Problem solving Decide

Pareto Analysis

Pareto analysis ranks causes by size to find the few that account for most of the effect. The 80/20 rule, applied to data.

The questionWhich few causes account for most of this effect, and which of those can we actually change?

Also called the 80/20 rule, the Pareto principle, or the vital few and the useful many.

Model · Vilfredo Pareto and Joseph M. Juran, 1896 to 1951

Short shipments

44 percent
$1.06 million

Damage in transit

29 percent
$696,000

Pricing errors

11 percent
$264,000

The rest

16 percent
Nine small codes

Share of $2.4 million in customer credits, ranked by cause. The first two columns are the vital few.

When to use it

You have counts, cost, or time for a long list of causes, products, customers, or defects, and you need to choose where the next effort goes.
A program is treating every cause as equal, so scarce fixing time is spread across items that differ by an order of magnitude.
A fishbone or an issue tree has produced many branches, and you need a size ranking before walking five whys on all of them.
You need a one-page picture for a steering committee of which few items make up most of a loss, a delay, or a complaint volume.

When to avoid it

You have no measured effect, only opinions about which causes matter. Ranking the opinions does not create a Pareto.
The few largest items cannot be changed in this engagement, and the next items can. Rank by size times tractability, or you will brief a vital few you cannot touch.
The categories are mixed units, such as mixing count of defects with cost of defects. Re-rank on the unit the decision is about.
The question is whether a product or customer is profitable after overhead. Activity-based costing answers that. A Pareto of revenue will highlight volume, not margin.

Overview

Pareto analysis ranks the causes of a measured effect from largest to smallest, so a team can put effort on the few that account for most of the result. The 80/20 rule is a memory aid for that concentration, not a ratio the data must match. Juran’s useful distinction is between the vital few, which deserve cause analysis and a project, and the useful many, which still need a standard way of working.

The ranking is only as good as the categories and the unit. Overlapping codes split a large cause into medium ones. Ranking by count when the decision is about cost promotes the wrong items. Once the few are marked, split the largest one by site, shift, or product, and then find the cause. A chart without that next step is a sorted list.

The ranking

The effect

One measured outcome for a stated period, such as dollars of credits, hours of downtime, or number of complaints. Changing the unit changes which items look vital, so pick the unit that matches the decision.

The categories

A MECE set of causes or items that add to the effect. If two codes can describe the same event, the ranking will split a large cause into two medium ones and hide it.

The vital few

The first items on the ranked list that together account for most of the effect. They are the ones worth a cause analysis and a fix. Juran's point was concentration, not a literal 80 percent.

The useful many

The long tail. Each item is small. Together they may still be a fifth or more of the effect. Do not ignore them. Standardize, automate, or batch them, and do not give them the same project attention as the vital few.

What it is not

A law of nature that every list will split 80/20. That ratio is a memory aid. Compute the actual cumulative share.
A reason to ignore the tail. Juran renamed it the useful many because a thousand small leaks still have to be run as a process.

The steps

Fix the effect and the period

Name the outcome in one unit and the months it covers. If last quarter is a peak season, say so. What comes out is a total that later rows must add to.
Build a MECE list of causes

Use the codes the operation already records, then check for overlap and for an "other" bucket that is quietly large. Split "other" before you rank, or the vital few will be incomplete.
Rank and cumulate

Sort from largest to smallest. Write each item's share and the running total. Mark the point at which you have most of the effect. That cut is a judgment, not a rule that it must be 80 percent.
Split the largest item once

Take the top cause and break it by site, shift, product, or customer. The vital few often hide inside one code. Stop when the next split would not change the action.
Choose the few you will fix

Keep the items that are large and that this team can change. Give each an owner and a measure. Put the tail under a standard process, not a project.

In practice

A distributor was issuing $2.4 million a year in customer credits, across twelve reason codes. The operations director wanted a packing-standards program in every warehouse. The consultant ranked the codes by dollar value for the last twelve months. Short shipments were 44 percent, $1.06 million. Damage in transit was 29 percent, $696,000. Together, 73 percent. The remaining ten codes, including pricing errors, made up 27 percent.

Short shipments were then split by site. One of three distribution centers, on its night shift, produced 80 percent of the short-shipment dollars, about $850,000. Pick-face slotting on that shift had not been updated after a layout change six months earlier. The first action was to reslot that building and audit night-shift picks for thirty days, not to rewrite packing standards for the network. Damage stayed on the list as the second project, with a carrier claim rate attached. The packing program was dropped.

What you leave with

A ranked table of causes that adds to the stated effect, with shares and a cumulative line.
A marked vital few, and the reason the cut was drawn there.
One extra split on the largest item, by site, shift, or product.
Named actions for the few, and a way of working for the many.

Failure points

Ranking by count when the decision is about cost

A frequent cheap defect outranks a rare expensive one. Use the unit the sponsor is trying to move.

An "other" bar that is the tallest

The leftover category was never split, so the chart says the main cause is unclassified work. Break it before you present the ranking.

Treating 80/20 as a target

The team massages categories until the chart looks like the textbook. The real concentration was 62/12, which was already enough to choose.

Vital few that cannot be touched

The largest cause is a contract, a regulation, or a supplier the client will not change. The ranking is correct and the plan still has nothing to do. Score tractability next to size.

Stopping at the chart

The bars are presented as the finding. Nobody asks why the largest bar is large, so the next action is "focus on X" with no change to the process that produces X.

The failures above are usually Framing: the wrong problem was chosen before this framework was applied.

Origin

Vilfredo Pareto, an Italian economist, showed in Cours d'économie politique (1896 to 1897) that a large share of Italy's land was held by a small share of the population, often remembered as roughly 80 percent by 20 percent. The pattern is a power-law distribution. Pareto was writing about wealth, not defects.

Joseph M. Juran applied the same imbalance to quality. He later wrote that he encountered Pareto's work around 1941 and named the "Pareto principle" in the early 1950s, including in the Quality Control Handbook. He also coined "the vital few and the trivial many," then changed trivial to useful, because the long tail still has to be managed. In a 1974 essay, "The Non-Pareto Principle; Mea Culpa," he said the name was a misattribution: Pareto had not claimed a universal, and the cumulative curve Juran used was closer to Max Lorenz. The 80/20 split is a shorthand, not a law. Real data may run 70/30 or 90/10, and the two numbers need not add to 100. There is no trademark on the principle or the chart.