Part 6 — Bid to margin, measure with MER
tROAS vs tCPA, margin-aware conversion values, server-side tracking, and MER as the number that cannot lie.
Everything so far — economics, feed, structure — exists to make this part work. Smart Bidding is exceptional at optimizing whatever you feed it; the operator’s leverage is entirely in what you feed it and how you verify the results. Two disciplines carry the weight: margin-aware conversion values in, blended measurement out.
Choosing the strategy: tROAS vs tCPA
For ecommerce, target ROAS is usually right because order values vary and the value signal matters. Target CPA wins when order values cluster tightly or when you are optimizing to a fixed-value event like a first purchase in a subscription business. The decision logic — including the volume thresholds where each becomes reliable — is in tROAS vs tCPA: when each wins. Set the actual target from part 1’s breakeven plus your profit requirement, never from what the account "used to do."
POAS: make the machine chase profit
The highest-leverage change most ecommerce accounts can make: send margin-reflecting conversion values instead of revenue. Per-order gross profit if your data supports it; margin-banded revenue weights if not. Target ROAS then becomes target POAS without the platform knowing the difference, and Smart Bidding starts scaling the products that actually make money. The implementation ladder is in bidding to margin — start at whatever rung your data allows and climb.
Feed the algorithm reliably: server-side tracking
Browser-side tracking loses a meaningful share of conversions to blockers, ITP, and consent friction — and Smart Bidding treats missing conversions as failures, bidding down exactly the traffic that converted invisibly. Server-side conversion tracking with proper deduplication restores the signal. The 2026 setup guide covers the architecture; treat it as infrastructure, not optimization.
MER: the referee
Platform ROAS decides allocation between campaigns; it cannot be trusted on the question "is the whole system profitable" because attribution flatters. Marketing efficiency ratio — total revenue over total marketing spend, straight from the P&L — cannot be gamed by attribution at all. Set a MER floor from blended margin and overhead, watch it monthly by spend tier, and when platform numbers and MER disagree, believe MER. That single habit catches more expensive drift than any dashboard.