Conversion lift works like a drug trial. One group gets the treatment (ads), another does not, and you compare. It separates people who bought after seeing an ad from people who bought because of it.
Per the developer docs, when you create a study (ad_studies), Meta randomly assigns the Account Center accounts reached by your ads to a test group or a control group. In each cell, treatment_percentage is the share that receives ads and control_percentage is the holdout share that does not; they must sum to 100. After the study, Meta computes the difference in conversions, converting accounts and, where available, sales revenue between the groups. You can run a single test group to see the extra effect of ads, or several cells to compare approaches.
The schedule has a start_time, an end_time and an observation period (observation_end_time) in which conversions keep being matched after delivery ends. Once a study starts, start_time and treatment_percentage cannot be changed. Conversion data has to be shared through the pixel or app events.
The docs state that access is currently limited and you should contact your Meta representative. Unlike report numbers that count attributed conversions, it measures the extra conversions the ads caused.
When to use
Use it to tell whether ads create new conversions or only reach people who would buy anyway; confirm access with your Meta representative.