Benchmark report · 2026
Australian paid advertising,
in real numbers.
Aggregated benchmark data from the engagements we run across Australian operators in the $2M to $20M revenue band. Median ROAS by industry, where channel spend is actually flowing, the attribution leaks we measure on every audit, and Event Match Quality benchmarks for Meta CAPI.
Published 1 April 2026 · Updated 29 April 2026
CC BY 4.0 · Share with attribution
Headline metrics
The six numbers worth quoting.
Median monthly ad spend (our client base)
$72K AUD
Across DTC, lead-gen, B2B, automotive, and health categories.
Typical attributable revenue recovered post-rebuild
8% to 18%
Year one. Combination of server-side rebuild, CAPI, consent, and offline conversion uploads.
Median Event Match Quality before audit
4.1 / 10
Most untouched setups score in the 3.0 to 5.0 band.
Median Event Match Quality after rebuild
7.9 / 10
After six to eight weeks of clean data on the new server-side stack.
Typical Meta-vs-CRM revenue gap before reconciliation
23%
Median across our last 18 audits. Worst case observed: 51%.
Typical Meta-vs-CRM revenue gap after rebuild
Under 5%
The threshold we will not cut over below.
ROAS by industry
Median and good ROAS, by category.
"Good" is the 75th percentile across our client base. Both numbers are platform-attributed ROAS, not contribution-margin ROAS. Use the Ad Spend Calculator to translate either into the break-even ROAS your business actually needs.
| Industry | Median ROAS | Good ROAS (p75) | Sample share |
|---|---|---|---|
| Direct-to-consumer ecommerce | 2.7x | 4.2x | ≈40% of our client base |
| Lead-generation services (home, professional) | 3.4x | 5.1x | ≈25% |
| B2B SaaS with paid acquisition | 2.1x | 3.6x | ≈15% |
| Automotive services and trades | 3.9x | 5.8x | ≈10% |
| Health, wellness, and education | 2.5x | 4.0x | ≈10% |
Channel mix
Where the $72K-a-month median spend actually goes.
Share of paid spend by channel across the client base. Meta is still dominant but slipping. Google PMax and TikTok are taking most of the share that's leaving.
Meta
42%Down from ~52% in 2023. Still the dominant channel for most ecommerce, but the share is slowly bleeding to TikTok and Google.
Google Search
28%Stable for high-intent categories. The most reliable single channel for lead-gen services and trades.
Google PMax
11%Climbing fast in the last 18 months. Good for ecommerce with strong shopping feeds. Bad for brands that need to control where the spend goes.
TikTok Ads
9%Up from ~3% in 2023. Strong for DTC and education categories. Reporting maturity still trails Meta.
LinkedIn
5%Concentrated in B2B SaaS. Expensive per click; conversion economics work for high-ACV deals only.
Other (Pinterest, X, Reddit, niche)
5%Long tail. Worth testing for category-specific brands; rarely a primary channel.
Attribution leaks
Where the attributable revenue actually goes missing.
The five categories of leak we measure on almost every audit. Each percentage is what's typically lost from platform-reported revenue against the CRM. Compounded across a quarter, the five routinely distort reporting by 14 to 38 percent.
01
Browser-side pixel loss to ad blockers and ITP
The headline iOS 14.5 effect, plus Safari ITP and ad blockers. Server-side tagging fixes most of it.
Typical leak
8% to 22%
02
GA4 misconfigurations (page-view conversions, currency mismatch, internal traffic)
Compounded across the seven errors we see most often. Audit-recoverable.
Typical leak
5% to 18%
03
Meta CAPI deduplication failures
Either undercounting (missing event ID) or double-counting (ID collisions). Easy to spot, hard to fix without server-side.
Typical leak
3% to 11%
04
Cross-domain handoff to checkout / payment processor
Stripe, PayPal, Afterpay, Zip, eWAY. Referral exclusions plus client ID propagation closes the gap.
Typical leak
4% to 15%
05
Offline conversion not uploaded back to ad platforms
Lead-gen and B2B specific. The single biggest leak when the conversion happens off-platform.
Typical leak
Often 100% of post-call revenue
Frequently asked
Five questions about the methodology.
Where do these benchmarks come from?
Aggregated, anonymised, and rounded from the engagements Profit Geeks runs across Australian operators in the $2M to $20M revenue band. Cross-checked against publicly available platform data (Meta Q1 transparency reports, Google Ads benchmarks) where it overlaps. Numbers are deliberately rounded to ranges, not point estimates, because attribution data does not deserve more precision than that.
What is a good ROAS for ecommerce in 2026?
Median DTC ecommerce ROAS in our client base is 2.7x. A 'good' ROAS for the same category is 4.2x. Both numbers depend on gross margin: a 50 percent margin business needs roughly 2.86x to break even after overheads, a 30 percent margin business needs 4.0x or higher. Use the Ad Spend Calculator to check against your own margin structure.
Why is Meta's share of paid spend declining?
Two reasons. First, post-iOS 14.5 the platform-side return per dollar dropped sharply for many advertisers and they reallocated some spend. Second, TikTok and Google PMax both grew quickly enough to take share. Meta is still the dominant channel for most ecommerce, but the share has slid from roughly 52 percent in 2023 to 42 percent now in our client base.
What is Event Match Quality and why does it matter?
Meta's score (0 to 10) for how well your conversion events can be matched back to a Meta user. Higher EMQ means better attribution and better algorithmic optimisation. Median EMQ in untouched setups is 4.1; after a server-side rebuild we typically see 7.9. The lift is the largest single source of recovered attributable revenue in most engagements.
How much attributable revenue can a server-side rebuild typically recover?
Eight to 18 percent in year one for most operators in the $2M to $20M band. The recovery is the sum of three things: ad-blocker and ITP bypass via server-side, CAPI match quality lift, and recovery of offline conversion uploads from the CRM. The exact figure depends on starting state.
Cite or use the data
Free to share with attribution.
The benchmarks above are released under Creative Commons BY 4.0. Quote them, embed them, build on them. Link back to this page so readers can see the methodology.