We split the account into two engines that work in tandem.
First, a tightly structured Search build: each ad group owns one product theme (e.g., “latex mattress” or “42-mm watch band”), seeded with exact and phrase-match keywords to hoover up obvious “buy” and “best price” queries, then broadened with controlled broad-match variants protected by a master negative list that blocks generic research terms.
Every responsive search ad includes dynamic keyword insertion in headline 1, calls out price or free shipping in headline 2, and layers in sitelinks to related categories, a promotion extension for seasonal offers, and price extensions to surface variants.
We launch on enhanced-CPC while query data matures, then switch high-volume ad groups to a target ROAS strategy once each has 30–50 sales in the look-back window.
Budgets flow to device segments and hours that out-convert, nudged by bid adjustments for high-margin products and returning visitors.
Running alongside is a multi-tier Shopping setup.
The product feed is tagged with custom labels for margin band, seasonality, inventory depth, and “hero” SKUs. We create three campaign layers that share the same feed but use Google’s priority setting to funnel traffic:
a high-priority “prospecting” campaign that blocks brand terms and harvests upper-funnel queries cheaply;
a medium-priority campaign that bids aggressively on generic-plus-brand phrases; and
a low-priority campaign reserved for exact product names and SKU codes where conversion rate is highest.
Each tier uses item-level negatives to prevent cannibalisation and lets us push spend toward best sellers while throttling low-margin long-tails.
Once both Search and Shopping achieve stable ROAS, we spin up Performance Max. Asset groups are organised by product-line, populated with top-performing creative (lifestyle images, short-form video, benefit-first headlines), and constrained to the same geography, language, and inventory rules already working in Search.
We exclude recent purchasers for some days, set an initial target ROAS equal to the blended account average, and let the system find incremental volume without undercutting the proven campaigns.
