When a campaign's ACOS exceeds the product's profit margin, every sale made through that campaign is losing money — this is the trigger condition for optimization, not an arbitrary "ACOS looks high" judgment call.
The fix proceeds in two ordered steps, not simultaneously:
Doing keyword surgery before fixing placement risks over-correcting keywords that were only underperforming because of a placement problem. See Leading Actions vs. Lagging Metrics (PPC Diagnosis Framework) for the broader four-metric diagnostic this sits inside, and Six-Campaign PPC Launch Structure for the campaign structure being optimized.
ACOS = (ad spend ÷ sales revenue) × 100. Break-even logic compares this per-keyword ACOS against the product's profit margin percentage: any keyword whose ACOS exceeds the margin is losing money on every sale attributed to it, regardless of how many total sales it's driving. This per-keyword margin comparison is the trigger condition for the optimization sequence — see also ROAS (Return on Ad Spend) for the equivalent check expressed as a return ratio instead of a cost ratio.
Quote underscoring the trigger condition for this sequence: "if the ACOS is greater than your profit margin, you're actually losing money each time you make a sale" — i.e., the moment ACOS crosses the margin line, the campaign is subsidizing sales rather than profiting from them.
ACoS should never be judged against a universal 'good' number — it should be judged against the seller's actual profit margin, since a 40% ACoS is fine at a 60% margin and ruinous at a 20% margin. It's also correctly analyzed last in the funnel, not first: ACoS is the output of everything upstream (impressions, CTR, conversion rate), not the root problem itself, so fixing ACoS means fixing whichever upstream metric is actually broken. See Leading Actions vs. Lagging Metrics (PPC Diagnosis Framework).