Meta creates demand.
Google captures it.
You need both.

One honest caveat on every benchmark in this post. Public ad benchmarks are wide, vary by source, and bury a lot of variance inside category averages. Use them to sanity-check your own account, not to predict it. Your category, margin, creative, and offer move your numbers more than any industry median does. Where I cite a figure I have linked the source; where I give an operator view I have said so.

Where Meta
is unbeatable.

Meta's superpower is simple to state and hard to replicate: it can get a brand nobody has heard of in front of millions of likely buyers, fast, with nothing but good creative. No keyword has to exist. No search has to happen. If your product is genuinely good and your ad is genuinely scroll-stopping, Meta will find the people who will love it and put it in front of them at a scale no other platform matches.

That is why the modern DTC playbook was born on Facebook. For a launch-stage brand, there is no meaningful demand to capture, because the world does not yet know the brand exists. The entire job is creation, and Meta is the best creation machine ever built. This has not changed in 2026, even with CPMs up 20%. The cost went up; the unique capability did not move.

What Meta is genuinely best at

Where Meta hurts is measurement and patience. It was the platform hit hardest by iOS privacy changes, so its dashboard is the least trustworthy of the two, and its job, planting demand that pays off later, is the hardest to attribute cleanly. A brand that judges Meta only on last-click, same-session ROAS will almost always under-credit it and cut it too early, which, as section eleven shows, is one of the most expensive mistakes in DTC. Meta asks you to believe in a return you cannot fully see in the dashboard, and that requires either trust or, better, real incrementality testing.

Where Google
is unbeatable.

Google's superpower is the inverse of Meta's. It does not have to convince anyone of anything. By the time a person types "best magnesium supplement" or your brand name into Google, the wanting is done. Google's only job is to be the offer they click, and because intent is already there, the conversion rates and efficiency are structurally higher than anything prospecting on Meta can hit.

This is why Google quietly becomes more important as a brand scales. A bigger brand generates more searches, both for the category and for its own name, and every one of those searches is a cheap, high-intent capture opportunity. Paid search ROAS of 4x to 8x is common for DTC, with branded search reaching the top of that range and non-brand sitting lower (MHI Growth Engine, DTC advertising benchmarks, 2026). The catch, covered in section twelve, is that the branded portion of that is mostly demand you already created and would have captured anyway.

What Google is genuinely best at

Where Google hurts is its ceiling. You cannot spend your way to more demand on Google. If only 10,000 people a month search for your category and brand, that is the size of the prize, and once you own most of those clicks, additional Google budget just inflates your own bids. Google is efficient but bounded, and the only way to raise the bound is to create more demand somewhere else first. That somewhere else is usually Meta. Which is the whole point of this post.

The two black boxes,
and what they hide.

Both platforms have pushed hard into automated, AI-driven campaign types, and most DTC spend now flows through them. On Google it is Performance Max; on Meta it is Advantage+ Shopping. They look superficially similar, a single campaign that lets the algorithm decide placement, audience, and bid, but they inherit the same create-versus-capture split as their parents, and understanding that keeps you from expecting the wrong thing.

Performance Max captures existing purchase intent across Google's full inventory: Search, Shopping, YouTube, Display, and Gmail, driven by your product feed and Google's signals about what people are searching for. Advantage+ Shopping creates demand by finding likely buyers who are not searching, driven by your creative library and Meta's ranking system (Ryze, PMax vs Advantage+, 2026). PMax is an intent-capture machine wearing an automation costume. Advantage+ is a demand-creation machine wearing the same costume. They are complements, exactly like the platforms underneath them.

Now the uncomfortable part, because the automation hides two real problems. On Meta, a Haus study of 640 incrementality tests found that Advantage+ campaigns underperform manual campaigns over time, despite posting strong reported ROAS (Ryze, citing Haus, 2026). The dashboard looks great; the incremental lift is smaller than it claims. On Google, Performance Max has a documented habit of scooping up branded search traffic, claiming it as its own conversions, and reporting a ROAS inflated by demand that was already going to convert (AdsX, Search vs PMax cannibalization, 2026).

The practical defense is the same on both: do not trust the black box's self-report, and do not let it eat your cleanest traffic. On Google, run a separate branded search campaign so PMax cannot claim those conversions, and watch your branded impression share for the telltale decline that signals cannibalization. On Meta, keep some manual or structured campaigns running as a measurement control, and validate Advantage+ against holdout tests rather than the dashboard. Automation is fine. Automation you cannot audit is not.

What you actually
pull to win.

Here is the operational truth that most budget arguments miss entirely: the thing that determines whether each platform works for you is different, and you cannot win either one by fiddling with the other's lever. On Meta, the lever is creative. On Google, the lever is your product feed and keyword structure. Pour effort into the wrong one and you will conclude a platform "doesn't work" when really your inputs were wrong.

Meta is a creative business now

With automation handling targeting and bidding, the human job on Meta has collapsed into one thing: feeding the machine enough good, varied creative to find winners. By most practitioner estimates, roughly 80% of performance work in 2026 is creative operations and only about 20% is media buying (SuperAds, creative diversity, 2026). Meta's ranking system now tests far more ad variants in parallel than it used to, so it rewards volume and diversity, not clever audience setups.

The volume benchmarks are higher than most brands run. A study of over 200 DTC accounts found that brands producing 30 or more new creatives per month scaled roughly 3 times faster than those producing fewer than 10, and brands testing 60-plus a month saw meaningfully higher ROAS than those testing fewer than 20 (Billo, how many creatives, 2026). The honest floor is 8 to 16 new concepts a month with a few iterations on each winner; elite brands ship 50 to 100. Across the brands I have operated and advised, the account that wins is almost always the one with the deeper creative pipeline, not the one with the cleverest media buyer.

Google is a feed-and-structure business

Google's lever has nothing to do with creative in the Meta sense. Winning on Shopping and Performance Max is mostly about the quality of your product feed: accurate titles stuffed with the words people actually search, clean images, correct pricing, and rich attributes. On Search, it is about keyword structure, the discipline of separating brand from non-brand, and writing ad copy that matches intent. The art is unglamorous and it compounds. A well-built feed quietly out-earns a sloppy one at the same spend, every single day.

The mistake I see constantly is a brand treating both platforms with the same playbook. They run beautiful creative through Google's feed where it does little, or they push a thin product feed and weak keyword structure while obsessing over Meta creative. Each platform has exactly one lever that matters most, and they are different levers. Pull the right one for the platform you are on. If feeds and search structure are the bottleneck, that is the call to make on whether to run it in-house or bring in help, which I dig into in Google Shopping agency versus in-house.

Both dashboards
are lying to you.

If there is one section that will save you the most money, it is this one. The ROAS numbers Meta and Google show you are not neutral facts. They are self-graded report cards, and both platforms grade themselves generously. Allocating budget by trusting those dashboards is the single most common, most expensive mistake in DTC paid media, and in 2026 it is worse than ever.

Here is why it broke. Since iOS privacy changes in 2021, deterministic, user-level tracking collapsed. Multi-touch attribution coverage fell from over 90% to roughly 30% to 60% by 2026, and to fill the gap both platforms now model a large share of conversions (AdLibrary, ad attribution explained, 2026). Modeling is not neutral: each platform models in a way that flatters its own contribution, which is why your Meta dashboard plus your Google dashboard often sum to more revenue than your Shopify store actually made. They are both claiming the same orders.

The gap between reported and real is large. Platform-reported ROAS commonly runs 2x to 5x higher than incremental ROI measured by marketing mix models (Cassandra, Google Ads benchmarks not from platform attribution, 2026). Think about what that means for a budget decision: if you move money toward whichever dashboard shows the higher ROAS, you may be moving it toward the platform that lies more, not the one that earns more.

The 2026 measurement stack

The fix is not a better attribution tool. It is a different hierarchy. The best-practice 2026 stack is marketing mix modeling plus incrementality experiments plus tactical platform reporting, in that order of authority (Leadgen Economy, cookieless attribution stack, 2026). Marketing mix modeling allocates the big budget because it never needed user-level IDs and is largely immune to privacy changes. Incrementality holdout tests answer the causal question, did this spend actually cause sales. Platform dashboards drop to the bottom: useful for in-platform optimization, not for cross-channel allocation. Adoption is real, a 2025 EMARKETER survey found 52% of US marketers now use incrementality testing precisely because standard attribution fell short.

You do not need an enterprise MMM vendor at $2M in revenue to act on this. Even a small brand can watch blended marketing efficiency ratio, the simplest cross-channel sanity check, and run a basic two-week Meta pause in a holdout region to see what actually happens to total orders. The discipline matters more than the sophistication: decide budget on what the business made, not on what each platform claims it made.

The split moves
with your stage.

Now the part you came for. The right Meta-to-Google split is not a fixed ratio, it is a function of how far along the create-to-capture curve your brand sits. Early on you must create demand because there is almost none to capture. As you scale, you generate your own search demand and the capture side grows into a bigger, cheaper opportunity. The split follows that curve.

The pattern is well documented: most ecommerce brands start Meta-heavy and shift toward Google as they scale, with inflection points around $1M, $5M, $20M, and $50M (Darkroom, Google vs Meta budget guide, 2026). Here is how I think about each band, blending that data with what I have run.

Read these as starting points, not laws. The same source notes the split also depends on margin structure, creative velocity, and measurement maturity, which is exactly right. A high-margin brand with a 50-creative-a-month pipeline can run Meta harder than the table suggests. A brand with thin margins and a great product feed might lean on Google's efficiency earlier. The stage sets the gravity; your specifics set the exact number. And every one of these should be pressure-tested with the incrementality testing from section nine before you treat it as truth. For how the whole channel portfolio assembles beyond these two, the 2026 channel mix strategy piece is the wider map, and the Shopify vs Amazon decision is the other big channel call most DTC brands face.

Cut Meta, and Google
bleeds in six weeks.

This is the most important idea in the entire post, and the one almost nobody internalizes until they have lived it. Meta and Google are not independent line items you can optimize separately. They are interlocked. Cutting Meta prospecting does not just reduce Meta revenue. It reduces Google revenue four to eight weeks later, when the demand pool Meta was filling runs dry (Ryze, budget allocation, 2026).

Watch how this trap springs, because it is seductive. A brand looks at its dashboards. Google shows a clean 5x. Meta shows a murky 1.8x prospecting ROAS that the last-click view makes look like a loser. The obvious move, the move I have watched smart operators make, is to cut Meta prospecting and pour that money into the "efficient" channel. For a few weeks it looks brilliant. Blended efficiency ticks up because you stopped spending on the low-ROAS line.

Then the lag hits. The branded searches that fed Google's 5x were never free demand. Meta created them. With the demand pump turned off, the branded search volume falls, the cheap high-intent clicks dry up, and Google's efficiency quietly erodes, because now it is mostly capturing demand that no longer exists. The brand cut the engine that was filling the funnel and kept the drain open. Six weeks later, both lines are down and nobody can quite explain why.

The defense is to refuse to judge the platforms in isolation. Meta prospecting at a 1.8x last-click ROAS may be carrying a 5x branded-search line on its back. The only way to see that is incrementality testing, a real holdout where you cut Meta in some regions and watch what happens to total orders, including Google's, not just Meta's dashboard. When operators run that test honestly, they almost always find Meta is worth more than its last-click number says, because it is funding the capture they were about to over-credit to Google.

The honest
scorecard.

With the system understood, here is the blunt, use-case-by-use-case verdict. This is the closest thing to a "winner" you will get, broken down by job rather than declared overall.

Count the rows and it looks like a tie, which is the point. Meta wins the creation jobs, Google wins the capture jobs, and a real brand needs both kinds of job done. The one row that quietly outweighs the others is the last one. Because Meta fills the pool Google drains, a brand can survive a weak Google setup longer than it can survive a starved Meta program. You can always capture more of an existing pool; you cannot capture a pool that was never created.

The split is downstream
of your margin.

Every word above is secondary to one number: the most you can afford to pay for a customer and still make money. The Meta-versus-Google debate is a debate about how to buy customers efficiently, but if you do not know your ceiling, efficiency is meaningless. You can run a beautiful 4x ROAS and still go broke if your margin cannot support the CAC underneath it.

This is why I push operators to settle unit economics before they touch the platform split. Ad spend as a share of revenue varies enormously by stage: public DTC companies run a median around 13% of revenue on marketing, but private scaling brands in the $5M to $50M range commonly run 20% to 35% because they are buying growth, not optimizing for the public market (Eightx, marketing spend percent of revenue, 2026). The cleaner way to govern it is blended marketing efficiency ratio, which scales with maturity: early brands often run 1.5x to 2.5x and lose money on the first order, $5M to $10M brands run 2.5x to 3.5x, and $25M-plus brands push 3.5x to 6x or higher (Eightx, marketing efficiency ratio, 2026).

Notice what that early-stage MER implies. A brand running 1.5x to 2.5x blended is losing money on the first purchase on purpose, betting on repeat revenue to make the customer profitable over time. That bet is only sane if you know your real contribution margin and your repeat behavior, which is exactly the math the max allowable CAC calculator exists to do, and why CAC payback varies so much by category, as the CAC payback by vertical breakdown shows. Set the ceiling first. Then the question of whether Meta or Google gets the next dollar has an actual answer, because you can ask which one buys customers below that ceiling right now.