Reading a modelled number

A number on a screen carries no evidence of where it came from. The same typeface renders a figure counted yesterday and a figure assumed for 2031, and a reader has no way to tell them apart unless they are told.

The six classes

• Observed — counted or measured. A night taken, a payment received.
• Verified — asserted by someone and checked against a source document.
• Modelled — computed from stated assumptions. Correct arithmetic on inputs that may be wrong.
• Estimated — a judgement, made by a person, on incomplete information.
• Forecast — a modelled figure about a period that has not happened.
• Pending — expected, not yet arrived. Shown as absent rather than as zero.

The classes are ordered, and the order is the point. A figure never appears with a stronger class than the weakest input it was computed from.

A number quoted without its class has been laundered by being moved.

Why a modelled yield is not a yield

A modelled distribution is the output of an occupancy assumption, a rate assumption and a cost assumption, run through a waterfall. The arithmetic can be flawless while all three assumptions are wrong, and nothing about the presentation of the result would look different.

This is why a provisional figure carries its mark everywhere it appears rather than carrying it once at the point of origin. A number quoted without its class has been laundered by being moved.

What happens when sources disagree

Material prepared by a promoter sometimes contains figures that cannot all hold at once. The rule is: carry the conservative figure, state the inconsistency on the record, and compute the dependent figures rather than restating them, so the two can never disagree again.

Read next