Optimizing marketing spend - you can show how a ML model will beat a conventional past performance regression algorithm by increasing sales, reducing cost, etc.
Learning Prolog was one of the better decisions I have made. Not because Prolog jobs are plentiful. Rather, understanding logic programming by broadened my problem-solving skill set.
Good question! Combination of holdout groups (split testing) based on historical keyword traffic/conversion and tagging paid and natural search entries to compare cannibalization across campaigns.
I worked in ecommerce for 10 years and got to know the quantitative marketing folks well. All marketing campaigns were tracked and analyzed weekly based on the concept of "marketing income". Marketing income at the order level was defined as: Marketing Income$ = Order Total$ - Cost of Goods$ - Shipping$ - Marketing Expense$.
I would imagine many folks who are visiting cancer/ALS content are impacted by the diseases either directly or indirectly. Affiliate marketing (directed to non-profits) seems like a place to start.