In a data-driven attribution model, how can an advertiser understand the impact of any given keyword in a conversion path?

Take two paths that look exactly the same apart from one keyword; compare conversion rates to determine the uplift from that one keyword
Take an AdWords campaign with at least 30 conversions in two weeks; remove the keyword in question and re-run the campaign for two weeks
Take all the keywords in an AdWords account and rank them by click volume; use static rules to assign credits to touchpoints
Take one individual click path; look at the conversion rate of the last clicked keyword to determine how it performs for any given auction