Pareto Principle

Thinking Tools4.1 Root Causes and the Case for Going BackAlso known as 80/20 Rule, Pareto Analysis

Named by Joseph M. Juran (Quality Control Handbook, 1951) after Vilfredo Pareto's income distributions (Cours d'économie politique, 1896 to 1897).

Ranks causes, customers, or problems by their share of an outcome to find the few that drive most of it, often near an 80/20 split.

Three of eight complaint categories cause four in five complaints

An internet provider's 1,200 complaints a month in eight categories, each bar a category's share, largest first (Bills are billing errors, Setup missed installations, Hold long waits on the phone). The running total climbs steeply and flattens after the third bar. Slide how many categories to fix, largest first: the marked line follows the running total up to the cutoff and goes flat after it, and the gap above it is what is left. Illustrative.

3
Categories fixed
3 of 8
Complaints those categories cause
80%
Complaints a month still coming in
240

Each category's share, percentRemoved by the categories fixedRunning total, every category fixed

025507510045251075422BillsSetupSpeedHoldRouterStaffOutageOther025507510045251075422BillsSetupSpeedHoldRouterStaffOutageOther

An example

The call center of a regional internet provider logged about 1,200 complaints a month in eight categories, and the operations director had budget for two improvement projects that quarter. Her team pulled six months of tickets and counted them by cause. Ranked largest first, billing errors made up 45 percent, missed installation appointments 25 percent, and slow speeds 10 percent; the other five categories came to 20 percent between them. The running total climbed steeply through the first three bars and flattened after them.

She funded two projects: a fix to the proration rule behind most billing errors, and a text the evening before each installation confirming the window, with a one-tap reschedule. Slow speeds went to the network team's list. Within three months complaints fell to about 560 a month. The team reran the ranking and found the order had shifted: slow speeds, third before, now topped the list, and became the next quarter's project.

Why it works

Causes are rarely equal. In most operations a few sources produce most of the effect, because a flaw in a rule or a process fires every time the rule runs, while a rare fault touches only the cases it happens to reach. Ranking makes that inequality visible, and the running total turns it into a decision: where the line flattens, each further category buys less.

The ranking also protects the budget from the loudest complaint, which is often not the largest. And because the count is redone after each fix, the method keeps up with a moving target. Fixing the top category changes the shape of what is left, and a category that looked minor can rise to the top once the giants are gone. The rest of the list still gets attention, in order.

The thinking behind it

Vilfredo Pareto Vilfredo Pareto, an Italian economist of the Lausanne school, studied how income was distributed across many countries in his Cours d'économie politique (1896 to 1897). He found a regular pattern, a power law, in which a small share of people held a large share of the income and the land. His subject was income and land; the principle carries his name because of what Joseph Juran later made of the observation. Economists also know him for Pareto efficiency, a separate idea.

Joseph M. Juran Joseph M. Juran, a pioneer of quality management, generalized Pareto's observation in his Quality Control Handbook (1951), which became the field's standard reference. In his terms a small share of causes accounts for most of the defects: the vital few and the trivial many, a phrase he later changed to the useful many, since the remaining causes still matter. He went on to frame quality management as a trilogy of planning, control, and improvement.

How to run it

  1. Pick the outcome and its unit. Choose one measure, such as complaints a month or hours of downtime, over a period long enough to smooth out a bad week.
  2. Sort every item into categories. Use categories that point at something someone can fix, and keep the catchall pile under a tenth of the total.
  3. Rank the categories by size. Count each one and order them from largest to smallest, with the catchall last.
  4. Add the running total. Plot each category's share as a bar and the cumulative share as a line over the bars.
  5. Mark where the line flattens. The categories before the bend are the few that drive most of the outcome, whether the split is 80/20, 70/30, or 90/10.
  6. Work the top categories first. Fund fixes for the largest one or two and keep the rest on a dated list.
  7. Count again after each fix. The order shifts once the leaders shrink, so rerun the ranking before choosing the next project.

Where it goes wrong

The wrong measure ranks the wrong things. A count treats a billing error and a lost business account as equal, though one costs far more; when the costliest problem sits low on the chart, rank by cost or by customers lost.

The second is categories too broad to act on. A bar called service quality tells no one what to change. Split it until each category names a fix and an owner.

The third is treating 80/20 as a law. Teams stop at the top fifth of the categories whatever the data says, or claim a split that is not there. Read the bend from the running total, and when the bars are nearly flat, accept that no few causes dominate and look for a fix to the whole process.

When to use something else

  • Fishbone Diagram. When one large category needs its own causes traced before anyone can fix it.
  • Five Whys. When the top category is clear and the team needs to follow one instance of it down to its root.
  • Power-Law Reasoning. When the question is whether an outcome is heavy-tailed at all, such as revenue across customers, before acting on its top few.

Where it sits in SOLVE eX

Go to the source

  • Cours d'économie politique, F. Rouge, Lausanne, two volumes, 1896 to 1897, 1896
  • Quality Control Handbook, McGraw-Hill, 1951

Back to the catalog