How to scale Facebook and Google ads without killing ROAS
“I can't scale my ads profitably.” Push spend up: CAC climbs, ROAS slides, margin thins. Pull back: growth stalls. Nobody in the building can say, on any given morning, whether yesterday's ads made money — so the brand swings between the two failure modes on vibes.
THE CORE PROBLEM, AS DATADREW'S PROBLEM BANK STATES IT · JUL 2026For Shopify founder-operators, growth leads and paid-media operators at $500K–$30M+ GMV hitting the first real scaling push: it works at $500 a day, breaks at $1,500, and you retreat. Eight questions decide it; two numbers and two tools answer them. Sean Frank, CEO of Ridge: “You need to be able to survive, I think, a 1.5X ROAS.” That floor applies to the next dollar, and the average is the last number to tell you it has been crossed.
NO PERCENTAGE RULES Spend, Shopify revenue and your own margin. No benchmarks. Checks are daily, not real-time.
Eight scaling symptoms, and where each one is answered
One question hides eight. Pick yours.
Usually a ceiling: the raise bought people the concept doesn't work for. The nine ceilings name which.
EIGHT QUESTIONS THAT FIND YOUR CEILING ComingThree questions before any edit, and six intervention states, from Act now to Do not act.
SIGNAL OR NOISE →Not a percentage. The readiness check names the unmet factors and sizes the step.
THERE IS NO 20% RULE ComingTo the highest marginal contribution you can trust, never the higher platform ROAS.
CUT OR FEED →“every time i push past $1k/day my cpa doubles. back off and it heals. stuck.” (a composite line from our operator research). Back off and it heals: saturation. Back off and it doesn't: something broke, and the diagnostic order finds it.
SCALING OR SATURATING? ComingMarginal ROAS is the number that describes the next dollar. Two spend levels, matched windows, Shopify revenue at each.
MARGINAL ROAS → Which ROAS number answers this?Blended MER: is the business still efficient. Marginal ROAS: should I raise. Platform ROAS: direction inside one channel.
WHAT BLENDED ROAS CAN'T TELL YOU →Ten principles, previewed below.
TEN PRINCIPLES, HUMAN OR AI ComingMarginal ROAS decides the raise. Blended MER checks the business survived it.
Average ROAS blends every dollar, including the cheapest demand you were ever going to buy: brand search, cart abandoners, your warmest audience. Those hold the average up long after the increments stopped paying. Marginal ROAS is extra revenue divided by extra spend between two budget levels: the only ROAS that describes the decision in front of you.
Blended MER is total Shopify revenue over total ad spend across Meta and Google: after the raise, is the whole business still efficient? It is an efficiency ratio that credits every platform with all revenue, so it cannot allocate, and it is not attribution. Which number settles which argument: Marketing ROAS decoded, what blended ROAS can't tell you, how to pull it across Meta, Google and Shopify.
Curtis Howland, DTC growth operator: “So they make budget decisions based on bad data. They over-invest in whatever platform tells the prettiest story.” Platform ROAS reports one channel on its own attribution window; Shopify revenue, net of refunds, is what the business banked. Use the second for both numbers.
Illustrative numbers: a $6M GMV Shopify brand raises Meta prospecting from $2,000 to $2,600 a day
| Daily spend | Daily Shopify revenue (matched 14-day windows, Google held flat) | Average ROAS | Marginal ROAS on the step |
|---|---|---|---|
| $2,000 | $6,800 | 3.40x | baseline |
| $2,600 | $7,700 | 2.96x | ($7,700 − $6,800) ÷ $600 = 1.50x |
The founder shares a 45% contribution margin, so break-even is 2.22x. The average still reads 2.96x; the step earned 1.50x, Sean Frank's line exactly and 0.72x under the floor. The dashboard never blinked. The calculator reads it as Constrain.
Marginal ROAS calculator
Two spend levels, Shopify revenue at each, matched windows.
Spend and revenue only; bring the margins you share with Drew to the decision. Reads use four of the eight budget states (Invest aggressively, Scale cautiously, Constrain, Reduce) plus a cut read and a pause read; the full eight-state table, including Maintain, Harvest, Explore and Pause, is on the canonical calculator at Marginal ROAS: The Number That Decides Your Next Ad Dollar.
Before you touch the budget slider: the readiness check
The question is never “raise by what percent”. It is “what does the evidence support”: economics against the real target, conversion volume, time since the last edit, creative depth, the marginal trend, audience breadth, site conversion and inventory, conversion lag, the demand calendar, downside tolerance.
The “20% every few days” rule exists because a big single step can restart Meta's learning (operators put the reset threshold at roughly a 20–25% jump; Datadrew operator research, July 2026). That argues for sizing steps to evidence; it is not a sizing rule. 300 conversions a week, fresh creative and 60 days of stock can take a bold step. 30 conversions and 11 days of stock should take none.
The $6M brand above: $2,600 a day, last edit four days ago, 190 conversions, 11 days of cover, tracking healthy, marginal below break-even. Verdict: Not yet, don't step, three unmet factors named.
Budget-increase readiness check
Six inputs, one campaign. Output: Ready or Not yet, the unmet factors, a step size. Never a percentage.
It grades your inputs against what a buyer checks; no account access, no benchmarks. The conversion floor (roughly 50 a week) is the figure operators cite for Meta's learning phase, as recorded in Datadrew's operator research, July 2026; Meta's own published threshold was not re-verified on 10 September 2026, so treat it as reported. The budget-increase piece, when it publishes, will be its canonical home.
Run this on your own account, on real orders and stock: connect Shopify, Meta and Google, share your margins, and ask Drew what to scale, reduce, pause or leave alone this week. Free plan, no card.
Connect your store freeBudget is one of six ways to scale, and the only one most brands ever try
Vertical scale is the slider. The other five are where an account goes when the slider stops paying.
| Form | When it is the move |
|---|---|
| Vertical More budget through the existing system | Only when the marginal read holds, creative and demand remain, stock and site can take the volume, and no recent edit muddies the read. |
| Creative-led New concepts, so the system has more good options | When spend sits in two or three aging winners, or every raise degrades efficiency fast. Ash Melwani, CMO of Obvi: “Assume that you will have a 10% hit rate on all the new ads you test.” |
| Product-led More eligible products, better product sets | Room for products whose margin, stock and repeat behavior justify it, taken from ones that only look good on revenue. Any product-level ROAS is approximate: spend is not attributed per product. On Google, feed quality decides delivery before bidding does. |
| Offer-led Stronger offers raise conversion capacity | Recalculate contribution after the discount. A raise that only works at 25% off is a different business with a different break-even. |
| Geographic New markets, after they have been costed | Cost shipping, currency, tax, returns, localization, conversion rate and availability first. Cheap CPMs in a market you cannot deliver to profitably are a cheaper way to lose money. |
| Value-led Fix what the platform is optimizing toward | Improve the value signal so delivery favors higher-value orders. On Google, reported revenue is often the wrong value: margin, returns and whether the buyer is new or returning change it. |
On Google the target is its own lever: budget, target (tROAS or tCPA), value signal and coverage each bind separately, so find the binding one first. When Performance Max spend grows, check where it came from: brand or non-brand, new or existing customers.
The nine scaling ceilings, how each shows up, and the move that is not “more budget”
When more budget degrades performance, one of these is binding. Name it, then act on it.
| Ceiling | How it shows up | The move |
|---|---|---|
| Creative concentration | Most spend on two or three aging concepts; CTR sliding, frequency climbing | Creative-led scale; test budget separate from scale budget |
| Addressable demand | CPM rising with no targeting change; back off and it heals | New audience, market or product; harvest what is there |
| Offer strength | Clicks hold, conversion falls on cold traffic | Offer-led scale; recalculate margin after the discount |
| Website conversion rate | CVR down across every campaign at once; a checkout or theme change in the log | Fix the site. No media move fixes a site problem |
| Inventory | Hero SKU under two weeks of cover; ads still pointing at sold-out products | Constrain, or move spend to products with stock |
| Margin target | Marginal ROAS under 1 ÷ margin while the average looks fine | Reduce to the last step that paid; run the contribution-margin check |
| Market size | Non-brand impression share already high; brand search fully harvested | Geographic or product-led scale; accept the harvest state |
| Bid or cost control | Spend far under budget; the cost cap or tROAS refuses the auctions on offer | Relax the control deliberately, within your economics |
| Measurement quality | Conversions missing or duplicated; the raise cannot be read | Protect measurement first. Never scale on data you cannot trust |
The state your account is in decides the move before the numbers do
Four states make up the scaling arc, and the same marginal reading means something different in each.
-
Validated
It has met its evidence threshold and your economics. Does headroom exist?
Find out before spending into it: one measured step, read after a full cycle. Most accounts go straight from “it works” to “triple it”.
-
Scaling
Budget, coverage or creative supply is expanding. Is the marginal dollar still paying?
Measure marginal efficiency without destabilizing the account: one causal edit at a time, read after a full cycle.
-
Saturating
Marginal economics deteriorate as spend expands. Back off and it heals.
You have found the curve for this setup. Constrain the budget at the last level that paid and open another path: creative, product, offer, market or value signal.
-
Fatiguing
Creative, offer or audience response is decaying. At which level?
Ad, concept, audience or offer. Replacing an ad when the concept died buys a week. Name the level first: a daily read of CTR and frequency against the account's own baseline tells you which one is decaying.
Three more matter: destabilized (a big edit or tracking break broke the baseline; restore trustworthy observation first), promotion mode and recovery (do not read lagged outcomes as new deterioration). Reading your state gets its own piece (coming). Whether to act at all is the signal-or-noise piece: Act now, Act cautiously, Run a controlled experiment, Investigate first, Hold and observe, Do not act.
Ten principles for anyone, or anything, touching the account
For a founder, a media buyer, an agency or an AI agent: whoever holds the account can be held to these.
- Act at the cause. A landing-page problem does not get a targeting change.
- Protect the measurement system. Never fix a tracking break with a media move.
- Respect learning and conversion lag. Recent edits make short windows lie.
- Separate test budget from scale budget. One buys learning, the other volume.
- A profitable campaign may have no headroom. Average efficiency is not marginal capacity.
- Lack of spend is not lack of potential. Algorithms under-deliver a challenger before it has fair evidence.
- Don't force equal spend to make a test look clean. It has an opportunity cost.
- Stacked edits teach you nothing, so make one causal edit at a time unless you are containing an emergency.
- Preserve reversibility. Prefer the move with a written rollback condition.
- Waiting needs a reason, a window and a trigger. No action is a first-class decision.
Benjamin Wenner, on the job now that the platforms bid for you: “They do not manage bids. They manage the system that manages bids.” The ten principles are that system, written down.SEARCH ENGINE LAND · VERIFIED IN DATADREW'S OPERATOR RESEARCH, JUL 2026
What Drew does with this, and what it doesn't do yet
Ask Drew what to scale, reduce, pause or leave alone this week — its Budget Recommendations come with the reasoning, graded against the margins you share with it, your stock and repeat behavior, not platform ROAS alone.
In practice: once Shopify, Meta and Google are connected and Drew has the margins you want decisions graded against, ask in plain English: “Which campaign is at its ceiling, and which ceiling is it?” Drew reads yesterday's orders and both ad accounts every day and answers with the numbers and the argument. The frameworks here are published method; the raise is still your call.
Budget Recommendations
What to scale, reduce, pause or investigate, how much and why, with the reasoning.
Diagnose Performance
Why CPA, ROAS, CVR or AOV moved, across Shopify, Meta and Google in one pass, plus daily creative-fatigue detection against your own baseline.
Daily Ads Brief
A diagnosis-grade daily brief runs on pilot accounts. Scheduled Slack and email reports and alerts are live for everyone.
Execute Ads Changes
Built to make and execute approved changes with guardrails, rolling out now. Nothing here assumes it acts on your account by itself.
Start with the free store health check, read how Drew handles acquisition, or see pricing.
Scaling questions operators actually ask
Find the ceiling before the budget does
You have the two numbers, the readiness check and the nine ceilings. Ask Drew to run the same read on your real orders, stock and both ad accounts, refreshed daily, and it tells you which ceiling you are at, and why. Free plan, no credit card.
Book a demo