A report arrives with a table in it. One cell is blank. There are three ways to fill that cell, and all three of them are lies: print a zero, produce an estimate, or copy the number from last month.

Here is what we print instead, and why the blank is worth more than any of the three.

not reported — nothing has been estimated and nothing from another period has been substituted

That is not a hedge. It is the most informative line in the report, because it tells you exactly where the evidence stops.

Zero is a measurement

Zero says: we looked, and there was none. A missing source says something different: we did not look, or we could not.

Printing zero turns "unknown" into "none", and once that happens the two are hard to tell apart ever again. A zero also drags an average down, flattens a chart, and three months later nobody can remember whether it was a real zero or a missing one.

An estimate is a guess wearing a number

An estimate looks like data and cannot be audited. Nobody can check it, so nobody can act on it with confidence, and nobody can defend it when someone asks where it came from.

If an estimate is genuinely useful, it should be labelled as one, kept apart from the measured figures, and never averaged in with them.

Last period's number is not this period's

Substituting is worse than estimating, because the result looks exactly like a measurement. A number that describes August cannot describe September just because it was the last one available.

What a blank should carry

A blank on its own is an excuse. A blank with a reason is an instruction:

  • The source it was meant to come from.
  • The date it was last available.
  • What would make it available — connecting the account, granting access, waiting for the platform's own reporting cycle.
  • Who is responsible for it.

Why a vendor should say it out loud

Two reasons. It protects your decisions: a report that is honest about what it does not know is the only kind you can trust about what it does know. And it protects the vendor from having to defend a figure they invented.

How to test a vendor on it

Ask three questions:

  1. What does a missing number print as in your report?
  2. Show me one.
  3. Is any figure in this report an estimate, and which one?

A vendor who answers all three has a system. A vendor who answers none has a template.

What we do

Where a source cannot be read, we mark it as not checked rather than scoring it as zero. The Visibility Score is built from up to seven parts, and a part we cannot check is left out — which is why it says "up to seven" rather than pretending all seven were read. The findings behind the score name the source and the date they were read, so the ones that are there can be checked.

For how that works in detail, see how the BravData Visibility Score is built.

What this standard cannot tell you

  • Whether the numbers that are present are correct. Only the source and the date let you check that.
  • Whether the missing data will ever arrive. Some sources need your permission; some may never be readable at all.
  • Whether the work is any good. Honest blanks make a report trustworthy; they do not make it good.