FAQs (Forecaster)

Is there a minimum audience size that the Forecaster can forecast against?

The Forecaster requires an audience size of at least 300,000 unique users in total to produce reliable forecasts. If you apply restrictions such that the number falls below this minimum – as might be the case for a very small Geofence radius over a very short period of time – The Forecaster will return no data. Restriction sets that return no data are unrecommended to run as campaigns.

Is there a minimum pixel volume that the Forecaster can forecast against?

The Forecaster requires a pixel volume of at least 100,000 unique users per month to produce reliable forecasts. If you attempt to forecast against a pixel that falls below this minimum – as might be the case with a small retargeting audience – The Forecaster will return no data. Restriction sets that return no data are unrecommended to run as campaigns.

My forecast did not pull any data. Why?

There are a number of reasons your forecast query could return no or partial data:

• Data size is too small: The Forecaster is fed by a sampled data set. This means if the restriction set chosen is too small it may not make it into the sample. As a general rule, if the restriction set is too small to make it into the sample it is also too small to run a viable campaign against.

• If there is an insufficient volume of campaign data certain metrics may not populate as part of the forecast Forecaster data favors engaged users, ie users who click and view ads, as well as users we bid, win, and serve against.

• The forecast data returned is dependent on the size of the sample and these heuristics.

How large does my audience segment need to be to return a forecast?

300k unique users for an audience segment.

How large does my pixel audience need to be to return a forecast?

100k unique users per month per pixel.

Does the forecaster account for seasonality?

The Forecaster does not directly account for seasonality. However, we find most users forecast near enough to campaign delivery that it is irrelevant. Furthermore, most forecast requests are broad enough not to be impacted. This accounts for >95% of the Forecaster's use.

Why is the bid explode (Forecaster Chart curve) set with steps of 20?

While a bid step of 1 provides more accurate data and a denser curve, it was too costly for the avails service.  It was found after testing steps of 1, 5, 10, 20, 50, 75, and 100 that performance improved significantly at a step of 20 with minimal changes in the curve at several common restriction settings.

Why are only some of my selected and applied pixels and/or audiences showing up in Slice By Audience?

Only those pixels or audiences with data will surface when sliced by the Audience. It is safe to assume that if the Audience or Pixel is not present that there is no forecast data available.

Why are the Slice By Audience forecasts not the same as the forecast I see on the main page?

The Slice By Audience feature does not deduplicate User IDs that fall into two or more audiences. Or, stated differently, a single User ID that falls into two or more audiences will also be counted in both audiences in the Slice By Audience grid.

What do I need to select in order for Slice By Audiences feature to be enabled?

• You must select at least 2 audiences/pixels within the targeting selection screen in order for the audience option to be available in “slice by”

  • Selecting audiences and using the NOT operator on all but one audience is treated the same as selecting only one audience.

  • Selecting audiences and using only the NOT operator is treated the same as not selecting any audiences at all.

• Slice by Audiences results will only display audiences/pixels by qualified bid requests