Facebook’s ML can probably use these events and improve DoorDash’s ad delivery efficiency by a significant percentage. It potentially equates to millions of dollars of savings depending on how they use the ad platform. Even just allowing doordash to measure their advertising outcomes is a big deal compared to traditional advertising. You could spend $X million on tv and have no clue if it worked, or spend it on Facebook and use these events to get precise measurement to what degree your ad dollars are resulting in real business outcomes.
Step 1: Build a website that sells things. Step 2: Sell the data about those transactions to anyone and everyone willing to buy it.
After that, who cares what is done with it? Best not think about those things lest it bother your conscious.
Your question probably is more along the lines of why would someone want to buy that data? You add that info to the buyer's enormous trove of data profile. This person bought an Apple device = they have lots of disposable income (send them ads on other superfluous crap). This person bought a baby crib = send this person lots of adds about diapers, formula, toys, etc.