Diminishing Returns in Ad Spend: The Marginal ROAS Hiding in Your Average

Why Your Best Channel Might Be Your Most Wasteful

Your Google Ads dashboard shows 4x ROAS. Your team celebrates. You increase budget by 30%.

Six weeks later, ROAS has dropped to 2.8x. Still positive, so you hold steady. Another month, 2.3x. Still profitable in aggregate. But something changed.

What happened: you hit the point of diminishing returns. The first $10,000 per month returned $50,000 in revenue. The next $5,000 returned only $8,000. Your average ROAS of 2.8x is hiding the fact that your marginal ROAS — the return on the last dollars spent — dropped below 1.0.

You're spending money to lose money, and average ROAS is concealing it.

Dropbox engineers reported in IEEE Access (Chivukula et al., 2026) that click attribution overstates causal impact by 2 to 10x compared to geo-incrementality experiments. The channel ROAS your dashboard shows is already inflated before diminishing returns make it worse. Average ROAS smears the bad marginal dollars together with the good ones, and platform self-reporting over-credits the whole pile.

The Three Phases of Ad Spend

Every channel follows the same pattern:

Phase 1: Accelerated Returns (High Efficiency)

At low spend levels, you're reaching the easiest-to-convert audience. Your first impressions go to people most likely to buy. Cost per conversion is low. ROAS is highest.

This is where small businesses live — and it's glorious. Every dollar works hard.

Phase 2: Linear Returns (Declining Efficiency)

As spend increases, you've reached the easy audience. Now you're expanding into less qualified segments. Conversion rates drop slightly. Cost per conversion rises. Revenue still grows, but slower than spend.

Average ROAS is still healthy, but marginal ROAS is declining.

Phase 3: Plateau (Saturation)

You've reached most of the addressable audience on this channel with this targeting. Additional budget buys more impressions but not more conversions. Frequency climbs. The same people see your ads 5, 8, 12 times.

Revenue flattens. Spend keeps rising. Marginal ROAS drops below 1.0.

Eric Seufert at Mobile Dev Memo has documented this dynamic in mobile user acquisition for over a decade: audience pools are finite. Once you've reached everyone who can be efficiently converted at a given CPM, additional impressions buy you the same person twice rather than a new buyer.

Average ROAS vs Marginal ROAS: The Metric That Changes Everything

Avinash Kaushik — formerly Google's Digital Marketing Evangelist, now strategy advisor to Measured — frames this as the one ratio a CMO should defend to a CFO: cost per incremental sale, not average ROAS. The "incremental" qualifier kills the average-ROAS illusion. Only the next dollar's outcome matters for the next decision.

The dangerous zone is "at saturation" — where average ROAS looks healthy enough that nobody questions the budget, but marginal ROAS has collapsed. Most companies operate here without knowing it.

How Response Curves Work

The shape isn't ad-specific. Archibald Hill described it in 1910 to model how oxygen binds to hemoglobin. The same S-curve describes ad spend saturation, dose-response in pharmacology, and population growth in ecology. Meta's open-source MMM library Robyn, Recast's causal MMM, and Mutinex's marketing modeling all fit Hill functions to weekly spend data. The math is settled. What changes by tool is which curve fits your channels.

Mathematically, diminishing returns follow a response curve. The one mbuzz uses is the Hill function:

Revenue = K × spend^S / (spend^S + EC50^S)

Where:
- K = saturation point (maximum revenue the channel can produce)
- S = curve shape (how steeply returns decline)
- EC50 = the spend level at 50% of maximum revenue

You don't need to understand the math. What matters is the shape: revenue rises steeply at first, then bends, then flattens. The derivative (slope) at your current spend level is your marginal ROAS.

The curve is fitted by looking at 12+ weeks of weekly spend and revenue data. Each week is a data point. The model finds the curve that best explains the relationship between what you spent and what you earned.

How to Spot You're Past the Point

You don't need a response curve model to see the warning signs. Watch for these:

Rising CPA at Flat or Increasing Spend

If your cost per acquisition is trending up while spend is stable or growing, you're entering Phase 2 or 3. The channel is working harder for the same results.

Climbing Frequency Metrics

• Meta: Average frequency above 3x per week means the same people are seeing your ads repeatedly. Fresh audience pools are exhausted.
• Google: Impression share plateauing at 80-90% means you're already visible for most relevant searches. More budget won't create more searches.

mROAS Below 1.0

If you can calculate marginal ROAS (via response curves or by comparing incremental spend to incremental revenue week-over-week), a reading below 1.0 is definitive. You're past the point.

New Customer Acquisition Flattening

If total conversions are growing but new customer acquisition is flat, your additional spend is just converting the same people repeatedly (retargeting, remarketing). The channel isn't expanding your customer base anymore.

What to Do When You Hit the Ceiling

Diminishing returns doesn't always mean "cut budget." Four options:

1. Shift Budget to an Under-Invested Channel

If Google Search is saturated and Meta is in Phase 1, the marginal dollar produces more on Meta. This is what reallocation is: moving money from low-marginal-return channels to high-marginal-return channels.

Shift 10-15% at a time. Not 50%. The algorithm tax applies.

2. Find New Audiences on the Same Channel

You've saturated your current targeting. But the channel has more audiences. Try:
- Broader geographic targeting
- Lookalike/similar audiences at wider percentages
- Different demographics or interest segments
- Upper-funnel vs lower-funnel targeting

Each new audience gives you a fresh response curve. The channel isn't saturated — your targeting is.

3. Change Funnel Stage

If you've saturated retargeting (warm audience, high frequency), try prospecting (cold audience, lower frequency). The curve resets because you're reaching different people.

Conversely, if you've saturated prospecting, adding retargeting captures more of the demand you've already created.

4. Refresh Creative and Messaging

New creative — different hooks, headlines, or positioning angles — can pull more conversions from an audience that looked tapped out. Same people, different value proposition. Teams without in-house copy often bring in a specialist SaaS copywriter to generate angle variations faster than internal iterations.

5. Reduce and Reallocate

If none of the above apply — you've tried new audiences, new funnel stages, and the channel is genuinely maxed out — reduce spend to the efficient zone and put the savings elsewhere.

The efficient zone is where marginal ROAS is between 1.0x and 1.5x. You're still getting positive return on every dollar, but you're not wasting money past the saturation point.

The Case Studies

FanDuel: Tested 20%, 40%, and 60% of budget on the same audience. 40% was most efficient — beyond that, returns diminished significantly. The 60% cohort produced only marginally more conversions at 50% more cost.

Outdoor Solar Outlet: Cut Google Ads spend by 26%. Revenue increased 25%. How? They reallocated the savings to Microsoft Ads and social retargeting — channels that were under-invested and still in Phase 1. Less total spend, more total revenue.

Dropbox (IEEE Access, 2026): Replaced click-attribution-led budget allocation with a geo-incrementality framework on its Individual Account team. Reallocated $25M of annual paid spend. Reported a 53% improvement in LTV:CAC on that team. Bad attribution doesn't only misallocate at the margin. At enterprise scale, it misallocates millions before the dashboard finishes loading.

All three cases demonstrate the same principle: the goal isn't maximum spend. It's optimal spend.

Related Reading

• How to Reallocate Marketing Budget Using Attribution — the full reallocation process using marginal ROAS
• When to Change Your Marketing Budget — the decision framework for when to act on diminishing returns
• The Algorithm Tax — what budget changes cost you in learning phase disruption
• How Much Are Your Ad Platforms Over-Reporting? — why platform ROAS is already inflated before diminishing returns make it worse

Holly Henderson

Co-Founder, mbuzz

Holly Henderson is Co-Founder of mbuzz. With 10+ years in marketing including roles at Westpac, Avon, and Forebrite, she's obsessed with making measurement actually useful.

Harvard Extension School Forebrite Westpac Avon