Most of the value in analytics is not in building charts. It is in trusting the ones you have, and in knowing what to do when you do not.
Drill into it
Any widget opens into the records behind it.
Do this the first time you see any number that matters. Not because the engine is suspect, but because it is the only way to find out whether the measure means what you assumed.
The frequent discovery is not an error in the arithmetic. It is that the definition includes something you would have excluded, or excludes something you would have counted. Both are invisible in a total and obvious in a list of rows.
Drill before you act, and before you quote a number to somebody else. Thirty seconds.
When it is empty
An empty widget is not a failure, it is an unanswered question, and the surface tells you which question.
The usual causes, and what each means:
The filter excludes everything. Most common by far. A scope narrowed to a pipeline, a hub or a status that has no matching records.
The period has nothing in it. A flow measure over a window where nothing happened, which is frequently a true and useful answer.
The measure has nothing to add up. A sum over a field that is empty on every record. This is a data problem rather than a chart problem, and it is worth chasing upstream.
The records exist but not for you. Analytics respects your permissions, so a colleague can see a populated version of a chart that is empty for you. Not a bug.
Rather than a blank rectangle, the widget explains which of these applies and offers the drill so you can see for yourself.
What drilling actually shows
The rows behind the figure, with their fields, as they exist right now.
Three things worth checking in them.
Are these the records you expected? Sometimes the scope includes a category you would have excluded.
Is the field being summed populated? A sum over a mostly-empty field is a small number with a confident label.
Is anything obviously duplicated? Two rows for one real thing inflate every measure built on that record type.
Numbers that disagree
Two figures that should match and do not is the most common reason people stop trusting a dashboard, and it almost always has a boring cause.
Different scope. One includes a status the other excludes. Check both scopes before anything else.
Different period. Month to date against last full month, or a calendar month against a rolling thirty days.
Flow read as snapshot, or the reverse. Summing point-in-time values, or asking a flow measure for a moment.
Two definitions. Two measures with similar names doing different arithmetic. This is the one that governance exists to prevent and the one that hurts most when it happens.
Work down that list in order. The fourth is rare and the first is usually it.
Scope is the usual culprit
If you take one thing from this piece, take this: when a number surprises you, look at the scope before you look at anything else.
Not the arithmetic, not the data, not the engine. The scope. In the large majority of cases the chart is correctly answering a slightly different question from the one you had in your head, and the difference is a filter or a window.
That is not a criticism of anybody. It is what happens when a question is held in a head and a chart is built from a definition.
Asking in words
A number can be put into a sentence describing what it shows.
Worth using on anything you are about to present. The description makes the scope explicit, and a surprising share of misunderstandings are somebody assuming a scope that was never stated.
A number that looks too round
Worth its own note because it catches people out repeatedly.
Exactly zero, exactly one hundred per cent, exactly the same figure as last month. Occasionally these are real and frequently they are a filter matching nothing, a division with no denominator, or a default standing in for a missing value.
The check is always the same: drill in. Four records behind a round number is an answer; zero records behind it is a question.
Widgets that were fine yesterday
A widget that changes without anybody touching it has a short list of causes.
The data changed, which is the ordinary case and the point.
A measure was redefined, which changes every widget built on it retroactively.
A pipeline stage was added or renamed, which changes anything scoped to stages.
Somebody's access changed, which changes what they personally count.
Work down that list before concluding anything about the business.
Habits worth forming
Drill the first time, every time. New widget, new number, one look at the rows.
Check against a screen you trust. If a revenue chart disagrees with Finance, resolve it before it is quoted.
Distrust a round number. Exactly one thousand, exactly fifty per cent. Occasionally real, frequently a filter matching nothing and a default filling in.
Distrust a sudden change. Before concluding the business changed, check whether a definition, a pipeline stage or a permission changed.
Say the scope out loud. "Revenue, closed won, this quarter, excluding intercompany" is a sentence that prevents an argument.
Permissions shape what you see
Analytics respects the same access rules as the rest of the platform, which has one consequence that surprises people.
Two colleagues can open the same board and see different numbers, legitimately, because each is counting the records they are allowed to see.
That is correct behaviour and it is worth knowing before somebody concludes a chart is broken. If a total differs between two people, compare access before compare anything else.
It also means a board built by somebody with wide access can look empty or thin to somebody without it, which is worth checking before sharing one widely.
When something is genuinely wrong
A number that cannot be explained by scope, period or definition is worth escalating rather than working around.
Bring three things: the widget, what you expected, and what the underlying records show when drilled. That turns a report of "the dashboard is wrong" into something anybody can act on in minutes.
Did this answer your question?
No, ask a person