Charts

Histogram Generator

A distribution chart from a column of numbers, with the bin width chosen for you or by you.

Loading the tool…

Processing happens locally in your browser. What you paste or load is processed by this page and is not uploaded to a server. Nothing is stored unless you use a control that says it stores something, and you can clear anything this site has kept from the privacy page.

How to use this tool

  1. Paste your numbers, one per line or as a single column copied from a spreadsheet.
  2. Leave the bin count at 0 the first time and let the Freedman-Diaconis rule choose.
  3. Select Draw the histogram.
  4. Now try two or three other bin counts. A feature that survives all of them is real; one that appears at a single bin width is not.

What histogram generator does

An average tells you where a distribution sits. A histogram tells you what shape it is, which is usually the more important question: whether it has one peak or two, whether it is skewed, and whether the tail contains the handful of values that will define your p99 and everybody’s experience of the system.

Paste your numbers, one per line or as a single column, and this bins and draws them. Left to itself it picks the bin width by the Freedman-Diaconis rule, which reacts to the spread of the middle half of the data rather than to the count, so one extreme outlier does not shred the resolution. Set the bin count yourself and look again — a shape that changes with the bin width was never as clear as it looked.

Frequently asked questions

There is no single right answer, which is why the honest advice is to try several. Left alone this uses the Freedman-Diaconis rule, which sets the width from the spread of the middle half of the data and so is not thrown off by one extreme value. Then change it: any feature that only appears at one particular bin width is not a feature.

A bar chart compares separate categories, and the bars could be reordered without losing anything. A histogram shows one continuous variable cut into ranges, so the order is fixed and the width of each bar means something. Drawing categories as a histogram, or a distribution as a bar chart, misstates what kind of data you have.

Because a couple of very slow requests stretched the axis across a range almost nothing occupies. That is the shape of most latency data and it is exactly why percentiles are reported rather than averages. Chart the log of the values, or cut the extreme tail out and chart it separately, and the body of the distribution becomes readable again.

Only loosely. With a small sample the bins are noisy and a second peak can appear and vanish between one dataset and the next. A box plot is the sturdier choice at that size, because quartiles move far less than bin counts do — and the summary here gives you the mean and median to compare regardless.