miROAS (Marginal Incremental Return on Ad Spend)

miROAS answers a single forward-looking question: if one more dollar is invested in a channel, campaign, or ad set right now, how much additional revenue will it generate? Unlike iROAS (Incremental Return on Ad Spend), which is a backward-looking average over a historical period, miROAS is a point estimate that changes as spend changes — making it the correct metric for all spend allocation and reallocation decisions.

Google's Media Mix Modeling (MMM) framework Meridian uses the identical concept under the label mROI (marginal ROI). The terms are mathematically equivalent; the naming diverges between vendors.

---

Definition and formula

miROAS is defined as:

miROAS = ΔRevenue ÷ ΔSpend

where ΔRevenue is the incremental revenue generated by a small additional unit of spend, and ΔSpend is that additional unit. Formally, it is the derivative of the advertising Saturation Curve at the current spend level — the slope of the tangent at that point. (Sellforte, as-of 2025-09-17)

The "i" in miROAS is load-bearing. It indicates the revenue numerator is based on true incrementality — the causal, counterfactual lift from advertising — not on attributed or platform-reported revenue. A "marginal ROAS" (mROAS) calculated using click-attributed revenue will systematically overstate returns for lower-funnel channels, making the distinction practically significant for budget decisions. (Sellforte, 2025-09-22)

Google Meridian's definition of mROI: the incremental outcome from a small fractional increase δ (e.g. 0.01) in spend above the historical budget level, divided by δ × cost. (Google Meridian docs, as-of 2026-05-15)

---

How miROAS differs from iROAS

Dimension	iROAS	miROAS
Direction	Backward-looking	Forward-looking
Measures	Average return over a historical period	Return from the next additional dollar
Changes with spend?	No — fixed for a given period	Yes — declines as spend grows
Use case	Measuring past performance	All spend allocation and reallocation decisions

A channel can simultaneously exhibit a high iROAS and a low miROAS. This divergence is the saturation signal: the channel delivered strong returns historically, but at current spend levels it is near its Saturation Curve ceiling. (Sellforte, 2025-09-17)

Concrete example (Sellforte, 2025-09-17): Ad Set A has iROAS 5.0 but miROAS 3.0. Ad Set B has iROAS 6.0 but miROAS 0.5. Incrementing budget into Ad Set B returns only $0.50 per additional dollar despite its superior historical average. The correct reallocation is into Ad Set A — the lower-average but higher-marginal option.

The ROAS peak trap: iROAS (as an average) peaks later — at a higher spend level — than miROAS (the marginal). This creates a deceptive window during which average ROAS is still rising even as the marginal return has already started declining, giving misleading "room to grow" signals from in-platform reporting. (WorkMagic, 2026-02-10)

---

Diminishing returns and the saturation curve

Advertising performance follows the law of diminishing returns: as spend on a channel grows, each additional dollar yields less revenue due to audience saturation, limited inventory, and growing competition for placements. miROAS captures this non-linearity by moving as spend moves. (MuttData, 2025-06-30)

Ad spend response curves pass through three phases (mbuzz, 2026-04-04):

1. Accelerated returns — small spend, high efficiency; channel is underinvested
2. Linear returns — growing volume, decelerating efficiency
3. Plateau — additional budget no longer acquires meaningfully new customers

Modern Media Mix Modeling (MMM) tools use two mathematical transformations to model this: Adstock (carryover / lag effects of advertising) and Saturation (modeled via the Hill function). The Hill function formula: Revenue = K × spend^S / (spend^S + EC50^S), where EC50 is the half-saturation spend level and S controls curve shape — borrowed from pharmacology. Meta's open-source Robyn, Recast's causal MMM, Mutinex, and Google Meridian all fit Hill functions to weekly spend-and-revenue data. (Measured, as-of 2026-04-21; mbuzz, 2026-04-04)

Data minimum: at least 12 weeks of weekly spend-and-revenue data is required to reliably detect diminishing returns and fit a response curve. (mbuzz, 2026-04-04)

---

Budget optimisation: the equimarginal principle

The Equimarginal Principle governs optimal budget allocation: reallocate spend from channels with lower miROAS to channels with higher miROAS until miROAS converges across all channels. At that convergence point, no further reallocation can increase total incremental revenue. (Measured, 2026-04-21)

Measured illustrated this with a $50M DTC ecommerce budget scenario (as-of 2026-04-21):

Channel	Saturation level	Marginal ROI	Signal
Display	90%	1.2×	Heavily over-invested
Paid Search	85%	1.8×	Declining
Podcast	—	2.8×	Underinvested
Paid Social	70%	3.2×	Strong, room to grow
TV/CTV	40%	3.5×	Substantial room to grow

Projected reallocation outcome in the same scenario: +$16.3M incremental revenue (+13.4%) with zero additional total spend. (Measured, 2026-04-21)

Practical channel benchmarks from Measured's 274-experiment dataset (as-of 2025-08-27): CTV median incremental ROAS of $2.88, Google $2.39 (at 40.5% of budget), Meta $2.30 (at 32% of budget) — yet CTV accounts for only 3.5% of average enterprise brand media budgets.

When the primary goal is revenue maximisation (not efficiency), brands may intentionally operate beyond peak miROAS — continuing spend as long as the marginal revenue exceeds marginal cost after accounting for profit margins, COGS, and operating expenses. (WorkMagic, 2026-02-10)

---

How platforms and tools use miROAS

In-platform (Google, Meta)

Neither Google Ads nor Meta Ads natively surfaces miROAS in their campaign interfaces. In-platform ROAS figures are rule-based attribution numbers — Google claims last-click, Meta can claim the same conversions as other platforms. True miROAS must be derived externally using causal measurement methods: either Geo Holdout Testing / Conversion Lift experiments, or incrementality-calibrated Media Mix Modeling (MMM). (Sellforte, 2025-09-22; WorkMagic, 2026-02-10)

Operational saturation signals that can be read from in-platform data as leading indicators of low miROAS (mbuzz, 2026-04-04):

• Meta average frequency above 3× per week
• Google Search impression share plateauing at 80–90%

Google Meridian

Meridian's native planning tool for miROAS-style decisions is the "flexible budget scenario with target marginal ROI" — it determines the maximum spend on each channel that still meets a minimum target mROI. Response curves are generated by scaling historical spend up or down (e.g. 1.2×) while preserving the historical flighting pattern across time and geographies. (Google Meridian docs, as-of 2026-05-15)

Meridian caution: response curves require extrapolation for channels that have not varied spend historically; extrapolation risk increases significantly for spend levels well above historical. (Google Meridian docs, 2026-05-15)

Meridian's usage guidance:

• ROI: evaluate historical performance (backward-looking average)
• Response curves: visualise diminishing returns and optimise future budgets
• mROI: evaluate current saturation — if mROI is much lower than ROI, the channel is near saturation and new funds should go to higher-mROI channels

Third-party vendors

Sellforte Activate (launched 2026, as-of 2026-06-23): applies miROAS-based budget and bidding recommendations directly to Google Ads, Meta Ads, and TikTok without manual export. Delivers recommendations at campaign level for Google and ad set level for Meta Ads (across CBO, ABO, and hybrid setups). Each recommendation shows: the metric being changed (daily budget, bid value, or target ROAS), current vs recommended values, miROAS estimate, and expected daily incremental sales impact. Designed for brands with €50M+ annual media investment; as of 2026 claims >$1B annual media spend optimised across 22 markets.

Haus.io: provides iROAS measurement via always-on geo holdout experiments; as of the undated article, recommends 4–8 weeks minimum test duration for statistically reliable iROAS. Does not use the miROAS label but achieves the marginal concept implicitly by running experiments at different spend levels. Haus example (SaaS brand): scaling Google Ads spend from $10K to $100K/month kept traditional ROAS stable at ~3.0× while true iROAS fell from 2.5× to 1.2× — evidence of saturation and cannibalization of organic sign-ups. (Haus.io, undated)

Northbeam: launched "Incrementality by Northbeam" (2025–2026) — automated lift testing with continuous validation, unifying MTA and MMM signals; self-service incrementality tests were slated for Q1 2026. (Northbeam blog, undated)

---

Practical benchmarks: what is a good miROAS?

There is no universal miROAS benchmark or "right" threshold. The appropriate stopping point is determined by each business's profit margins, COGS, and operating expenses: the threshold is the miROAS level at which incremental revenue no longer exceeds incremental cost when those factors are applied. (WorkMagic, 2026-02-10)

One practical operationalisation: set an internal contribution ROAS floor (e.g. 1.80×), then scale spend until in-platform ROAS falls to that level — using ROAS as a rough proxy for a miROAS floor. This is a simplification, not a causal measure. (WorkMagic, 2026-02-10)

The Google Performance Max saturation effect: increasing PMax budget forces the algorithm to reach progressively less-relevant audiences, causing marginal returns to collapse below break-even even while the campaign-level average ROAS remains above target. This is a concrete example of high average ROAS coexisting with destructive marginal ROAS. (smec, 2026-02-27)

---

miROAS and MMM: the relationship

miROAS is an output of Media Mix Modeling (MMM), derived as the mathematical derivative of the saturation / response curve the model produces for each channel, campaign, or ad set. MMM tells you "what happened" (historical channel contribution); miROAS optimisation tells you "what to do next." The transition from the former to the latter is the critical shift from reporting to action. (Measured, 2026-04-21; Sellforte, 2025-09-17)

MMM models must be validated with incrementality testing — Geo Holdout Testing, Conversion Lift studies — to confirm that saturation curves reflect causal relationships rather than historical correlation. Without validation, optimised budget allocations carry significant uncertainty. (Measured, 2026-04-21)

Suggested cadence (vendor recommendation — reflects Measured product positioning): weekly monitoring of mROI trends and saturation signals; monthly budget adjustment; quarterly full model recalibration. (Measured, 2026-04-21)

---

Contradictions

---

Key terms

Term	Meaning
miROAS	Marginal Incremental ROAS — return on the next additional dollar spent
iROAS	Incremental ROAS — average causal return over a historical period
mROI	Google Meridian's label for the same concept as miROAS
mROAS	Marginal ROAS without the incrementality requirement — may overstate returns
Saturation curve	The non-linear response curve of revenue to spend; fits a Hill function
EC50	Half-saturation spend level in the Hill function — the spend at which revenue is half its maximum
Equimarginal principle	Optimal allocation: miROAS should be equal across all channels
Adstock	Carryover / decay effect: advertising from past periods continues to affect current sales

---

What practitioners report

• Sellforte (2025-09-17): A channel at 90% saturation can show average ROAS of 4× while miROAS is 0.6× — the next dollar spent returns only 60 cents.
• Haus.io (undated): Traditional ROAS held at 3.0× while iROAS dropped from 2.5× to 1.2× as a SaaS brand scaled Google Ads spend 10× — saturation and organic cannibalization.
• Measured (2026-04-21): In the $50M budget model, simply reallocating (no new spend) projected +$16.3M revenue. The constraint is not budget; it is spending at a point past saturation in high-efficiency channels (Paid Search) while under-spending lower-saturation channels (TV/CTV).
• Mbuzz (2026-04-04): Outdoor Solar Outlet cut Google Ads spend by 26%, saw revenue rise 25% by reallocating to Microsoft Ads and social retargeting still in the high-efficiency phase.

---

Next frontier (dangling links)

The following concepts are referenced in this page but have no dedicated vault entry yet: Diminishing Returns · Equimarginal Principle · Saturation Curve · Adstock · Budget Optimisation · Ghost Ads · Hill Function · Robyn (MMM)