ROAS (Return on Ad Spend)

ROAS is the headline efficiency metric of paid ecommerce media: revenue divided by ad spend. It is the most-referenced dangling concept across the run-103→106 measurement cluster (Retail Media → Incrementality → Media Mix Modeling (MMM) → Multi-Touch Attribution (MTA)), and the metric all of those methods exist to correct. The recurring theme across every source this run is that reported ROAS systematically overstates the commercial value of advertising — because it ignores profit (the POAS critique), double-counts revenue across platforms (the MER / blended critique), and credits sales that would have happened anyway (the Incrementality critique).

Firewall: every claim is what a source reports. This was a web-only, vendor-skewed run — almost all benchmarks come from agencies, ad tools, or measurement platforms with a commercial interest in the framing. See ../../CONTEXT.md Rule 1.

The three sibling metrics

Metric	Formula	What it answers
ROAS	revenue ÷ ad spend (per channel)	How much revenue did this channel report per £ spent?
MER (blended ROAS)	total revenue ÷ total marketing spend (all channels)	Is the whole paid program pulling its weight?
POAS	(Revenue − COGS − Shipping − Returns − Payment fees − Discounts) ÷ ad spend	Did the ad spend actually make money?
iROAS	incremental revenue ÷ ad spend	How much revenue did the ads actually cause?

JudeLuxe's recommended usage split: ROAS for tactical within-channel bidding, MER to sanity-check blended investment, POAS for the profit truth (JudeLuxe, 2026-05-25). POAS is a registered trademark of ProfitMetrics.

Break-even & target ROAS

The one piece of non-contested, non-volatile knowledge this run:

Break-even ROAS = 1 ÷ gross profit margin.

Gross margin	Break-even ROAS	Target (+20% buffer)
60%	1.67×	2.00×
50%	2.00×	2.40×
40%	2.50×	3.00×
35%	2.86×	3.43×
25%	4.00×	4.80×
10%	10.0×	—

(Midsummer/Contrast aggregate + AdAmigo, 2026.) Break-even ROAS is the floor, not the goal; a common rule sets target ROAS = break-even × ~1.3–1.5 to fund operating costs and profit (Contrast.digital, 2026). Fashion typically runs 35–50% gross margin, with net margin also absorbing fulfilment (8–15% of revenue) and payment fees (2.9–3.5%) — so headline ROAS overstates true profitability before Returns Management|returns are even counted. Raising AOV lifts ROAS without touching conversion rate ($45→$65 AOV ≈ +44% ROAS; AdAmigo, 2026).

Benchmarks (as-of 2026-06-27)

All vendor-sourced and volatile — directional only.

Overall ecommerce: average ~2.87:1, median 2.04:1 — half of businesses below 2:1 (Hawky.ai, snippet, low confidence).
By channel: Google Search ~4.5:1 / Shopping ~5.0:1; Meta 2.5–4.0×; TikTok ~1.4× (Hawky.ai, low confidence).
Fashion/apparel (Meta): median 2.18×, top 25% 4.4×, top 10% 6.0× (MHI Media analysis of 1,247 Meta accounts via AdAmigo).
Fashion by platform: Google 3.40–4.48×, Meta 2.18–2.90×, TikTok 1.69–2.80× (AdAmigo — note internal range-vs-median inconsistency).
Retail media: aggregate 6.1× held for five consecutive quarters (Skai Q1 2025); grocery RMN — Instacart 4.8–6.7×, Kroger 4.5–6.2× (Osmos, 2026, low confidence).

Market context: Meta CPMs rose ~~19.2% YoY in 2025; US CPMs (~~$20.48) ≈ 2× UK (~$10.85), so achievable ROAS varies sharply by market (AdAmigo, 2026).

The overstatement problem (why the other metrics exist)

Three distinct effects, often conflated — see the contradiction note below:

Profit blindness → POAS. A worked fashion example collapsed a 3.76× ROAS to a 1.55× POAS once COGS (45%), shipping, payment fees, discounts and returns (32% of orders) were deducted — £93k of costs ROAS never showed on £42k spend (JudeLuxe, 2026-05-25, illustrative, single agency).
Cross-channel double-counting → MER / blended ROAS. A single order can be claimed by Google, Meta and TikTok at once; platform-reported total ROAS is routinely 2–3× higher than blended ROAS from actual Shopify orders, and Meta alone is cited as overstating ~28% (Polar Analytics / Karbon Analytics, 2026, vendors, snippet).
Non-incremental credit → Incrementality. A channel can show 6× platform ROAS while nearly non-incremental; measured iROAS is typically 30–60% below platform-reported ROAS, worst for branded search / brand-defence. Geo-lift / matched-market testing is the preferred remedy (Measured, 2026-04-15; DigitalApplied, 2026, vendors).

Measured's iROAS decision bands

[!unverified] Vendor framework (Measured), not an industry standard.

3.0 strong (scale) · 1.5–3.0 profitable (maintain) · 1.0–1.5 marginal · <1.0 unprofitable (pause).

Contradictions

The 2.87 average-ROAS figure is dated inconsistently — Hawky.ai labels it "2026"; Upcounting frames it as "Dropped to 2.87 in 2025." Treat as a 2025–26 snapshot, not precisely year-stamped.

Magnitude of platform overstatement varies by source and measures different things — "~28%" (Meta per-channel inflation, Polar) vs "30–100% over-count when summed" (cross-channel double-counting) vs "1.5×–3× / 30–60%" (incrementality gap, Measured). These are three distinct effects, not one disputed number.

Key terms

Term	Meaning
ROAS	Revenue ÷ ad spend
iROAS	Incremental revenue ÷ ad spend (causal)
MER	Marketing Efficiency Ratio = total revenue ÷ total marketing spend (blended ROAS)
POAS	Profit on Ad Spend = contribution margin ÷ ad spend (ProfitMetrics TM)
Break-even ROAS	1 ÷ gross profit margin — the floor where ad spend covers cost
Blended ROAS	Total revenue ÷ total ad spend; avoids per-channel over-credit

What practitioners report

[!unverified] Reddit MCP not connected this run (tool_uses: 0) and the YouTube/Apify transcript actor was unavailable — no practitioner counter-narrative was collected. The "platform ROAS is a lie" / real-world target-ROAS debate remains a gap. Candidate videos logged for re-fetch: Tinuiti "Beyond the Click: Holistic ROAS" (3vf276mzQDI, 2026), "Daily Profit Beats ROAS" (wS_17fT0R-M, 2025).