Create & Run Experiment

Creates and runs a Google Ads custom EXPERIMENT (A/B test with a traffic split) on an existing Search or Display campaign: a control arm (the original campaign) and a treatment arm (a draft copy that gets the change), such as a different landing page (final URL) or ad text on a share of traffic.

Steps (one action each): 1. create — campaign_id, name, treatment_percent (1-99, default 50), optional start_date/end_date (YYYY-MM-DD). Creates the experiment in SETUP and the treatment draft campaign. Nothing serves yet. 2. set_final_url — experiment_id, new_final_url; optional ad_group_id (an ad group id of the ORIGINAL campaign) and/or old_final_url to limit which treatment ads change. Changes the landing page of the treatment ads only. replace_text — experiment_id, find_text, replace_text; optional ad_group_id. Replaces text inside the treatment RSA headlines/descriptions. 3. schedule — experiment_id. Starts the experiment on its start date (Google requires at least one change in the treatment first). • results — experiment_id. Control vs treatment clicks, impressions, CTR, conversions, cost and a significance test for CTR and conversion rate. Read-only. • status — experiment_id. Arms, split, draft/treatment campaign and the treatment ads with their final URLs. Read-only. To end, apply the winner (promote) or delete an experiment use list_experiments (action=end|promote|remove). Writes need confirm=true; confirm=false returns a preview. The original campaign is never edited by this tool; while the experiment runs Google locks its bidding/budget/targeting.