Similar to linear attribution model, in a time decay attribution model, all ad interactions will be taken into consideration, however, in this method of measurement, the time proximity to the conversion point plays an important additional factor, where advertisers will calibrate so that earlier ad interactions will recieve a lower score than ad interactions closer to the date of the conversion.

This method relies on the access to user level data, and extensive calibration and model training. 
Given the diminishing access to user level data - this method is extremely rare, and is often only used only in theoretical and academic settings.