Does Meta or Google deliver better ROAS in H1 2026?

Across revenue-tracked Synter-connected accounts in H1 2026, Meta's median return on ad spend reaches 3.1× and beats Google on cost base — median CPA of $35 vs $48 — despite a lower median conversion rate. Cheap-but-mediocre traffic wins when the price gap is wide enough.

Why is Performance Max called the over-adopted underperformer?

Performance Max is one of the most widely adopted Google campaign types, but at the account-level median it underperforms more targeted campaign types — so it quietly becomes default spend before it is audited.

How is this benchmark measured?

Per-account medians across every account connected to Synter over H1 2026 (Jan 1 – Jun 13), anonymized, with physically-impossible accounts excluded and revenue figures limited to accounts where ROAS is plausibly tracked.

Benchmarks

The State of Agentic Advertising · H1 2026

Where the money actually works

A data-backed, mid-year look at paid media across every account connected to Synter — $30M in clean spend, read at the median, not the average. This is the story of two platforms that started the half in the same market and ended it in two different worlds.

246: connected accounts (Google + Meta)
$30M+: clean ad spend analyzed
2.2M: conversions

Scroll to begin ↓

Per-account median, H1 2026 (through Jun 13). Google n=149, Meta n=97.

Chapter 1 · The price of attention

Two platforms enter H1 2026 with opposite economics.

Start with the simplest fact. Across 246 guarded Google and Meta accounts — $30.2M in clean spend, 2.19M conversions through Jun 13 — Meta's median CPM is $13.48 against Google's $37.60. Attention on Meta costs roughly a third of what it costs on Google. That single gap sets the table for everything that follows: it gives Meta room to be wrong and still win, and it gives Google no margin for error. Hold that thought — because the half didn't stay still.

Chapter 2 · The turn

Month by month, Google slid toward break-even while Meta pulled away.

This is the beat the averages hide. Track the median account inside each month and the two platforms diverge in slow motion. Google's median ROAS drifts — 1.39× in March, 1.57× in April, 1.38× in May — then drops to 0.98× in June: the typical Google account ended the half losing money. Its median CPC climbed the other way, from $1.64 to $2.05. Meta did the reverse, holding CPC near $0.70 while its median ROAS climbed from 3.1× to 5.5×. Same calendar, opposite trajectories.

Chapter 3 · Cheap still converts

Even averaged over the half, Meta's cheaper traffic wins on CPA.

Zoom back out to the whole window and the cost-base advantage holds where it counts. Meta's median cost per acquisition is $35 against Google's $48 — and it wins despite converting at a lower median rate (0.91% vs 2.61%). That's the lesson buried in the headline: cheap-but-mediocre traffic beats expensive-but-better traffic when the price gap is wide enough. Conversion rate is a tactic; cost base is the terrain.

Chapter 4 · The return gap

On the dollar that matters most, Meta's median nearly triples Google's.

Enough Meta accounts now report revenue to publish ROAS honestly — 38 qualify — and the typical Meta account returns 3.10× against Google's 1.18×. Notice the tails almost touch (Meta p90 12.85×, Google p90 12.98×): the best Google accounts are world-class. But the median Google account is barely above break-even, and as the monthly arc showed, falling. The platform isn't the problem. The distribution is.

Chapter 5 · Follow the money

A third of Google spend — $8.3M — is running underwater.

Medians describe the typical account; dollars describe the damage. Sort every revenue-tracked Google account into ROAS bands and weight by spend, and the picture turns stark: 34% of clean Google spend ($8.3M) sits in accounts returning under 1.0× — a literal loss — and 72% sits below 2.0×. Only 18% earns 4×+. Meta is the mirror image: 46% of its spend lives in accounts returning 4×+. The same operators, running both platforms, concentrate Meta budget on winners and let Google budget bleed.

Interlude · The money map

Trace every Google dollar: $17.6M flows into at-risk accounts, $6.4M into working ones.

Here is the same truth as a flow. Every clean Google dollar leaves the budget on the left, lands in a ROAS band in the middle, and is judged on the right. The two fattest ribbons — under-1× and 1–2× — carry $17.6M into accounts that are losing money or barely breaking even, while only $6.4M reaches the healthy and strong bands. We have no geography in the warehouse, so this maps where money goes by outcome, not by place — and the outcome is where the leak is.

Chapter 6 · The average hides the danger

Inside Google, the gap between weak and strong dwarfs the platform gap.

Here's why the median undersells the stakes. Google ROAS runs from 0.07× at the 25th percentile to 4.83× at the 75th — a spread wider than the entire Google-vs-Meta median difference. Whether you 'do well on Google' has almost nothing to do with choosing Google and almost everything to do with how the account is run. Performance is an execution problem wearing a platform costume.

Chapter 7 · Two operators

The losing quartile pays 2.5× the CPC of the winning quartile.

Split Google accounts into ROAS quartiles and the difference is behavioral, not magical. The bottom quartile returns 0.01× and pays a $3.15 median CPC; the top quartile returns 9.28× on a $1.24 CPC. The losers overbid into expensive auctions — and they over-concentrate: 91% of their budget sits in Search and uncategorized campaigns, with almost no Shopping or Performance Max structure. The winners spread budget across Search (27%), PMax (24%) and Shopping (22%). Discipline on bids, diversity in structure. That's the whole game.

Chapter 8 · The mechanism

Your CPC isn't set by the platform — it's set by the auction you enter.

Why does the losing quartile pay so much more? Because CPC is wildly path-dependent. Google CPC runs from $0.25 at the 10th percentile to $7.05 at the 90th — a 28× spread between accounts on the same platform, with Meta nearly as wide at 25×. The median is almost meaningless here. Bidding into expensive auctions without the conversion rate to support it is the single most common, most expensive mistake in the whole set — and the one most directly within an operator's control.

Chapter 9 · Where budget should route

Search remains Google's cleanest conversion engine.

If the winners diversify, what should they diversify into? By inferred campaign type, Search delivers a $41 CPA at 3.87% CVR across 108 accounts — the efficient core. Performance Max is close on CPA ($48) but converts at barely half the rate, while Shopping, Display and Video/Demand Gen are materially weaker. The problem was never Google. It's how budget gets routed inside Google.

Chapter 10 · The automation default

Performance Max is everywhere — and converting at half of Search.

76 of the ~150 Google accounts run Performance Max, nearly the whole base. Yet it converts at 1.82% versus Search's 3.87%, and costs ~18% more per conversion. PMax isn't bad — handing the machine a clean objective and good signals is exactly the agentic playbook. But it's being switched on as a default instead of chosen as a tool, and that quiet reflex is the single biggest optimization opportunity in the set.

Chapter 11 · The wrong-objective tax

Optimizing Meta for clicks costs $578 a conversion.

The same default reflex shows up on Meta as a brutal tax. Its strongest objectives are Leads ($22 CPA) and Sales ($30). But 44 accounts still optimize to Link Clicks, where CPA is $578 and conversion rate is effectively zero. Ask the algorithm to find clicks and it finds clicks; the buyers simply don't follow. This is the most fixable line in the entire report — graduate those accounts to Leads or Sales and the cost base does the rest.

Chapter 12 · The creative drop-off

Most video viewers who reach a quarter still never finish.

Creative theme labels are too under-instrumented to benchmark — ~96% of named creatives fall into a generic bucket — so the honest creative signal is completion decay. Of 28.4M viewers who reach 25%, only 8.1M finish: roughly 29% make it to the end, and the steepest fall is between 25% and 50%. The hook works; the hold doesn't. The next dollar of creative effort belongs in the middle of the video, not the first frame.

Chapter 13 · Who this reaches · Watt Data

The reachable market skews Technology, with SaaS as the dense core.

Switching sources for the final act: these bars are not performance. They come from the Watt Data audience graph and describe the addressable B2B universe behind these campaigns — the terrain every account above is competing over. By firmographic mix, Technology is the largest reachable segment, with SaaS, B2B services and CRM forming the high-intent core. The dollars chase a crowded, well-defined market.

Chapter 14 · Who this reaches · Watt Data

Most of the reachable market is light-traffic — the field is won on discipline, not scale.

Sized by digital footprint, the addressable universe is overwhelmingly small: 2.41M reachable businesses sit in the low website-traffic tier against just 183K high and 77K very-high. The same skew shows in paid behavior — far more businesses run low-to-medium monthly Google ad-spend (≈57K low, ≈86K medium) than run at scale. The takeaway mirrors the whole report: the bulk of the market is modest spend competing on execution, not a handful of giants. Context, not a performance ranking.

Unlock the full report

See who this market actually reaches

The remaining 8 chapters cover the firmographic, technographic, demographic and buying-intent makeup of the addressable B2B audience behind these campaigns. Enter your work email to keep reading — no spam, unsubscribe anytime.

Free · 8 more chapters · audience intelligence

Source: Synter connected accounts · TimescaleDB · H1 2026 (Jan 1 – Jun 13) · per-account medians, anonymized · audience context from Watt Data

The takeaway

It's not the platform. It's the discipline.

The story isn't that one platform wins and another loses. It's that agentic advertising only works when the system is pointed at the right economic target: the right objective, the right campaign type, the right bid, the right creative signal, and enough clean data to tell the difference. The skills below encode the fixes from this edition — audit PMax before it becomes default spend, kill link-click optimization, watch video decay, and route budget by account-level economics instead of platform folklore.

Next edition: geography and landing-page layers (not yet in the warehouse), and joining the Watt audience graph to account-level CPA and ROAS — so you can benchmark against the accounts that actually look like you.

All skills & data How we measure this

Synter connected accounts · TimescaleDB · H1 2026 (Jan 1 – Jun 13) · per-account medians, anonymized · audience context from Watt Data

FAQ

Common questions

Does Meta or Google deliver better ROAS in H1 2026?: Across revenue-tracked Synter-connected accounts in H1 2026, Meta's median return on ad spend reaches 3.1× and beats Google on cost base — median CPA of $35 vs $48 — despite a lower median conversion rate. Cheap-but-mediocre traffic wins when the price gap is wide enough.
Why is Performance Max called the over-adopted underperformer?: Performance Max is one of the most widely adopted Google campaign types, but at the account-level median it underperforms more targeted campaign types — so it quietly becomes default spend before it is audited.
How is this benchmark measured?: Per-account medians across every account connected to Synter over H1 2026 (Jan 1 – Jun 13), anonymized, with physically-impossible accounts excluded and revenue figures limited to accounts where ROAS is plausibly tracked.