By default, experiment results include only programmatic indirect demand. However, you can include ad server line items to understand the impact on different slices of your traffic.
- Sign in to Google Ad Manager.
- Click Optimization and then Experiments.
- Under "Experiments," click an experiment.
- Next to "Select types of line items to include in results," click No line items and select an option from the menu:
- No line items (Default): The experiment results don't include ad server line item traffic. The results do include Open Auction, Private Auction, Open Bidding, and Header bidding trafficking transactions.
- Non-guaranteed line items: In addition to the default slice, the experiment results include non-guaranteed line items (Price Priority, Network, Bulk, and Programmatic Direct). House and Bumper line items are excluded.
- Guaranteed and non-guaranteed line items: In addition to the default slice and non-guaranteed line items slice, the experiment results include all guaranteed line item types (Standard, Sponsorship, and Programmatic Guaranteed). House and Bumper line items are excluded.
- All line items: In addition to the default slice, the experiment results include all line item types, including House and Bumper.
Note that CPD revenue is not included in experiment results.
Supported experiments and opportunities
The following experiments and opportunities currently support ad server metrics:
- Unblock categories
- Unified pricing rules
- Native ad style
- Formats
- Video ad rule
- Header bidding trafficking
- User message
- Optimized competition (opportunity)
About the experiment results
To understand experiments results with ad server line items, keep in mind the following two points:
- We use the rate or value CPM of a line item, similar to reporting. This means that if a line item has an inaccurate CPM, the experiment results will also be inaccurate.
- We advise exercising caution when including guaranteed line items in your experiment results. Guaranteed demand is guaranteed by definition, and therefore should not be impacted by most experimental changes. However, since our pacing system is designed to ensure line item delivery across all traffic targeted by a line item, looking only at the subset of traffic an experiment applies to could introduce noise. The noise could present a misleading conclusion about actual performance.