Operations & process
Pareto Analysis
Juran's application of Pareto's 80/20 observation to operational problems. Categorise the defects or losses, rank them by frequency or cost, cumulate the percentages, and concentrate effort on the vital few causes that drive most of the effect, without abandoning the useful many.
Also known as 80/20 rule, Pareto principle, Vital few analysis, ABC analysis. First set out by Joseph M. Juran, after Vilfredo Pareto in 1951; the primary source is cited in full below.
Where this is contested
The naming is a historical accident: Pareto observed concentration in wealth, and it was Juran who generalised it to quality and attached Pareto's name, a mistake Juran publicly conceded in 'The Non-Pareto Principle; Mea Culpa' (1975), where he also softened 'the trivial many' to 'the useful many'.
- Format
- Scoring model
- Level
- Business unit · Team
- Best for
- Prioritise · Allocate resources
- Decision stage
- Diagnose · Review
- Difficulty
- Introductory
- Time to apply
- A few hours with clean data; allow days where the category coding needs rebuilding first, which it usually does.
Plate · The model
The components
Collect and categorise
Assemble the raw incidents or costs for a defined period and classify each into one category of an agreed scheme. The quality of the entire analysis is set here, by the discipline of the coding.
Signals of strength
Categories are mutually exclusive and add up to the whole · The 'other' bucket is under roughly 10% of the total · One defined counting period and counting rule throughout · Free-text reasons re-coded rather than dumped into 'miscellaneous'
Rank by frequency or cost
Sort the categories descending by the chosen measure. Running the ranking both ways, by count and by cost, exposes the cases where the most common problem is far from the most expensive one.
Signals of strength
A deliberate choice of measure, argued rather than defaulted · Both frequency and cost rankings produced and compared · The ranking stable when the period is split in half · Severity or harm considered where money understates it
Cumulate percentages
Express each category as a share of the total and accumulate down the ranking to reveal how concentrated the effect is. The cumulative line is the analytical content of a Pareto chart; the bars alone are just a sorted histogram.
Signals of strength
A cumulative line plotted over the sorted bars · The concentration read from the curve, with no forcing towards 80/20 · Percentages of a stated, reconcilable total · A flat, undifferentiated curve recognised as a real and reportable result
Separate the vital few
Divide the ranked categories where the curve flattens: the vital few get focused problem-solving and named owners, the useful many get routine measures. Juran's own revision of 'trivial many' to 'useful many' is part of the method, a warning against writing off the tail.
Signals of strength
A cut-off justified by the curve, and never by the slogan · Named owners and actions against each of the vital few · An explicit, proportionate plan for the useful many · A date to re-run the analysis after action, since the ranking moves
When it earns its keep
- You face many competing problems, defect types, complaint reasons or cost lines, and need a defensible basis for tackling some before others.
- An improvement effort is spreading itself evenly across every issue. Effort allocated uniformly across causes of unequal size is misallocated by definition.
- You are in the Analyse phase of a DMAIC project and need to know which verified causes carry the weight of the defect.
- You are reviewing whether past effort landed where the losses actually were: a before-and-after Pareto chart is one of the simplest honest progress reports in operations.
And when it doesn't
- The categories are all roughly the same size. Some distributions are genuinely flat, and forcing a 'vital few' onto them manufactures a priority that is not there.
- Frequency is a poor proxy for importance and you cannot yet weight by cost or harm. A rare defect that injures a customer outranks a common cosmetic one; rank on the wrong measure and the chart will confidently mislead.
- The long tail is the point. In some settings, product range economics, safety near-misses, rare severe failures, the aggregate of small categories is where the value or the risk lives.
- You need causal understanding rather than ranking. A Pareto chart says which category is biggest; it says nothing about why. Pair it with a fishbone diagram or Five Whys.
How to run it
Before starting, gather the inputs the analysis depends on:
- A defined effect to analyse, defects, returns, complaints, downtime, cost, over a stated period.
- A category scheme that is mutually exclusive and reasonably complete, with an 'other' bucket small enough not to embarrass the analysis.
- Counts per category, and ideally a cost or impact weighting per category, since frequency and cost often rank differently.
- Enough data volume for the ranking to be stable rather than an artefact of a quiet fortnight.
- 1
Collect and categorise
Gather the incidents or losses for a defined period and assign each to one category. This is where most Pareto analyses are silently won or lost: vague codes, overlapping categories and a bloated 'other' bucket will surface as a chart that ranks data-entry habits rather than problems.
- 2
Rank by frequency or cost
Count each category and sort descending. Then do it again weighted by cost or harm. Where the two rankings disagree, the cost ranking usually deserves the decision, and the disagreement itself is a finding.
- 3
Cumulate percentages
Convert counts to percentages of the total and accumulate down the ranking, the classic Pareto chart overlays this cumulative line on the sorted bars. The curve's shape tells you how concentrated the problem is; a steep early rise means a few categories dominate.
- 4
Separate the vital few
Draw the line where the cumulative curve flattens, wherever that falls. The point is the concentration, never the specific numbers: 70/25 and 90/10 are both Pareto patterns. Name the vital few, assign them owners and effort, and give the useful many routine controls rather than nothing.
Reading the result
A ranked, cumulated view of where a problem actually concentrates: a short list of vital-few categories that justify focused effort, an explicit judgement about the useful many, and a baseline chart to re-run after action to prove the mix has shifted.
- Read the cumulative curve, then the bars. A steep early curve licenses concentration of effort; a shallow one tells you this problem does not have a vital few, which is equally worth knowing.
- Check which measure the ranking uses before trusting it. A chart ranked by count answers 'what happens most', ranked by cost it answers 'what hurts most', and they are frequently different charts.
- Treat the 80/20 numbers as a mnemonic for concentration, never as a law. Real distributions land anywhere from mild to extreme concentration, and the shares need not sum to 100.
- A successful intervention shows up as a reshuffled ranking on the next run. If the same category tops the chart quarter after quarter, the analysis is being admired rather than used.
A worked example
An online electronics retailer ranks the causes of product returns
A UK online electronics retailer selling audio gear, monitors and small appliances is running an 11.2% return rate, well above its 8% plan, and the returns operation is absorbing margin through refunds, inspection labour and write-downs. The operations lead pulls 90 days of data, 3,412 returns, and runs a Pareto analysis by both count and net cost before the next trading review.
- Collect and categorise
- The twelve return-reason codes prove unreliable: 38% of returns sit in 'other' because the codes are awkward to select at the returns desk. A week re-coding free-text comments and inspection notes cuts 'other' to 7% and adds two categories the original scheme lacked, 'item not as described' and 'damaged in transit'.
- Rank by frequency or cost
- By count: 'item not as described' 24%, 'changed mind' 21%, 'faulty on arrival' 19%, 'damaged in transit' 9%, then eight smaller reasons. By net cost the order changes sharply: 'damaged in transit' jumps to second at 27% of return cost, because damaged monitors are write-offs, while 'changed mind' items restock at near-full value.
- Cumulate percentages
- On the cost ranking, the top three categories, faulty on arrival, damaged in transit, item not as described, cumulate to 71% of the 90-day return cost of 214,000 pounds. The curve then flattens hard: the remaining nine categories average under 4% each. Inspection notes tie most 'faulty on arrival' cost to a single supplier's wireless earbuds batch.
- Separate the vital few
- Vital few, with owners: the earbuds supplier batch (buying lead, claim and delisting review), courier packaging for monitors over 27 inches (despatch manager, packaging trial), and spec-page accuracy on audio products, where 'not as described' concentrates (ecommerce manager, template rewrite with connectivity compatibility tables). The useful many, including 'changed mind' spread thinly across hundreds of SKUs, get routine measures: clearer size imagery and no project effort.
The read. The cost-weighted ranking, and only the cost-weighted ranking, tells the truth: had the team ranked by count alone, 'changed mind' would have earned a project it did not deserve while transit damage on monitors kept quietly writing off stock. Three focused actions cover roughly 71% of return cost. The honest caveats are recorded on the chart itself: the earbuds batch is a one-off that will wash out of the next quarter's ranking, and the 'useful many' tail still carries 29% of cost that routine controls, not neglect, must hold.
Pitfalls
- Ranking on the measure that is easiest to count. Frequency charts are cheap; cost and harm charts change decisions. Where they disagree, the frequency chart is usually the wrong one to act on.
- Garbage categories: overlapping codes, a giant 'other', or coding habits that vary by person or shift. The chart then ranks the recording process rather than the problem.
- Forcing the data to say 80/20. The principle is about concentration in general; announcing 'the top three are our 80%' when they are 55% is fiction with a chart attached.
- Writing off the tail. Juran corrected 'trivial many' to 'useful many' precisely because the small categories collectively matter and can contain rare, severe risks that frequency ranking buries.
- Treating the ranking as causal analysis. The biggest bar identifies where to look; a fishbone diagram, Five Whys or data drill-down still has to establish why.
What the critics say
The 80/20 numbers are folk empirics. Pareto observed a specific regularity in income and land data; the universalised 80/20 rule is a heuristic that gets stated as a law, misused as a forecast, and often 'confirmed' by analysts who stop counting when the ratio appears. The principle's serious content is only that effects tend to be concentrated, in proportions that must be measured case by case.
Sanders, R. (1987) 'The Pareto Principle: Its Use and Abuse', Journal of Services Marketing, 1(2), pp. 37-40.
Vital-few thinking can systematically undervalue the long tail. Anderson's work on tail economics showed that low-frequency items in aggregate can rival or exceed the head, and in operations the same logic applies to risk: rare, severe events sit at the bottom of a frequency-ranked Pareto chart, which is exactly the wrong place to ignore.
Anderson, C. (2006) The Long Tail: Why the Future of Business Is Selling Less of More. New York: Hyperion.
Juran himself flagged the framework's most common abuse. In 'The Non-Pareto Principle; Mea Culpa' he conceded that attaching Pareto's name to the principle was historically mistaken, and his replacement of 'trivial many' with 'useful many' was a substantive correction: the tail is to be managed, never dismissed.
Juran, J. M. (1975) 'The Non-Pareto Principle; Mea Culpa', Quality Progress, 8(5), pp. 8-9.
Work it through
List the categories with their count and, where you have it, their cost. Choose which measure to rank by; the cumulative percentages and the vital-few cut compute below. Where the count ranking and the cost ranking disagree, the cost ranking usually deserves the decision, and the disagreement is itself a finding.
| Category | Count | Cost (£) | Share | Cumulative |
|---|
Sources and further reading
- Juran, J. M. (ed.) (1951) Quality Control Handbook, 1st edn. New York: McGraw-Hill.
- Pareto, V. (1896-1897) Cours d'economie politique, 2 vols. Lausanne: F. Rouge. ↗
- Juran, J. M. (1975) 'The Non-Pareto Principle; Mea Culpa', Quality Progress, 8(5), pp. 8-9. ↗
- American Society for Quality, 'What is a Pareto Chart?'. ↗