The dominant failure mode in paid media isn't bad targeting or weak creative. It's optimizing against a metric that has no relationship to whether you made money. Platform-reported ROAS excludes cost of goods, shipping, returns, discounts and payment fees, and it double-counts across channels because every platform claims the same conversion.
Once you replace that with contribution margin, most of the hard questions become arithmetic. Break-even ROAS is one divided by your contribution margin percentage. Your maximum acceptable CAC is your lifetime contribution. Whether a budget target is reachable is a calculation, not a debate.
What remains hard is creative. Meta and TikTok now infer targeting primarily from the creative itself, which makes production throughput the binding constraint in most accounts. That's a volume problem with an unpleasant arithmetic of its own: if winners fatigue faster than your hit rate replaces them, the account declines no matter how good any single asset is.
What decides outcomes here
Optimize to the event that makes money
Not the easiest event to fire. Application starts, form fills, and add-to-carts are cheap to buy and frequently uncorrelated with revenue. Feed platforms funded accounts, closed-won deals, or purchases.
Never sum ROAS across platforms
Attribution models overlap. Compare summed platform-claimed revenue against your actual revenue: the gap is typically 30–60% and it's the size of the error you're budgeting against.
Creative volume beats targeting sophistication
Broad targeting plus high creative throughput consistently outperforms narrow targeting plus low throughput. The lever moved from the media buyer to the studio.
Test incrementality on your best-performing channel
The channels showing the highest attributed ROAS (brand search and remarketing) are usually the least incremental. That's the test most likely to change your budget allocation.