Experiment

Before and after comparisons mix the change you made with everything else that moved: season, competitors, budget. An experiment splits the same traffic at the same time, so the difference is the change.

It is the honest way to test a new bid strategy, landing page or structure on an account with enough volume.

1. Open Campaigns, then Experiments.
2. Create a custom experiment on a campaign, change one thing in the trial, and set the split, usually 50/50.
3. Run it for several weeks and watch the significance shown in the results.
4. Apply the trial to the original campaign if it wins.

A campaign tests Target CPA against Maximize conversions on a 50/50 split for five weeks. Target CPA brings 12% more conversions at a similar cost, with Google showing the result as significant, so it is applied.

• Changing several things in one trial, so nobody knows what worked.
• Stopping the test after a few days.
• Running experiments on campaigns with too few conversions to give an answer.

The audit checks Oppy runs on this, each with its own page.

• An account big enough to test that has never tested
• An ad test with a real winner

• Bid strategy: How a campaign sets its bids: by hand, or automatically toward a goal such as clicks, conversions, conversion value, a target CPA or a target ROAS.
• Incrementality: The part of a result that would not have happened without the ad.
• Learning period: The time Smart Bidding needs to recalibrate after a new strategy or a large change.