Pareto Chart Pro answers one question well: how much of the result comes from how few of the entities? This page covers how to frame that question for four common business scenarios, how to scale to large models, and the mistakes that make a Pareto chart misleading.
Question: how exposed are we if our largest customers leave?
Setup: Entity = Customer, Value = Revenue (12-month rolling),
Tooltips = Gross Margin, Order Count. Reference line 1 at 80%.
Read it like this: the earlier the cumulative line crosses 80%, the more concentrated — and the more fragile — the revenue base. If the first bin alone (top 20% of customers at the Free bin size) already carries ~80%, that is a classic Pareto distribution. If it crosses only at the fourth bin, revenue is unusually spread out, which is a resilience story rather than a risk story.
Pair it with: a table of the actual customer names on the same page. Click the first bin to cross-filter — the table then shows exactly who sits inside that share. Binning gives you the shape; the table gives you the names.
Question: how much of the catalogue can we retire without losing meaningful sales?
Setup: Entity = SKU, Value = Units sold,
Tooltips = Revenue, Stock value, Days of cover. Reference line 1 at 95%
(not 80% — the decision threshold here is "what can we lose", not "what drives the business").
Read it like this: the share of entities to the right of the 95% line is your candidate list for delisting. Set line 2 at 99% to see the true long tail.
Watch out: units and revenue rank SKUs differently. A cheap high-volume SKU dominates a units Pareto and disappears in a revenue one. Build both, side by side — the SKUs that are in the tail of both are the safe ones to cut.
Question: which few causes explain most of the failures?
Setup: Entity = Defect code (or incident category),
Value = Incident count, Tooltips = Downtime hours, Cost of rework.
Reference line 1 at 80%. Use a Pro bin size of 5% if you have enough distinct defect codes to justify it.
Read it like this: this is the original Juran use case. The bins left of the 80% line are where corrective action buys the most. Adding cost as a tooltip measure often reverses the priority order — the most frequent defect is frequently not the most expensive one.
Pair it with: a drilldown hierarchy of Defect family → Defect code, so you can see whether concentration lives at the family level or only inside one family.
Question: is negotiating attention going where the money is?
Setup: Entity = Supplier, Value = Spend,
Tooltips = Contract count, Payment terms, On-time delivery %. Reference lines at 50% and 80%.
Read it like this: the entity share at the 50% line is your strategic supplier count — typically a handful. The gap between the 50% and 80% lines is the tactical middle, which is usually where consolidation savings hide.
Use outlier exclusion (Pro): if one framework agreement dominates spend and is negotiated separately, exclude the top 1–2% to see the structure of the remaining base. Always say so in the visual title when you do — see the pitfall below.
Drag the raw entity key (customer_id, product_id, SKU) straight into
Entity. The visual handles ranking, binning, cumulative percentages and outlier trimming itself,
and it recalculates on every update from data Power BI has already filtered — so the Pareto is of whatever
your slicers currently select. Do not pre-sort or pre-aggregate; you will only fight the visual's internal sort.
Bind a numeric key where you have one. An integer costs far less per value than a long text code, both when streaming toward Power BI's 100 MB ceiling and when cross-filtering.
Clicking a bar asks Power BI to filter the report by every entity behind it, carried as a list of values. Past roughly 10,000 values that query becomes too heavy, so bin size decides whether a model of a given size stays clickable:
maximum bin size % = 1,000,000 ÷ number of entities
At 500,000 entities that means 2% bins — 50 bars. The full measured table, for imported models and live connections separately, is in the documentation.
Free bin size (20%, 5 bars) · reference line 1 at 80% · value labels off · dots on · X and Y axis labels on. Five bars read instantly at a glance and survive being shrunk into a dashboard grid.
Pro bin size 5% (20 bars) · reference lines at 50 / 80 / 95% · value labels on, 10pt · bar gap 2px. Give it at least half a report page — twenty bars plus labels need horizontal room.
Turn on IBCS Mode. Neutral charcoal bars stop the chart from overstating drama, and the cumulative line and reference lines carry the message. Pair it with a plain-language title that states the conclusion — "38% of customers generate 80% of revenue" — rather than describing the chart.
Excluding the top or bottom % is a legitimate analytical move and a misleading one if the audience does not know. Put it in the title or a text box: "excludes top 1% (national accounts)". A Pareto chart with a hidden trim looks identical to one without.
A bar is a share of ranked entities, not one customer or one SKU. "The first bar is 42%" means the top 20% of entities carry 42% of the total — not that one entity does. If the audience needs names, put a table next to the chart.
Pareto logic assumes contributions that add toward a total. Returns, credit notes and reversals break the cumulative line. Filter them out, or use an absolute-value measure, and decide deliberately which one answers your question.
With 12 suppliers and a 1% bin size there is nothing to bin. The visual adapts the bin count automatically, but the honest answer below roughly 30–40 entities is usually a plain sorted bar chart with a cumulative line — at that size the audience can read the names directly.
80/20 is a heuristic, not a law. Plenty of real distributions are 90/10 or 60/40. The value of the chart is finding out which your business actually is — so set the reference line to the threshold your decision depends on, then read the entity share off the X axis.