The wrong question is "which platform is better"
Marketers often ask whether Google or Meta delivers better ROAS, but the more useful question is what job each platform is best suited to do, because they capture fundamentally different types of demand.
Google Search: harvesting existing intent
Search ads work because the person is already looking for something — they typed the query. This makes Google Search excellent for capturing demand that already exists, but it does very little to create new demand. Budget here should scale with search volume for relevant terms, and tends to have the most predictable, measurable ROAS of any channel.
Meta and Instagram: creating demand from scratch
Feed and Reels ads interrupt someone who wasn't looking for the product at all. This is harder to measure in last-click terms, but it's how most new-to-brand customers are introduced to a product in the first place. Businesses that under-invest here because "the ROAS looks worse than Search" often end up starving the top of their own funnel, which eventually shows up as declining search volume for their brand name.
A practical starting split
For most e-commerce and DTC businesses without strong existing brand search volume, a rough starting allocation of 60% to upper-funnel demand-generation (Meta/TikTok) and 40% to demand-capture (Google Search/Shopping) is a reasonable test point — adjusted from there based on blended CAC, not platform-isolated ROAS.
Shopping ads as the middle ground
Google Shopping campaigns sit between the two: they capture search intent but also function as visual discovery, especially on the Search results page itself. For product-based businesses, Shopping campaigns often deliver disproportionately strong returns relative to spend and deserve a meaningful share of the Google budget on their own.
Reassess quarterly, not weekly
Platform performance shifts with algorithm updates, seasonality, and competitive bidding pressure. Reacting to week-to-week fluctuations usually causes more harm than it prevents; a quarterly budget reallocation based on trailing blended CAC trends is generally a more stable approach than constant week-to-week shifting.