Histogram card

The Histogram card groups a device's readings into ranges and shows how many fell into each one. Where a Graph card answers "what happened and when", a histogram answers "how often, and at what magnitude".

Use it when

Frequency matters more than sequence. How many events exceeded a threshold, how readings are distributed across a range, whether activity clusters at one end or spreads evenly.

It is particularly useful for event-driven measurements, where a time series is mostly empty with occasional spikes and tells you little at a glance. The same data as a distribution shows immediately where the activity sits.

Configure it

Count or percentage

Count gives the number of readings in each bin. Percentage gives each bin's share of the total.

Count is the better default when the absolute number matters — how many events exceeded a level, how many readings you have at all. A bin holding three readings and a bin holding three hundred look very different, and percentage hides that.

Percentage is for comparing distributions rather than volumes. Two cards showing different devices, or the same device across different time ranges, are only comparable as percentages if the two sets contain different numbers of readings. In count mode the busier period simply has taller bars everywhere, which tells you about activity rather than shape.

Note: Percentage is a share of the readings in the dashboard's time range, not of some fixed total. Change the range and the percentages change, even though the underlying data has not.

Bin width is the whole card

Everything about how a histogram reads comes down to bin width, and it is worth trying two or three before settling.

Too narrow and every bar is one or two counts — you are looking at individual readings again, which is what the Graph card is for. Too wide and everything collapses into two or three bars that hide the shape entirely.

A reasonable starting point is a width that gives you somewhere between eight and twenty populated bars across the range you care about. Adjust from there based on whether the distribution has structure worth seeing.

Note that changing bin width can change the story. A distribution that looks bimodal at one width can look flat at another, so if a shape is going to inform a decision, confirm it holds at more than one setting.

Reading it

The card summarises the dashboard's time range as a whole, so it shows no sequence. A histogram of a week cannot tell you whether the large readings came on one day or were spread evenly across seven.

That makes it a good companion to a Graph card of the same metric rather than a replacement. The graph tells you when; the histogram tells you how often.