Your ROAS is still above target. The account is still producing customers. Nothing looks broken. So you raise the budget again. That’s exactly where an average can mislead you: it tells you how the whole program has performed, not whether the next block of spend deserves another dollar.
This article is about that next block — how to tell when more ad spend stops being a good decision, even while the dashboard still looks healthy.
Stop increasing spend when the next increment no longer brings in customers at economics the business wants to accept, or when inventory, capacity, cash or measurement makes more demand a poor use of capital.
Blended ROAS can’t tell you where that point is. It’s a backward-looking average of everything you’ve spent. The question is what the most recent increase actually bought: its marginal CAC and the contribution those added customers leave. There’s no universal ROAS at which to stop.
In this article
What’s the difference between average CAC and marginal CAC?
Average CAC tells you what customers cost across the whole program. Marginal CAC tells you what the added customers from your latest increase cost. Scaling decisions are about the second.
- Blended (average) CAC
- Total acquisition spend ÷ total new customersWhat did customers cost us across the whole current program?
- Marginal CAC
- Additional spend ÷ additional new customersWhat did the added customers from the latest increase cost us?
This is ordinary economics: most decisions are made on the margin, a little more or a little less, and money already spent shouldn’t drive the next choice. One caution belongs here, though. Ordinary reports can’t prove how many of the added customers the extra spend actually caused. In practice you estimate marginal CAC from a controlled budget change, first-party new-customer counts and cohort data. Call it an estimated or practical marginal CAC, and don’t treat it as more precise than it is.
Block width = customers; height = CAC; so each block’s area is its spend.
- Core spend$50,000 · 1,250 customers
$40.00 CAC
- Next increment+$20,000 · +350 customers
$57.14 marginal CAC
- Blended account$70,000 · 1,600 customers
$43.75 CAC
Can ROAS still look good while I’m overspending?
Yes. A blended ROAS can stay attractive while the newest spend is far less efficient.
- Base$50,000 → $250,000 attributed
5.0x
- Newest increment+$20,000 → +$60,000 attributed
3.0x
- Blended$70,000 → $310,000 attributed
4.43x
Illustrative — not a benchmark. Attributed revenue isn’t proof of incremental revenue; this shows the arithmetic, not causation.
The dashboard still shows about 4.4x. The owner’s question is whether the 3.0x increment works economically. And there’s no universal line: “stop at 3x” and “scale while above 4x” are rules of thumb, not business rules. The return you need depends on contribution margin, fulfillment costs, discounts, returns, acquisition management and creative costs, how many buyers are genuinely new, measured repeat purchase, overhead, cash and strategy. Why a strong-looking ROAS still doesn’t answer the profitability question covers that gap in detail.
If not ROAS, what sets the stopping point?
Contribution. Move the decision below ROAS, to what the added customers actually leave.
Marginal contribution after acquisition ≈ new-customer first-order revenue − incremental product and fulfillment costs − marginal acquisition cost
Contribution is the portion of revenue left after variable costs to cover fixed costs and profit. It isn’t net profit. The scaling question becomes: does the next increment leave enough to justify the capital, risk and operational load it brings? The ceiling is what a first order contributes before acquisition. How to calculate break-even and target CAC walks through that number, including when a first order is deliberately allowed to lose money.
First-order contribution before acquisition: $70 per new customer.
- Base$38 CAC blended all-in$32 left per customer
- Increment 1$44 CAC marginal$26
- Increment 2$51 CAC marginal$19
- Increment 3$58 CAC marginal$12
- Next decisionStill positive. Is $12 the best use of the next dollar?
If the next customers are still profitable, why not keep going?
Because positive contribution is a threshold, not proof that paid acquisition is the best use of the next dollar.
At increment 3, each added customer still leaves about $12. That clears the bar. But the same capital and operating capacity could go elsewhere, and the owner’s real question is whether $12 is the best available return on it. Competing uses might include:
- Inventory for products that sell reliably
- Production capacity
- Product development
- Storefront and conversion improvements
- Retention and email
- SEO and other durable demand
- Creative capability that lifts every campaign
- Debt reduction or a cash reserve
This isn’t financial advice, and there’s no universal answer. Sometimes $12 per customer is an excellent use of money. The point is to make it a decision rather than a default.
Why does CAC often get worse as spend increases?
Often, not always. The first customers are usually the easiest to reach; each expansion reaches people who need more persuasion.
Early spend tends to find obvious buyers, high-intent searchers and people already familiar with the category. Growth then means reaching less aware buyers, broader audiences and less obvious search demand, and those customers can rationally cost more. Auction dynamics, creative fatigue, geographic or product expansion, and the simple finiteness of demand at a given efficiency can all push the edge up. Google’s own planning tools let the conversion rate change as budgets rise, to account for diminishing returns as investment increases.
But it isn’t a law that every account follows smoothly. A larger budget can improve learning, unlock volume, stabilize delivery, support broader creative testing or capture seasonal demand. Meta describes a learning phase during which delivery is less stable and cost per result is usually worse; ad sets that can’t gather enough results to get through it can perform worse at low budgets, and large budget changes can restart it. Audience expansion isn’t automatically bad either. The question is always whether the added customers’ value still justifies their cost.
Can lower ROAS or higher CAC still be the right call?
Yes, when the lower efficiency fits an intentional business model. “ROAS dropped” shouldn’t automatically become “cut spend.”
- Filling unused capacitywhen added orders still clear their economics
- Entering a new market or launching a productwhere early customers cost more
- Buying customers with measured repeat valuewhere cohorts reliably pay back later
- Expanding distribution or market shareas a deliberate strategy
- Testing a channelthat needs enough scale to learn
- A planned growth periodwith the cash to support it
What separates these from overspending is intent and evidence. Repeat purchase has to be measured by cohort, not assumed from a theoretical lifetime value, and the payback has to arrive on a timeline the business can fund.
What besides acquisition economics can make the next increment a bad idea?
Even excellent marginal CAC doesn’t justify demand the business can’t fund, fulfill or measure.
- Capacity
- Production, materials, labor, warehouse, 3PL, customer service, lead times and quality all cap how much demand you can responsibly take. For manufacturers, whether unused factory capacity justifies more ad spend works through the same marginal logic from the operations side.
- Inventory
- If a core product is low, slow to replenish or held up by supplier lead times, scaling acquisition invites stockouts, backorders, a skewed product mix and wasted spend. The cost isn’t only the lost order: in one field study, stockouts reduced customers’ later purchasing as well. Dashboard economics don’t override physical inventory.
- Cash
- More advertising spends cash on media, inventory, materials, freight, production and fulfillment before it all comes back. A program can be economically positive over time and still squeeze near-term cash, and the gap is longer when payback depends on repeat purchases. Whether you can fund the next increment matters alongside whether it eventually earns a return.
- Measurement
- If Meta, Google, GA4 and store data tell materially different stories that nobody can explain, a large increase based on one platform’s ROAS is hard to justify. You don’t need perfect attribution, but larger decisions deserve more rigor. See which revenue number should judge a scale test.
Can the bottleneck move while I’m scaling?
Yes, and scaling is usually what moves it. At $50,000 a month demand may be the bottleneck; at $80,000 marketing may have relieved it, and the binding constraint is now production, inventory, conversion, cash or fulfillment. Scaling shouldn’t continue mechanically just because marketing used to be the constraint. How to tell whether marketing is still your growth bottleneck covers the full framework.
How do I know whether the next $10,000 is worth it?
Run it through the Next-Increment Test, then scale in controlled steps rather than by percentage.
- 01EconomicsWhat did the last meaningful increase in spend actually produce?
- 02CACWhat did the additional new customers cost?
- 03ContributionWhat did those customers leave after acquisition?
- 04CapacityCan the operation absorb another increment?
- 05CashCan we fund another increment until it pays back?
- 06MeasurementAre we confident enough in the numbers for a larger decision?
- 07Opportunity costIs more paid acquisition the best use of the next dollar?
- 08TestIf yes, increase again by one controlled increment.
- 09ReassessDon’t assume the next increment behaves like the last.
What does a controlled scale test look like?
- Establish a baseline.For a defined period, record spend, actual new customers, CAC, conversion, contribution and capacity.
- Increase spend by a meaningful but controlled increment.There’s no universal percentage; big enough to measure, small enough to reverse.
- Hold other major variables as steady as practical.If promotions, creative, pricing, products or channel mix change at the same time, you can’t tell what the budget did.
- Measure what changed.Use first-party new-customer counts and company sales, not only platform attribution.
- Estimate the economics of the added customers.Practical marginal CAC and contribution after acquisition.
- Check inventory, capacity and cash.Did the operation absorb it well?
- Decide.Scale again, hold, or fix something first.
Controlled comparisons make the numbers far more useful, but even they are noisy. Across 25 large retail advertising experiments, economists found the median confidence interval on return on investment was more than 100 percentage points wide. For the largest decisions, lift studies such as Google’s Conversion Lift, which compares people who see ads with a held-out group who don’t, can add stronger evidence. Match the rigor to the size of the decision.
Don’t scale by percentage alone. “ROAS is still good, so add 20%” isn’t a business rule, and neither is “we have 20% more capacity, so add 20% to ads.” Spend, customers and physical capacity don’t move one-for-one; measured increments tell you how they actually move.
When should I stop scaling, and when should I keep going?
Hold when something becomes binding. Keep scaling in controlled steps while the edge still clears your bar.
- Marginal CAC exceeds the economic ceiling
- Marginal contribution becomes unacceptable for your goals
- Recent increments keep deteriorating without a strategic reason
- Conversion weakens materially
- Inventory for key products is constrained
- Production or fulfillment reaches practical capacity
- Working capital becomes the limit
- Payback runs longer than your cash can carry
- Measurement is too weak for the size of the next decision
- Another investment offers a better expected return
- Marginal CAC remains acceptable
- Marginal contribution is attractive enough for your goals
- New-customer quality holds
- Conversion stays healthy
- Repeat economics hold up, where you rely on them
- Inventory and capacity can absorb the orders
- Cash can fund the next increment
- Measurement is adequate for the decision
- The added demand serves the company’s strategy
One weak week isn’t a reason to stop. Look for a pattern across increments, not a single noisy reading. For the numbers to watch each month, including CAC trend, contribution and capacity headroom, see what marketing numbers an owner should review every month.
So when should you stop?
Not when a dashboard crosses a magic number. Stop when the next increment no longer earns its place: when the added customers cost more than you’d willingly pay, leave too little behind, strain the operation or cash, or can’t be measured well enough for the size of the bet. Until then, keep scaling deliberately, one measured step at a time.
The average tells you how the program has performed. The edge tells you whether to buy more of it.
Before the next material increase, the Profitable Demand Diagnostic is a quick way to check the rest of the system, not just the ROAS.
Frequently asked questions
When should I stop scaling ads?
When the next increment of spend brings in customers at economics you no longer want to accept, or when inventory, capacity, cash or measurement makes more demand a poor use of capital. Blended ROAS alone can’t show that point.
What is marginal CAC?
The cost of the added customers from an increase in spend: extra spend divided by extra new customers. In practice it’s an estimate, best taken from a controlled budget change and first-party customer counts.
What is the difference between average CAC and marginal CAC?
Average CAC is total spend divided by total new customers across the program. Marginal CAC is what the latest added customers cost. The average can look fine while the margin is much more expensive.
Can ROAS still look good when I’m spending too much?
Yes. Blended ROAS averages efficient older spend with less efficient new spend, so it can stay attractive while the newest increment returns much less.
Should I stop scaling when ROAS drops?
Not automatically. Lower ROAS can be rational when the added customers still leave acceptable contribution, fill unused capacity, carry measured repeat value, or serve a deliberate growth strategy. Check what the latest increment actually produced.
Should I keep scaling if CAC is below break-even CAC?
Below break-even means the added customers are contribution-positive, which is necessary but not sufficient. You still need the contribution to be worth the capital, and inventory, capacity, cash and measurement to support another increment.
Why does CAC rise when I increase ad spend?
It often does, because the easiest customers are reached first and expansion reaches less responsive audiences. But it doesn’t always: more budget can sometimes improve learning and stabilize delivery. Measure each increment.
How does inventory affect ad scaling?
Demand for products you can’t ship turns into stockouts, backorders and wasted acquisition, and stockouts can reduce customers’ later purchases too. Scale acquisition in line with what you can stock and replenish.
Should I use MER or ROAS to decide when to stop scaling?
Neither alone. MER is a useful company-level trend and ROAS helps inside a platform, but both are revenue-to-spend ratios. The stopping decision rests on marginal CAC, contribution and the operating constraints.
Sources & further reading
- Principles of Microeconomics, Chapter 2 key conceptsOpenStax. Marginal analysis and sunk costs.
- About forecasts in Reach PlannerGoogle Ads Help.
- About the learning phaseMeta Business Help Center.
- About Conversion LiftGoogle Ads Help.
- The Unfavorable Economics of Measuring the Returns to AdvertisingLewis and Rao, Quarterly Journal of Economics, 2015.
- Contribution Margin: What It Is, How to Calculate It, and Why You Need ItHarvard Business Review, 2017.
- Measuring and Mitigating the Costs of StockoutsAnderson, Fitzsimons and Simester, Management Science, 2006.
Every CAC, ROAS, contribution figure and dollar amount in this article is illustrative, created to show the reasoning. None is a benchmark, a threshold or client data.
