Zovda
An audience panel reports 62.4% female, 18 to 24, urban. The decimals imply a measurement. Behind them is usually a model trained on a panel that is itself a sample of who agreed to be measured.
That does not make the number useless. It makes the precision fictional.
Where the number comes from
Three common methods: declared profile data, a panel, and inference from behavior. Each has a different error profile, and they are usually blended into one field with no label.
Blending is what destroys the ability to reason about error later.
Keep the method as a field
Declared, panel, inferred, or unknown. A buyer comparing two creators needs to know whether they are comparing two measurements or one measurement and one guess.
Report the interval, not the point
If the method has an error of several points, say so, and round to the nearest five. A number without an interval invites decisions it cannot support.
Check the panel against something known
Where you have a ground truth, such as a survey with a real response rate, compare. The gap tells you how far to trust the inference elsewhere.
Zovda assembles these collections through Datuva, where the method and the interval travel with the estimate. The creator rate cards collection shows the creator record structure.
Tell us what you need to know about an audience. We reply within one business day.