A measure is what a number means, written down once. It is the most important surface in Analytics and the one with the least ongoing work.
What a measure declares
What it counts. Which records.
Which arithmetic. Six choices: count, sum, average, minimum, maximum, and count of distinct values.
Which field, where the arithmetic needs one. Count works on rows alone; sum and average need a number.
How it behaves over time. This is the one that matters most, and it has its own section.
How it is displayed. Currency, percentage, plain number.
Flow and snapshot
Every measure is one or the other, and getting it wrong produces figures that look plausible and are wrong.
Flow accumulates over a period. Revenue, deals closed, tickets resolved. Asking for a month gives you that month's worth. Asking for a quarter gives you three times as much. The period is part of the question.
Snapshot is a value at a moment. Headcount, open pipeline, account balance. It does not accumulate, and adding January to February does not give you a bigger number, it gives you a nonsense one.
The classic error is applying a period filter to a snapshot and reading the result as a total. Declaring the class on the measure is what stops the engine doing that, and it is why the field exists rather than being inferred.
If you are unsure, ask whether summing two months makes sense. If it does, it is flow. If it does not, it is snapshot.
Platform and organisation measures
Platform measures ship with NearSync. They cover the things every business counts, and they are read-only in your workspace.
That is deliberate. A shared definition of a common number is worth more than the ability to redefine it, and a platform measure that each tenant edited would make cross-organisation benchmarking meaningless.
Organisation measures are yours. Create as many as your business genuinely needs.
Where a platform measure is close but not right for you, make your own alongside it and name it distinctly. Two measures called "Revenue" is the failure this whole surface exists to prevent.
Naming, which is most of the work
A measure's name is what somebody reads on a board six months from now with no context.
Say the qualifier. "Revenue, closed won" beats "Revenue". Somebody will assume the unqualified one means something else, and they will be right to.
Say the unit if it is not obvious. "Average deal size" is clear; "Deal value" is not.
Do not name it after the team that asked for it. "Marketing pipeline" stops making sense when marketing reorganises.
Never ship two measures whose names could be confused. If you need "Revenue, booked" and "Revenue, recognised", the names have to make the difference unmissable, because a board only shows the name.
Filters on a measure
A measure can carry its own filter, which is different from filtering a widget.
On the measure the filter is part of the definition and travels everywhere. "Revenue, closed won" is a measure whose filter is the point.
On the widget the filter is a view of a general measure. "Revenue, this quarter, north region" is one look at a measure that is otherwise unrestricted.
Put it on the measure when it is part of the meaning, and on the widget when it is part of the question. Getting this backwards produces either a proliferation of near-identical measures or a general measure nobody can safely use.
Formats
Choosing currency or percentage is not decoration. A figure without a unit gets read as whatever the reader expects, and a rate shown as 0.42 gets reported as forty-two in one conversation and nought point four in the next.
Currency measures follow the platform's currency handling, so a multi-entity organisation gets figures in the base currency rather than a meaningless sum of mixed ones.
Cohorts
A measure can be narrowed to a defined group of records rather than everything of that type.
Useful where a number only makes sense for a subset: revenue from new customers, tickets from one tier, headcount in one entity.
The discipline is the same as naming. A measure narrowed to a cohort must say so in its name, because the board will not, and a figure that silently excludes half the business is the most damaging kind of wrong.
A reasonable set
Start with under ten. The measures every conversation already uses.
One owner per area. Finance owns the money definitions, sales owns the pipeline ones. Definitions with no owner drift.
Review once a year. A measure whose underlying process changed is worse than no measure, because people still trust it.
Delete rather than leave. A measure nobody should use will be used.
Where a measure is used
Once defined, a measure is available everywhere: the studio, the pivot builder, the analysis engine, each HQ's own reporting surface, and the assistant when somebody asks a question in words.
That reach is the argument for spending time on the definition and the reason a careless one is expensive. A measure is not a chart, it is a claim about what your business means by a word, and it will be quoted by people who never open this screen.
Editing one
Changing a definition changes every widget built on it, immediately and retroactively, because widgets store the reference rather than a copy of the number.
That is what you want and it is worth pausing over. A measure that has been on a board for six months has been quoted in meetings; redefining it silently makes those quotes wrong.
For a genuine change of meaning, prefer a new measure with a clear name and retire the old one. For a correction, edit and tell the people who use it.
Why this is worth doing properly
Everything downstream inherits from here. A widget is a measure with a shape, a board is a set of widgets, a scheduled report is a board on a timer.
Get a definition wrong and it propagates silently into every one of them, and it will be discovered in a meeting rather than in the studio. Get it right once and nobody has to think about it again, which is the entire point of a governed layer.
Did this answer your question?
No, ask a person