AI Deployment
Feature 02  ·  Automation  ·  Edition Q1 2026

Google's automation won.
Meta's lost half its share.

Google's automated shopping product now takes two-thirds of Shopping budgets. Meta's equivalent has lost half its share in twelve months. Same promise, same quarter, opposite verdicts - and the incrementality evidence explains which advertisers changed their minds and why.

For three years the industry argument about campaign automation has been conducted as a single question: is the black box good or bad. It is the wrong shape of question, and Q1 2026 demonstrates why with unusual clarity.

Two of the largest advertising platforms sell essentially the same proposition. Hand over targeting, placement and bidding. Supply creative and a product feed. Let the system allocate. Google calls it Performance Max. Meta calls it Advantage+ Shopping Campaigns. The pitch is close to identical and the underlying machinery is comparable.

In the same quarter, advertiser behaviour toward the two products moved in opposite directions - and not marginally.

Performance Max took 67% of Google Shopping spend. Advantage+ fell to 20% of retail Meta spend, from a 38% peak twelve months earlier.

One product consolidated its position. The other lost roughly half its budget share inside a year. Any explanation resting on "advertisers are warming to automation" or "advertisers are rejecting automation" fails immediately, because both happened simultaneously to two versions of the same idea.

The divergence, quarter by quarter

The Advantage+ decline is not a single bad quarter. It is a consistent three-quarter trajectory: 38% of US retail Meta spend in Q1 2025, 27% in Q4 2025, 20% in Q1 2026.1,2

Performance Max moved the other way and then held. It accounted for 67% of Google Shopping ad spend in Q1 2026 and 68% of Shopping sales - a share slightly above its spend share, which is the first useful signal in this dataset.

Figure 01
Automated campaign share of platform spend
Performance Max as share of Google Shopping spend; Advantage+ Shopping as share of US retail Meta spend.
Grouped bar chart: automated campaign share of platform spend across Q1 2025, Q4 2025 and Q1 2026.
Source: Tinuiti Digital Ads Benchmark Report, Q1 2026, and Q1 2025 edition. Performance Max share is reported for Q1 2026 only in the material available to us; earlier quarters are not shown rather than estimated. The two series measure different denominators - Google Shopping spend versus US retail Meta spend - and are plotted together to compare direction, not level.

What each platform's own numbers say

On Google's side the reported case is straightforward. Among advertisers running both Performance Max and Standard Shopping, Performance Max delivered slightly better return on ad spend, and its 68% share of sales against 67% of spend indicates it is not simply absorbing budget without producing proportionate revenue.

That comparison has a real virtue: it is a within-advertiser comparison. The same brands, in the same quarter, running both campaign types, with PMax coming out marginally ahead. It is not a controlled experiment, but it is considerably better than comparing adopters to non-adopters.

On Meta's side the reported case is, on its face, stronger still. Agency-reported figures put Advantage+ Shopping at roughly 4.5× ROAS against 3.7× for manually configured campaigns - a lift in the 15–25% range, with materially lower cost per acquisition across ecommerce verticals.3

Which produces the puzzle. If Advantage+ reports a 20%-ish ROAS advantage, why did advertisers cut its share of their budgets by nearly half in a year?

The evidence that resolves it

The most useful counterweight comes from incrementality testing rather than platform reporting. Haus ran 640 incrementality tests over eighteen months across mid-market and enterprise brands averaging roughly $1M per month in Meta spend.4

The result inverts the reported picture. Advantage+ Shopping outperformed manual campaigns in 42% of head-to-heads. Manual won 58%.

Figure 02
Reported advantage versus tested advantage
The same product, measured two ways, produces opposite conclusions.
What platform-attributed reporting shows
4.5× vs 3.7×
Advantage+ Shopping ROAS against manually configured campaigns - a 15–25% reported lift, with around 32% lower cost per acquisition.
Agency-reported · vendor-adjacent · lowest source tier
What incrementality testing shows
42% of 640 tests
Share of head-to-head incrementality tests in which Advantage+ beat manual. Manual won the remaining 58%.
Haus · 640 tests · 18 months · ~$1M/mo brands
58%Manual won
42%Advantage+ won
Source: Haus incrementality study as reported in secondary coverage; ROAS comparison figures are agency-reported and should be treated as the weakest evidence tier in this feature. We have not seen the Haus methodology in full and cannot independently verify test design, brand mix or how ties were classified.

This is the same gap that Feature 01 of this edition described in a different context: the number the platform reports and the number the business recognises are measuring different things, and they can point in opposite directions without either being fabricated.

Why attributed ROAS flatters automated campaigns structurally

The mechanism is not mysterious, and it is not evidence of bad faith by the platforms. Broad automated campaigns are given permission to buy across the full funnel, including audiences that were already going to convert - existing customers, recent site visitors, branded searchers. Those conversions are cheap to win and they are attributed to the campaign that touched them last.

A manual campaign structure that deliberately excludes those audiences will report a worse ROAS while potentially producing more incremental revenue. It is being penalised in the report for the discipline that makes it valuable.

Incrementality testing exists precisely to separate these. When the test disagrees with the report, the test is measuring the thing the business cares about.

Why the two products diverged

If the attributed-versus-incremental gap applies to automation generally, why did Performance Max hold its share while Advantage+ lost half of its?

Three structural differences are visible in the data, and we can support two of them.

1. The comparison set is not the same

Performance Max's competitor inside Google Shopping is Standard Shopping - itself a feed-driven, largely automated format with limited manual levers. The gap in operator control between the two is narrow.

Advantage+'s competitor on Meta is a manual campaign structure with genuine control over audience exclusions, placement, budget splits and creative-to-audience mapping. The gap in control is wide. An advertiser moving from Standard Shopping to PMax gives up little. An advertiser moving from manual Meta to Advantage+ gives up a great deal - including the ability to exclude existing customers, which is exactly the lever that separates attributed from incremental performance.

2. Google's auction got cheaper. Meta's advertisers got tested

Q1 2026 was a benign quarter for Google search economics. Paid search CPCs were flat while clicks grew 14%, and Shopping CPCs were flat on 18% click growth. When unit costs are not rising, the cost of leaving a campaign type alone is low, and the incentive to audit it is weak.

Meta's advertisers, by contrast, are the population where incrementality testing has spread fastest - driven by the post-ATT measurement gap and by the arrival of affordable geo-test tooling. The Haus sample is instructive here: brands averaging around $1M a month in Meta spend are precisely the cohort that can afford to run 640 tests and act on the results.

Advantage+ did not lose share because automation stopped working. It lost share in the segment that acquired the ability to check.

3. The prospecting problem

The third difference is the one we can least evidence and flag accordingly. New-customer acquisition economics are where automated Meta campaigns are most frequently reported to underperform, because the system optimises toward conversion probability and existing customers are the highest-probability converters available.

We have no dataset that isolates this cleanly, and we are not asserting it. We note it because it is the mechanism most consistent with both the incrementality result and the direction of advertiser behaviour, and because it is testable in any individual account within a quarter.

The two products, side by side

Figure 03
Performance Max and Advantage+ Shopping compared
Q1 2026 position and the structural differences behind it.
 Performance MaxAdvantage+ Shopping
Share of relevant spend67% of Google Shopping20% of US retail Meta
Direction, 12 monthsHeld / consolidated38% → 27% → 20%
Share of sales68% - above spend shareNot separately reported
Platform-reported performanceSlightly better ROAS than Standard Shopping, within-advertiser~4.5× vs ~3.7× manual (agency-reported)
Incrementality-tested performanceNo comparable public test setWon 42% of 640 head-to-heads
What the alternative offersStandard Shopping - also largely automatedManual - full audience exclusion control
Control surrenderedNarrow gapWide gap
Auction pressure, Q1 2026CPCs flatCPM −3%
Source: Tinuiti Q1 2026 for share, sales share and auction figures. Haus for the incrementality result. Agency-reported for the ROAS comparison. The empty cell is genuinely empty - we found no public incrementality test set of comparable scale for Performance Max, and did not substitute a weaker source to fill the row.

What this means operationally

Automation share is not a strategy signal. The fact that 67% of Shopping spend runs through Performance Max tells you what the market does, not what your account should do. Herd data is a prompt to test, never a substitute for testing.

The control you give up matters more than the automation you gain. The clearest difference between these two products is not sophistication - it is what the advertiser can still exclude. Where an automated product removes your ability to exclude existing customers, expect reported performance to improve and incremental performance to deteriorate, and plan the measurement accordingly.

If you cannot test, weight the reported number down. Not to zero. Platform-attributed ROAS is a real measurement of a real thing; it is simply measuring last touch across a funnel the campaign was permitted to buy across. Where no incrementality capability exists, the honest position is that automated campaign performance is unresolved, rather than good.

Re-run the decision annually, not once. Advantage+ share fell across three consecutive quarters. Advertisers who evaluated it in Q1 2025 and never revisited are running a conclusion drawn in a different auction, at different creative volumes, with different measurement tooling available.

The test worth running this quarter

For any advertiser with meaningful Meta spend, one design settles the local version of this question: run the automated campaign with existing-customer exclusions applied, against the same campaign without them, and measure on incremental new-customer acquisition rather than blended ROAS. If reported ROAS falls while incremental new customers rise, the exclusion was doing work that the report was punishing.

That test is not free, but it is cheaper than a year of budget allocated on a number that was measuring the wrong thing.

How we did this

Where this comes from
Three tiers, weighted differently. Straight from the source: Tinuiti Q1 2026 benchmark data - their own advertiser sample, directly reported. A named study, reported by someone else: the Haus incrementality study, a named study with stated scale, reported through secondary coverage. From a company that sells into this market: agency-reported ROAS comparisons, which are vendor-adjacent and treated as the weakest evidence here.
Sample
Tinuiti figures reflect their advertiser base, weighted toward US retail. The Haus set is mid-market and enterprise brands averaging around $1M per month in Meta spend - not representative of smaller advertisers, for whom the economics of both automation and testing differ.
Period
Q1 2026, with Q1 2025 and Q4 2025 for the Advantage+ trajectory. The Haus tests span eighteen months and are not confined to this quarter.
Comparability
Performance Max share is measured against Google Shopping spend; Advantage+ against US retail Meta spend. Different denominators. We compare direction of travel, not absolute level, and Figure 01 states this on the chart.
Not verified
We have not reviewed the Haus methodology in full and cannot confirm test design, brand mix or tie handling. The result is reported because its scale and specificity make it the strongest available counterweight to platform-attributed figures, not because we have audited it.

What this doesn't prove

  • That Performance Max is better than Advantage+. No public test set compares them, and they operate on different platforms against different alternatives. Nothing here supports a cross-platform ranking.
  • That automation underperforms. Manual won 58% of the Haus head-to-heads, which means automation won 42%. That is a meaningful minority, not a rout, and account-level factors likely determine which side a given brand lands on.
  • Why Advantage+ share fell. The decline is reported and consistent. Our incrementality explanation is the most plausible mechanism we can evidence, but no source states the cause, and creative fatigue, setup quality and conversion-volume thresholds are all cited elsewhere as contributing factors.
  • Anything about the prospecting mechanism. Section three of the divergence analysis is explicitly flagged as unevidenced. It is a hypothesis worth testing in-account, not a finding.
  • Anything for advertisers below roughly $1M/month on Meta. The incrementality sample sits well above most accounts. Smaller advertisers face different conversion-volume constraints that plausibly change the answer.

Sources for this feature

  1. Tinuiti, Digital Ads Benchmark Report, Q1 2026. tinuiti.com Straight from the source
  2. Tinuiti, Digital Ads Benchmark Report, Q1 2025. tinuiti.com Straight from the source
  3. Agency-reported Advantage+ ROAS comparisons, as summarised across practitioner coverage, 2026. From a company that sells into this market - vendor-adjacent
  4. Haus, incrementality test programme - 640 tests, 18 months, reported via secondary coverage, 2026. A named study, reported by someone else - named study, methodology unreviewed
  5. Karooya, Digital Ads Benchmark Report by Tinuiti, Q1 2026: Key Highlights. karooya.com Secondary
AL
The practice behind this desk

Ads Legendary

We run the exclusion test described above as part of onboarding, before recommending any change to automated campaign allocation. Paid media inside your accounts, at 6% of media spend.