instead of guessing it.
What did Pareto actually find?
In the second volume of his Cours d’économie politique (Lausanne, 1897), Vilfredo Pareto examined income tax returns. In his own words, despite the uncertainties that taxpayers’ returns carry, they were still the surest basis available for knowing, at least approximately, how wealth is distributed. He denoted a given income level by x and the number of taxpayers with an income above that level by N, and plotted the logarithms of the two variables on the same chart. The points lined up almost on a straight line; he called the slope of that line the “inclinaison” and denoted it α.
Pareto’s England data were not a census of wealth or land. As he himself noted, only for Schedule D (commerce and professions) was there an extended classification of taxpayers by size of income in England, and in the 1843 example the minimum income in the calculation was 150 pounds sterling. The curve therefore described incomes above the tax threshold, not the whole population. The Italian data were not land ownership either, but family tax records from cities such as Florence and Perugia and aggregate income data for Italian cities.
Pareto saw that the slope varied little from country to country and argued that this could not be due to chance alone; but he added that where purely empirical laws are concerned, one cannot be too cautious. A decrease in the slope indicated less inequality of incomes. What Pareto measured, in other words, was not a fixed ratio but a parameter that changes from country to country and from year to year.
Table 1. Slope values published by Pareto (selected rows, as printed in the source)
| Country / city, year | Slope α |
|---|---|
| England, 1843 | 1.50 |
| England, 1879-80 | 1.35 |
| Prussia, 1852 | 1.89 |
| Prussia, 1894 | 1.60 |
| Saxony, 1886 | 1.51 |
| Florence | 1.41 |
| Italian cities (combined) | 1.45 |
| Basel, 1887 (wealth data) | 1.24 |
| Paris (rents) | 1.57 |
| Augsburg, 1526 (uncertain data) | 1.13 |
| Peru, late 18th century (uncertain data) | 1.79 |
Source: Vilfredo Pareto, Cours d’économie politique, second volume, 1897, p. 312, §959. Of the 22 values in the table, Pareto marked 5 as “derived from rather uncertain data”; this article calls the rest firm values. Because the OCR text of this table is garbled, the values were read from the scanned page image.
Does Pareto’s own slope say 80/20?
It is possible to move from slope to ratio. If a distribution follows a Pareto law of the form N(x) = A / x^α and α is greater than 1, the share of the total held by the top fraction p is p^(1 – 1/α). This is not a ratio Pareto wrote in this form but a standard result derived from the distribution he described; the CEOtudent calculation was checked both with the closed form and with numerical integration.
Table 2. Which “x/20” ratio do Pareto’s slopes correspond to? (CEOtudent analysis, using Pareto’s 1897 data)
| Case | α | Share of the top 20% (derived) | Reading |
|---|---|---|---|
| England, 1843 | 1.50 | 58.5% | About 58/20 |
| Median of the 17 firm values | 1.57 | 55.7% | About 56/20 |
| Lowest firm value (Basel, 1887, wealth) | 1.24 | 73.2% | Most unequal firm case |
| Highest value (Prussia, 1852) | 1.89 | 46.9% | Most equal case |
| Slope required for an exact 80/20 | 1.16 | 80.0% | None of the firm values is this low |
| Augsburg, 1526 (uncertain data) | 1.13 | 83.1% | The only row above 80/20 |
CEOtudent analysis. The share is calculated among the units included in the tax records and does not apply to the whole population. The Basel row is based on wealth data; the other firm rows are based on income data. The calculation and numerical check are in calc.py.
Inside the tax records, Pareto’s own data show not 80/20 but, in 13 of the 17 firm values, a concentration between 50/20 and 65/20. The pattern, a minority receiving a disproportionate share, is real; the figure “80” does not come from anything Pareto wrote.
Where did the name “Pareto principle” come from?
Juran tells the story himself in his 1974 text “The Non-Pareto Principle; Mea Culpa.” In the mid-1920s, as a young engineer, he observed that quality defects are unequal in frequency: when a long list of defects was arranged in order of frequency, a relative few of the defects accounted for the bulk of the defectiveness. He saw a similar phenomenon in employee absenteeism and in the causes of car accidents. He first encountered Pareto’s work at General Motors, in a study showing that the executive salary pattern fit one of Pareto’s mathematical models surprisingly closely.
In the late 1940s, while preparing the manuscript for the first edition of the Quality Control Handbook, he needed a short name for this universal pattern and chose Pareto’s. After challenges from Dorian Shainin and others, he examined what Pareto had actually done and conceded that he had given the principle the wrong name: Pareto’s work had been in the economic sphere, and his models were not intended to be applied to other fields. He wrote that the cumulative curves should have been properly identified with Lorenz, and complained that contemporary authors had attributed to Pareto things he did not do. The story that 80% of the land in Italy belonged to 20% of the population is a typical example of such embellishments: no such sentence was found in the primary text examined. The correct attribution is this: the mathematics of the distribution belongs to Pareto, and the generalization that turned it into a management principle belongs to Juran.
How 80/20 is the 80/20 rule in real data?
Nine datasets from different fields, each with a primary source that could be opened, were placed side by side. The share printed by each source was kept; CEOtudent added two columns: the concentration factor (the top group’s share divided by its population share; 4 under 80/20) and the share that would be expected at the same cut if the 80/20 rule held across the whole distribution. The second column allows a fair comparison with sources that publish a 10% or 15% cut.
Table 3. How 80/20 is the 80/20 rule in real data? (CEOtudent analysis; shares as published in the sources)
| Dataset | Cut | Share of the outcome (as published) | Concentration factor | Share expected at the same cut under the 80/20 rule | Reading |
|---|---|---|---|---|---|
| Global wealth, 2025 (World Inequality Report 2026) | Richest 10% of adults | 75% | 7.5x | 72.7% | Slightly more concentrated than 80/20; the top 1% hold 37% of wealth |
| Global income, 2025 (World Inequality Report 2026) | Top 10% | 53% | 5.3x | 72.7% | Less concentrated; the bottom 50% get 8% of income |
| US healthcare spending, 2022 (AHRQ MEPS) | Top 10% of people by spending | 65.9% | 6.59x | 72.7% | Less concentrated; the top 5% account for 49.7% of spending |
| Consumer packaged goods brands (Kim, Singh and Winer 2017; Nielsen panel, over 100,000 households, 6 years) | Top 20% of buyers | 73% (mean) | 3.65x | 80% | Less concentrated; niche brands run higher |
| 339 publicly traded non-CPG companies (McCarthy and Winer 2019) | Top 20% of customers | 67% | 3.35x | 80% | 59% for subscription businesses, 68% for non-subscription businesses |
| UK brands, Kantar data (Ehrenberg-Bass Institute) | Top 20% of buyers | 40% in a quarter, over 50% in a year, just over 60% in five years | 2.0 to 3.0x | 80% | The ratio rises as the measurement window lengthens |
| Faults in large software systems (Ostrand, Weyuker and Bell 2005) | The 20% of files selected by the model | 71% to 92%, average 83% | 4.15x | 80% | True concentration is at least this high |
| Scientific citations (Seglen 1997; three biochemistry journals) | Most cited 15% of articles | 50% | 3.33x | 76.9% | The most cited half accounts for 90% of citations |
| England tax incomes, 1843 (Pareto α = 1.50) | Top 20% of taxpayers | 58.5% (derived) | 2.92x | 80% | CEOtudent calculation in Table 2 |
CEOtudent analysis; the sources are items 1, 3-7, 9 and 11 in the Sources section. The “expected share” column assumes that 80/20 is a Pareto law that repeats itself at every cut (α about 1.16). In the Ostrand row the 20% was selected by the model’s prediction, so the true top 20% contains at least this many faults. The healthcare spending row is only an example of a distribution; this article is general information and is not medical or financial advice.
Three conclusions follow from the table:
- The pattern is almost everywhere. In all nine rows the top group takes at least 2 and at most 7.5 times its population share.
- The ratio is rarely 80/20. Only two rows reach or exceed the share the 80/20 rule predicts at the same cut: global wealth and software faults. In customer and brand data the ratio runs between 60/20 and 73/20 over long windows and falls as low as 40/20 over short ones; income and citation data also stay below the share 80/20 predicts.
- Measurement choices set the number. The same UK brand data come out at 40/20 in a one-quarter window and above 60/20 in a five-year window. In Kim, Singh and Winer, the ratio tends to be lower for large brands and categories and higher for niche brands.
The message is not “80/20 is wrong” but “80/20 is a measurement result”: in your work the real ratio could be 55/20, or it could be 85/20.
Why does the ratio move so much?
According to the abstract of Schmittlein, Cooper and Morrison’s 1993 paper “Truth in Concentration in the Land of (80/20) Laws,” interpreting these statistics correctly is not nearly as easy or immediate as computing them. Four factors stand out.
Measurement window. According to a summary by Byron Sharp and Charles Graham of the Ehrenberg-Bass Institute, based on Kantar data, for the average brand the share of the top 20% of buyers is typically only 40/20 in one quarter, rising to over 50/20 for an analysis period of one year and to just over 60/20 over a five-year window.
Unit and sample. Pareto counts taxpayers, Kim and colleagues count households, MEPS counts people. According to the Ehrenberg-Bass authors, loyalty card data from a single retail chain can inflate the ratio, because it does not capture households’ purchases in other stores.
Drift and regression to the mean. According to Sharp and Graham, this year’s top 20% of buyers, who generated 60% of sales this year, will be worth about 45% next year; the 15-point gap is the effect of regression to the mean.
Which outcome is measured. According to the abstract of Fenton and Ohlsson’s 2000 study, in a major commercial software system most of the faults discovered in prerelease testing were concentrated in a small number of modules; but the modules that were most fault-prone prerelease were among the least fault-prone postrelease, and vice versa. The “most problematic 20%” can be an entirely different set depending on the stage you measure. On the other hand, the model of Ostrand, Weyuker and Bell, using each file’s fault and modification history from previous releases, predicted quite accurately where faults would concentrate in the next release; with the right metric, the past is a good guide.
What does 80/20 change for the CEO and the student?
CEO lens: attention is capital. An executive does not split a budget evenly; they weight it toward areas with measured returns and set criteria in advance for exiting low-return areas. The individual equivalent is concentrating your weekly hours on the few clients, projects and tasks that produce the highest contribution. The kill criteria method, for deciding in advance what you will stop, is the natural complement to a Pareto audit.
Student lens: the ratio is a hypothesis. A good student finds the “most important 20%” by experiment, not intuition, and retests the result with new data. In learning, 80/20 is not an excuse to skip the fundamentals; how to decide what you will deliberately not learn is covered in strategic ignorance.
How do you run a Pareto audit for knowledge work?
This protocol is for measuring your own distribution instead of assuming 80/20. The scoring tables for automation candidates appear in the 80/20 of knowledge-work automation, so they are not repeated here; this protocol is the measurement step that comes before that scoring.
Table 4. Quarterly Pareto audit for knowledge work (CEOtudent editorial framework)
| Step | What to do | Why |
|---|---|---|
| 1. Choose the domain and the metric | One domain (clients, tasks, products, information sources, causes of interruption) and one real metric (contribution share, hours, output actually used, minutes lost). | If the metric changes, the vital few changes too (Fenton and Ohlsson). |
| 2. List 20 to 30 items | Write down every item in the domain; do not skip the small and boring ones. | If the tail is invisible, the ratio comes out higher than it really is. |
| 3. Fix the window | Use at least one quarter of data and note the window. | The length of the window changes the ratio (Ehrenberg-Bass). |
| 4. Sort and compute the cumulative share | Rank the items from largest to smallest; record the share of the top 20% and how many items it takes to reach 80%. | This is your real ratio. |
| 5. Decide on each item | Keep and grow, delegate or automate, drop, or mark as a “protected tail” (Table 6). | A measurement that does not lead to a decision is only a report. |
| 6. Put the decision on the calendar | Write an end date for what you drop and an owner for what you delegate; track them in your weekly review. | A decision that never reaches the calendar does not get carried out. |
| 7. Repeat at the end of the quarter | Measure again with the same metric and window; count how many items from last quarter’s top 20% are still there. | Regression to the mean shows that the vital few shifts. |
A small example shows why step four matters. The ten clients and values below are hypothetical inputs (illustrative assumption); they only show how the calculation is done.
Table 5. Hypothetical example: cumulative share in a ten-client portfolio (all inputs hypothetical)
| Client | Annual contribution (hypothetical input, units) | Share | Cumulative share | Cumulative share of items |
|---|---|---|---|---|
| A | 38 | 38% | 38% | 10% |
| B | 22 | 22% | 60% | 20% |
| C | 12 | 12% | 72% | 30% |
| D | 8 | 8% | 80% | 40% |
| E | 6 | 6% | 86% | 50% |
| F | 5 | 5% | 91% | 60% |
| G | 4 | 4% | 95% | 70% |
| H | 2 | 2% | 97% | 80% |
| I | 2 | 2% | 99% | 90% |
| J | 1 | 1% | 100% | 100% |
CEOtudent editorial framework; the values are hypothetical inputs (illustrative assumption), and the calculation is in calc.py.
In this portfolio the first two clients, that is 20% of the items, make up 60% of the contribution: the ratio is 60/20. The 80% mark of contribution is reached only at the fourth client, at 40% of the items; the bottom five clients together account for 14%. Here the rule “focus on the top 20% and drop the rest” would mean giving up clients C and D, that is 20% of the contribution.
Table 6. Default decision by position in the cumulative ranking (CEOtudent editorial framework)
| Position in the cumulative ranking | Question to ask | Default decision |
|---|---|---|
| First 50% of the contribution | Is the contribution durable and scalable, and is dependence on a single item a risk? | Keep it, give it your best hours, monitor the dependence |
| Between 50% and 80% cumulative | Is the item standard and repetitive? | Delegate it, template it or automate it |
| Tail beyond 80% | Does it have strategic, learning or risk value? | If not, drop it; if so, mark it as a “protected tail” and write down the reason |
When does 80/20 mislead?
This list is a CEOtudent editorial framework; in six situations, applying the result of an audit blindly does harm.
- When the total value of the tail is not small. In the Ehrenberg-Bass data, light and occasional buyers are the majority of buyers and still account for 40% of long-run turnover; according to the authors, they are also the primary source of brand growth.
- When the vital few shifts. This year’s top 20% takes a smaller share next year (from 60% to about 45%). Base the decision not on a single measurement but on a comparison of two consecutive ones.
- When the risk is not where you measure. In Fenton and Ohlsson, modules with few faults in testing were among those with the most faults in operation. Security, compliance and reputational risks usually hide in the tail.
- When returns are compounding and delayed. Foundational concepts, relationships and long-term skills look low-return over a short window; a one-quarter measurement does not show returns that accumulate over years.
- When the window is short or the sample is incomplete. A distribution that looks like 40/20 in one quarter exceeds 60/20 over five years; single-channel data do not see behavior in other channels.
- When the “trivial many” is not optional. Legal obligations, safety checks and basic relationship maintenance may come out low in the ranking but cannot be dropped; the right move is not to drop them but to do them at the lowest cost.
What happened to the claims in the old article?
Every concrete claim in the 2020 version was checked against a primary source. No claim that could not be verified was carried over into this version.
Table 7. Audit of the old claims (CEOtudent analysis)
| Old claim | Status | What the primary source says |
|---|---|---|
| Pareto argued that 80% of the land in Italy belonged to 20% of the population | Not verified, removed | There is no such sentence in the income-curve section of the Cours (§958-961); the Italian data are cities’ tax and income records |
| Pareto found that 80% of the wealth in England belonged to 20% of the population | Corrected | The England data are not wealth but Schedule D commerce and professions incomes; α = 1.50 for 1843, i.e. about 58.5/20 among taxpayers (CEOtudent calculation) |
| Juran said that 80% of effects come from 20% of causes and called this the Pareto principle | Partly verified, corrected | Juran gave it the name and later wrote that it was the wrong name; his own text has not a fixed 80/20 ratio but the phrase “vital few and trivial many” |
| We spend 80% of our time with 20% of our friends | Unsourced, removed | Not based on any study |
| 20% of drivers cause 80% of traffic accidents | No primary source found, removed | Peck, McBride and Coppin (1971, California driver records, abstract): highly selective treatment programs can never be expected to result in a dramatic reduction of the overall accident rate |
| 86% of the world’s wealth is held by 8% of the population | Unsourced, replaced with current data | World Inequality Report 2026: the richest 10% own 75% of wealth, the top 1% own 37%, the bottom 50% own 2%; the report has no 8% cut |
| The wealth of the top 3 of the 10 richest people equals that of the other 7 | Undated and unsourced, removed | Not needed for the topic |
| The Pareto principle can be linked to marginal utility (200-person team example) | Conceptual error, removed | The example describes the law of diminishing returns, not marginal utility; neither is the same thing as the Pareto pattern |
| Most exam questions came from the 10 pages you skipped | Not evidence, removed | An anecdote; the limit in learning is covered in item 4 of the “when it misleads” list |
| Solve the problems that cause the most trouble and you solve about 80% of your problems | Unsourced, removed | Replaced with a measurement-based Pareto audit (Table 4) |
Marginal utility describes the extra satisfaction from each additional unit consumed, and diminishing returns describes the shrinking extra contribution of each worker added to production; the Pareto pattern, by contrast, is about how outcomes are shared among units.
Why was this article rewritten?
The original version of this page was published in 2020 as an article of about 410 words. Most of its figures were unsourced, its only source line was an external link to a student news site, and the text contained ad code, embedded videos and first-person narration. The topic and the search intent were kept. The origin of the concept was grounded in Pareto’s 1897 text and in Juran’s own account, current global data and a new CEOtudent analysis were added, and a protocol was included that lets readers measure their own distribution; unverifiable claims, ad code, images, videos and external links were removed.
Frequently asked questions
Is the 80/20 rule a scientific law?
No. It is a common pattern of skew, but the ratio is not fixed; of the nine datasets here, only two reach the concentration that 80/20 predicts at the same cut.
Who discovered the Pareto principle?
Vilfredo Pareto published the curve of the income distribution in the late 1890s. The person who turned the pattern into a management principle applied to every field, and who named it the “Pareto principle,” was Joseph M. Juran; Juran later wrote that the name was a mistake.
Did Pareto say that 80% of the land in Italy belonged to 20% of the population?
No such statement was found in the primary text examined. Pareto’s Italian data are cities’ tax and income records, and his result is not a ratio but a slope parameter.
How do I measure 80/20 in my own work?
Choose one domain and one metric, measure 20 to 30 items with at least one quarter of data, sort them from largest to smallest and compute the cumulative share. The share of the top 20% and the number of items it takes to reach 80% are your real ratio (Table 4 and Table 5).
Does 80/20 mean I can skip 80% of the topics when I learn?
No. 80/20 helps set priorities, but the return on foundational concepts is delayed and compounding; skipping them makes later learning harder.
How much of the world’s wealth do the richest hold?
According to the World Inequality Report 2026, in 2025 the richest 10% of adults held 75% of global wealth and the top 1% held 37%; the share of the bottom 50% was 2%. In income, the top 10% received 53% of global income and the bottom 50% received 8%.
Sources
- Vilfredo Pareto. Cours d’économie politique, Tome second. F. Rouge, Lausanne, 1897. “La courbe des revenus”, §958-961, pp. 304-313; slope table on p. 312 (Internet Archive scan; table read from the page image).
- Joseph M. Juran. The Non-Pareto Principle; Mea Culpa. Text dated 1974, Juran Institute archive copy.
- Lucas Chancel, Ricardo Gómez-Carrera, Rowaida Moshrif and Thomas Piketty (coordinators). World Inequality Report 2026. World Inequality Lab.
- Adriana Hernandez-Viver and Emily M. Mitchell. Concentration of Healthcare Expenditures and Selected Characteristics of People with High Expenses, United States Civilian Noninstitutionalized Population, 2018-2022. Statistical Brief No. 560, Agency for Healthcare Research and Quality, March 2025.
- Baek Jung Kim, Vishal Singh and Russell S. Winer. The Pareto rule for frequently purchased packaged goods: an empirical generalization. Marketing Letters, 28(4), 491-507, 2017 (abstract only).
- Daniel M. McCarthy and Russell S. Winer. The Pareto rule in marketing revisited: is it 80/20 or 70/20? Marketing Letters, 30(2), 139-150, 2019 (abstract only).
- Byron Sharp and Charles Graham. The value of Pareto’s bottom 80%. Ehrenberg-Bass Institute for Marketing Science, University of South Australia; undated institute article, a research summary based on Kantar data.
- David C. Schmittlein, Lee G. Cooper and Donald G. Morrison. Truth in Concentration in the Land of (80/20) Laws. Marketing Science, 12(2), 167-183, 1993 (abstract only).
- Thomas J. Ostrand, Elaine J. Weyuker and Robert M. Bell. Predicting the location and number of faults in large software systems. IEEE Transactions on Software Engineering, 31(4), 340-355, 2005 (abstract only).
- Norman E. Fenton and Niclas Ohlsson. Quantitative analysis of faults and failures in a complex software system. IEEE Transactions on Software Engineering, 26(8), 797-814, 2000 (abstract only).
- Per O. Seglen. Why the impact factor of journals should not be used for evaluating research. BMJ, 314(7079), 497-502, 1997 (archive copy of the eBMJ full text, and the published abstract).
- R. C. Peck, R. S. McBride and R. S. Coppin. The distribution and prediction of driver accident frequencies. Accident Analysis and Prevention, 2(4), 243-299, 1971 (abstract only, TRID record).
Note: Tables 2, 3 and 7 are CEOtudent analysis; Tables 4, 5, 6 and the “When does 80/20 mislead?” list are a CEOtudent editorial framework; the values in Table 5 are hypothetical. Table 1 was transcribed by hand from the scanned page image of Pareto’s 1897 edition because of garbled OCR. Removed legacy claims: Italian land, traffic accidents, the 86%/8% wealth figure, the 10 richest people, time spent with friends and the “80% of problems” promise (Table 7). Because the 2019 “60/20” working paper by Sharp, Romaniuk and Graham could not be opened, only the values in the Ehrenberg-Bass summary article were used; because the UBS Global Wealth Report could not be accessed, no UBS figures were used.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
This post is also available in:
){kind=link}
&description=&image=https://ceotudent.com/wp-content/uploads/2021/07/is2kvf.jpg){kind=link}