→2. Multi Keyword / Target (1:1:n): when you already know which product to advertise (chosen strategically, or it has the superior CTR × CVR combination), when targets have low volume — under roughly one conversion a day — and you consciously give up per-target placement control, or when you want to test a group of keywords without a campaign for each. Fewer campaigns, easier manual bidding, budget control on the advertised product; but no granular placement/B2B/audience adjustment, and one budget can throttle the delivery of some targets.
→3. Multi Ad (1:n:1): when you don't yet know which product to advertise, or want to test a hypothesis (better CTR or CVR with a different product) and let Amazon pick. Usually temporary — the exception is genuinely comparable products with similar performance and no strategic reason to separate them, such as differing inventory levels. Cost: without Marketing Stream data there's no placement / search term / ad report combination, and you lose budget control at ASIN level.
→4. Multi Ad & Keyword (1:n:n): a combination of the two above — keyword testing, low-volume targets, near-identical products, or when it's unclear which product should be advertised. Quick, easy testing setup that keeps the account lean, at the cost of blurrier reporting and dramatically less precise bid control.
→5 and 6 — the multi ad group structures (n:1:n and n:n:n): I personally don't use these at all. They do let you separate match types or targeting types at ad-group level and give more daily-budget control through fewer campaigns, but you get no granular placement, B2B or audience adjustment, no meaningful evaluation at campaign level, hardly any placement evaluation, and manual control becomes genuinely difficult.
Certain structures can be the right answer temporarily. Challenge them regularly — and be willing to dismantle or consolidate them again.