Rendering Diversion

[jeffkaufman]

In developing new creative formats and tweaking existing creatives, it is common to run rendering experiments. Should the text go here, or here? Should the text be bold? What should be the shape of the button? We are interested in running these experiments in response to rendering_url requests.

The browser would request the HTML (or, eventually, Web Bundle) for an ad, and the server would return a fully rendered ad. Each ad is the product of many, perhaps 30, experiment decisions. For example, users might be randomly assigned to either a "bold text" experiment group (id=12345) or a "no change" control group (id=67890). We're imagining the server could return the IDs for each treatment group it applied in its response.

Once the ad rendered in the browser, it could pass the treatment ids to the reporting worklet which could, with event level reporting, be logged along with other information the worklet has. After the transition to aggregate reporting, the reporting worklet could send separate pings for each treatment group.

This relies on some mechanism for the creative rendering in a fenced frame to pass information to the reporting worklet. Does that sound like something FLEDGE could support?

[JensenPaul]

Closing as I believe this is supported in via the eventData field of the Fenced Frame Ad Reporting API.