Goal forecast FAQ
Because of subtleties in how AdMob simulates the ad server, you may see some numbers that you don’t expect. Following are some of the most frequently asked questions about forecasting.
AdMob campaign goal forecasts use a sample of the last 30 days of your network’s impression activity as a basis to model future traffic patterns. The forecasting simulation pits the prospective campaign for which you’re running a forecast against campaigns that are already running or scheduled, using the 30-day sample as the expectation of how many clicks or impressions will be available.
When a forecast is run, AdMob:
Looks at the existing state of the network.
- Runs a simulation for the campaign to be forecasted:
- As part of the simulation, AdMob first places the campaign at a priority lower than the other campaigns in the network. This artificially lowered priority means that the forecasted campaign isn't able to "steal" impressions from other campaigns.
- Next, AdMob runs another forecast with the forecasted campaign at its given priority.
Analyzes the simulation results to get the forecast estimate.
All active campaigns are used. Paused, deleted, or ended campaigns are ignored when the simulation is run.
You may be able to make more clicks or impressions available for a campaign by making changes to the campaign's targeting settings. For example:
- Target more ad units.
- Make changes to some of the targeting restrictions, such as adding countries to geo location targeting or adding more devices).
- Make changes to the targeting of competing campaigns. This will potentially impact those campaigns, so you may want to rerun the forecast for those campaigns before saving any changes to ensure they'll still meet their goals.
Even though two campaigns have the same targeting, the Available, Competing, and Unavailable numbers can be different. This is largely a result of the surpluses and shortages of impressions and clicks caused when lowering the priority of a campaign for the first step of the forecast simulation (see How do campaign goal forecasts work?). When that happens, other campaigns can claim the impressions and clicks that otherwise would have gone to the forecasted campaign, thus leaving other impressions and clicks unused.
Should a forecast over one wide date range be close to the sum of multiple, smaller forecasts that cover the same dates?
Let's say you:
Run a forecast from April 1-14.
Then run 14 forecasts — one for each day from April 1-14 — and add the forecast numbers together.
While the numbers may be close, they may not exactly match because of factors like frequency capping. For example:
A campaign has a frequency cap of one impression every 14 days.
A user session has several impressions on each of those 14 days.
The 14-day forecast assigns only one impression in the session, while each one-day forecast might assign an impression.
Yes, frequency capping is considered. However, because AdMob uses a 30-day sample, monthly frequency caps are not captured as part of the sample data. Consequently, in cases with monthly frequency caps, forecasting can overpredict availability.