Part 3 — Structure the account around margin, not vibes

Account structure has one job in the Smart Bidding era: give the algorithm clean, sufficient signal while keeping control where your economics demand it. The old instinct — mirror the site taxonomy, one campaign per category, dozens of micro-segmented ad groups — starves every campaign below learning volume and hands you back noise. The modern instinct — one campaign, everything in it — hands Google decisions it will resolve in its own favor. The right structure sits between, and your part 1 margin bands define exactly where.

The consolidation rule

Automated bidding needs roughly 30+ conversions per month per campaign to clear the learning phase; below that, targets are guesses and performance whipsaws. So consolidate until each campaign clears the floor — then and only then consider splitting. The legitimate reasons to split are economic, not organizational: different margin bands (which need different ROAS targets), different strategic roles (acquisition vs. clearance), or genuinely different demand pools. "The site has a menu item for it" is not a reason. The fuller argument, including how account standards encode this, is worth the detour.

Brand and non-brand: never blended

Brand search converts at multiples of non-brand because the customer had already decided; blending them lets brand’s easy wins subsidize non-brand’s real costs and flatters every average the account reports. Isolate brand into its own campaign with its own budget and modest targets, and read growth from non-brand — the traffic that represents customers you did not already have. We covered the measurement distortion in brand vs non-brand; structurally, the separation is non-negotiable and costs nothing to do on day one.

A reference shape for a typical store

1. Brand Search — exact and phrase brand terms, capped budget, high impression-share target.
2. Non-brand Search — the head terms your category genuinely wins, consolidated until conversion volume justifies splits.
3. Shopping or PMax segmented by margin band via custom labels — each band with a ROAS target derived from ITS breakeven, not the account average (part 4 and 5 cover the Shopping/PMax split itself).
4. Remarketing/retention kept deliberately small and honestly measured — it harvests decided demand and should never be allowed to inflate blended results.