Pick a problem with a countable outcome
Pareto works on counts or magnitudes: defects, complaints, costs, time-spent, dollars-lost. Vague qualitative problems don't Pareto cleanly.
The 80/20 principle, made practical — when it applies, how to build a Pareto chart that actually points to action, the five most common misuses, and how AI agents run continuous Pareto across every dimension of your data at once.
Vilfredo Pareto noticed in 1896 that 80% of Italy was owned by 20% of the population. Joseph Juran carried the idea into quality management in the 1940s with a memorable phrase: 'the vital few and the trivial many.' Most defects, complaints, costs, and incidents concentrate. A small number of categories drive most of the pain.
The Pareto chart is the visual that makes this actionable: a bar chart sorted by frequency descending, with a cumulative-percent line that shows where the natural break between the vital few and the trivial many sits. It's one of Ishikawa's seven basic tools of quality and appears in the Measure phase of every well-run DMAIC project.
The 80/20 ratio is shorthand. The principle is concentration; the specific ratio varies.
Pareto works on counts or magnitudes: defects, complaints, costs, time-spent, dollars-lost. Vague qualitative problems don't Pareto cleanly.
The categories you choose determine what the chart can tell you. Defects by part number, by shift, by supplier, by failure mode — pick the dimension where action is possible. Run multiple Paretos if you're unsure which dimension matters.
Count occurrences in each category. Sort descending. Add a cumulative-percent line on a secondary axis. The drop-off should reveal the vital few.
The 80/20 ratio is the popularized version; the real pattern is 'a few categories dominate.' The elbow on the cumulative line is where to draw the line between the vital few and the trivial many.
A category like 'Defect Type A' is a starting point, not a solution. Apply a fishbone or 5 Whys to each top category to find the cause worth fixing.
Once you've reduced the top category, the second category often becomes the new top. Pareto isn't a one-shot tool — it's a sequence.
If you Pareto by defect type and the real driver is shift, the chart hides the truth. Try several dimensions — by part, by operator, by time of day, by supplier — before committing.
The most common defect is rarely the most expensive. Run Pareto twice: once by count, once by cost or impact. The vital few differ.
Pareto says 'focus here first,' not 'ignore the rest.' Categories that are individually small can collectively be larger than the top one, and some of them may be high-severity (think AML false-negatives).
Process drift moves the chart. A Pareto from January won't match July. Treating the Pareto as a fixed artifact is how teams keep working on yesterday's vital few.
Sometimes it's 70/30, sometimes 90/10. Force-fitting the ratio is theater. The principle is concentration of impact, not a specific number.
Pareto defects by failure mode to find the dominant cause of scrap. Re-Pareto by line, shift, and supplier to localize root cause.
Pareto false-positive alerts by detection rule to find which rules generate the most noise. Tune or retire the top contributors.
Pareto tickets by contact reason to find the top deflection opportunities. Re-Pareto by handle time to find the cost-weighted top.
Pareto incidents by service to find which systems generate the most pages. Re-Pareto by user impact to weight by actual customer pain.
Every dimension. Continuously. Cost-weighted.
Agents Pareto the data across every plausible dimension — failure mode, time, location, operator, supplier, customer segment — in parallel. The dimension that reveals the strongest concentration is the one the team should act on.
The Pareto chart updates in real time as the process runs. Drift is visible the day it starts, not the next quarterly review.
Agents compute cost-weighted Pareto by attaching the financial impact of each category — handle time, dollar loss, regulatory fine exposure — automatically.
When the obvious Pareto is unhelpful (flat, no concentration), the agent proposes the next dimension to try based on what variables correlate with the outcome.
The Pareto principle is named after the Italian economist Vilfredo Pareto, who observed in 1896 that 80% of Italy's land was owned by 20% of its population. Joseph Juran adapted the idea to quality management in the 1940s, coining the phrase 'the vital few and the trivial many' to describe how a small number of causes typically drive most defects. The 80/20 ratio is shorthand for that concentration — the actual ratio varies (70/30, 90/10) but the principle that impact concentrates is the durable insight.
A Pareto chart is a bar chart sorted descending by frequency or magnitude, with a cumulative-percent line overlaid on a secondary axis. The two things together make it diagnostic: the bars show which categories dominate, the cumulative line shows where the natural cutoff between vital few and trivial many sits. A plain unsorted bar chart loses both insights.
Pareto picks the target — which category of failure deserves attention. Fishbone enumerates the possible causes of that target. FMEA scores and prioritizes failure modes by severity, occurrence, and detection. In a typical DMAIC project they appear in sequence: Pareto in Measure (to choose the focus), fishbone in Analyze (to map causes), FMEA in Analyze and Control (to prioritize risk and design protections).
No. The principle of concentration usually holds — a small number of categories usually drive most impact — but the specific ratio varies. Some processes show 90/10 (heavy concentration), others 60/40 (modest concentration), occasionally 50/50 (no concentration, every category matters). When the Pareto is flat, that's itself a finding: the problem doesn't have a vital few, and a different improvement strategy is needed.
Two shifts. First, agents run the analysis across many dimensions simultaneously rather than the team guessing which dimension to chart. Second, the chart stays live — drift shows up the day it starts, not next quarter. Cost-weighting and severity-weighting also become trivial, since the agent already has the data. The team's role moves from charting to deciding which top-category to act on.
Bring a defect, an alert volume, or a queue. We'll Pareto your event data across every plausible dimension and show you where the real concentration is — and what it costs.