Estimation Basics

Forecasting and estimation entails two concepts:

  • inventory estimation: Inventory refers to the available number of bid requests (and with that potential impressions) of a site, which are extrapolated from a sample of a few percent of the received bid requests during yesterday.
  • user estimation: Users refer to unique browsers, estimated with a hyperloglog data structure.

Best Practice

  • Use the RTB Sitelist Inventory Estimation Service to estimate the available number of bid requests for a given sitelist and geo target
  • Use the User Estimation Service to estimate the available number of users for a given sitelist and an optional audience expression
  • If called without an audience expression, the User Estimation Service will return the total number of unique users in a given sitelist. Combined with the estimated user results from a given audience expression, this can be used with the inventory estimation numbers as a scale factor to estimate the number of actual available impressions for a given sitelist and audience.

Limitations to Estimation

  • Intersections (AND) of data sets in the Data Volume Estimation Service are limited to three sets. This is due to statistical error limitations of hyperloglog data structures.
  • User estimation is based on data collected through sampling of the real time bid request stream. This means that newly created audiences will not be possible to estimate. For this scenario, we recommend estimating based on an expression of the individual audience components, if possible.
  • Sampling of segment and cookie group presence on a particular site is always done for incoming bid requests at a rate of 5%.
  • Sampling of audience presence of a particular site is not done by default. Contact your account manager to activate a sampling rate.
  • User tracking is cookie based, which may lead to an inflated representation of devices not allowing cookies.