What it means
Building one means choosing ranges, called bins, and counting how many observations land in each. Payment delays might be grouped into 0 to 10 days, 11 to 20 days and so on, with the bar height showing how many invoices fell in each band.
The reason it matters in business is that averages conceal almost everything interesting. An average payment time of 27 days sounds fine until a histogram reveals a cluster paying within a week and a stubborn tail paying past 50 days, which are two entirely different problems needing two different responses.
Shape is the thing to read. A single central hump suggests one consistent process, two humps suggest two different processes mixed together, and a long tail to the right warns that occasional extreme values will dominate your cash flow or your costs.
Bin width is the main judgement call and it changes the story. Too few bins and everything flattens into a featureless block, too many and normal randomness looks like meaningful structure, so a sensible starting point is somewhere between five and fifteen bins for a few hundred observations.
Histograms sit at the heart of quality control, credit risk and operational reviews for the same reason: they turn a column of numbers nobody reads into a picture people argue about productively. They are also the natural companion to standard deviation, since the chart shows the spread the statistic only summarises.
In practice
Real-world examples
Example 1
A manufacturer plots 500 measurements of a machined part and sees two separate humps rather than one. Investigation shows two machines are producing to slightly different settings, a difference the average measurement had masked entirely.
Example 2
A subscription business charts customer lifetime in months and finds a large spike at month one alongside a broad group lasting over two years. The marketing team stops treating churn as a single problem and builds a dedicated first-month onboarding programme.
Example 3
A lender plots the credit scores of applicants approved last quarter and finds an unexpected cluster just above the cut-off. The risk team reviews whether scores are being nudged over the line during manual overrides.
Think of it
Imagine sorting a giant pile of loose change into separate jars based on coin values, then lining the jars up by height to see instantly which coin is most common.
Formula
Calculation
There is no single formula, but two calculations do the work: bin width = (largest value - smallest value) / number of bins, and relative frequency = count in bin / total observations.
A finance team reviews 60 invoices and records the days taken to pay, ranging from 3 days to 58 days. Aiming for six bins, bin width = (58 - 3) / 6 = 9.17 days, which is rounded to a tidy 10 days.
The counts come out as follows: 0 to 10 days, 8 invoices; 11 to 20 days, 15; 21 to 30 days, 22; 31 to 40 days, 9; 41 to 50 days, 4; 51 to 60 days, 2. The total is 8 + 15 + 22 + 9 + 4 + 2 = 60, which confirms nothing has been missed.
The tallest bar is 21 to 30 days, holding 22 / 60 = 36.7% of invoices. More usefully, invoices paid after 30 days total 9 + 4 + 2 = 15, which is 15 / 60 = 25% of the book. On payment terms of 30 days, that single number says a quarter of customers are late, which the average payment time of 27 days completely hid.Case study
Seen in the real world
Fernbrook Interiors is an invented furniture retailer used for this illustrative example. Its finance director reports average debtor days of 27 against 30-day terms and treats collections as healthy, while the cash flow forecast keeps missing by five figures each month.
An analyst plots the last 60 invoices as a histogram. The bars show 8 invoices settled within 10 days, 15 within 20, 22 within 30, then a tail of 9, 4 and 2 running out to 58 days. A quarter of invoices are paid late, and because the late payers happen to be the largest accounts, they represent a much bigger share of the money than of the invoice count.
Fernbrook changes tack. Rather than chasing everyone, it puts the 15 late-paying accounts on a weekly call cycle and offers a 1% discount for payment within 10 days to the biggest three. In this illustrative outcome average debtor days barely move, falling from 27 to 25, but the tail beyond 40 days almost disappears and the forecast starts landing within 3% of actual cash. The moral of the fictional case is that the average was never the problem; the shape was.
Watch out
Common mistakes
- Confusing a histogram with a bar chart. A histogram splits one continuous measure into ranges with touching bars, while a bar chart compares separate categories with gaps between them.
- Choosing bin widths that flatter the story. Adjusting bins until the chart shows what you hoped for is a well-worn way to mislead an audience, including yourself.
- Reporting the average without ever looking at the distribution. Two data sets with identical averages can behave completely differently, and only the shape reveals which one you are dealing with.
Questions
People also ask
How many bins should I use?
Between five and fifteen suits most business data sets, and it is worth redrawing at two or three bin widths to check the shape is stable.
Can a histogram handle small samples?
It can, but with fewer than about 30 observations the shape is unreliable, and a simple list or dot plot often communicates more honestly.
What does a long right-hand tail mean?
It means occasional very large values exist, which usually matters more for cash and risk planning than the bulk of the distribution does.