Not every figure is equally well supported. A margin calculated from fully costed jobs and one calculated from jobs missing their labour are both numbers, and only one of them is worth acting on.
So figures carry a confidence, drawn from how complete and how corroborated the data behind them is.
The bands
Excellent — the data behind it is complete and agrees with itself.
Good — well supported, with minor gaps.
Fair — usable, with enough missing that the figure should be read as approximate.
Below fair, a finding is held back rather than shown.
Why hold something back rather than show it with a warning
Because a caveat next to a number does not survive contact with a busy week. The number gets read, the caveat does not, and a decision gets made on a figure nobody would have used if they had understood how thin it was.
Being cautious by default is the more honest failure. You are told something is missing, which is actionable, rather than told something uncertain, which is not.
What raises confidence
The same three things every time, and none of them are settings:
Connect the sources. A figure computed from the bank alone is less well supported than the same figure with the bank and your accounting software agreeing.
Clear the review queue. Uncategorised money is money the breakdown cannot see.
Cost the jobs. Materials and labour attached to the job rather than the month is what makes per-job figures possible at all.
Why a number can disappear
If a figure was visible last month and is not now, the usual cause is data rather than the business: a connection stopped syncing, or a period has more uncategorised activity in it than the one before.
The evidence chain is where to look, because it shows what the figure was built from — and an empty answer there is the finding.