Every account I've taken over from an in-house team or a generalist agency has the same scaling story: performance was good, leadership asked for more volume, someone raised budgets 20-30% across every campaign, and blended ROAS dropped within two weeks. Nobody planned for that. It happened because budget increases were applied evenly instead of where the account could actually absorb them.
Scaling spend isn't a single decision. It's campaign-by-campaign math. Some campaigns have room to grow at the same efficiency. Others are already squeezing every profitable click out of their audience and any extra dollar buys you a worse conversion. Treating them the same is how a healthy account turns mediocre right when leadership is watching the numbers closest.
Find the marginal CPA, not the average CPA
Your average CPA tells you what a campaign has cost so far. It tells you nothing about what the next dollar will cost. That's the number that actually matters when you're deciding where to add budget. A campaign averaging $40 CPA might be converting its cheapest clicks at $25 and its most recent, budget-constrained clicks at $60 — the marginal cost is climbing even though the blended number still looks fine.
I pull this by looking at impression share lost to budget alongside a week-over-week CPA trend for campaigns that have been raised recently. If a campaign is losing significant impression share to budget and its CPA has stayed flat as spend grew, that's a campaign with real headroom. If a campaign has low budget-lost impression share and its CPA has been creeping up even without new budget, it's already saturated — you're bidding your way into weaker auctions, not being held back by a spend cap.
- High budget-lost impression share + stable CPA = room to scale profitably
- Low budget-lost impression share + rising CPA = saturated, more budget won't fix it
- High budget-lost impression share + rising CPA = audience is thinning out, scale cautiously in small increments
- Low budget-lost impression share + stable CPA = campaign is at its natural ceiling, leave it alone
Blended ROAS can look completely healthy while every individual campaign is quietly getting worse — that's the whole trap of scaling budgets evenly.
Sequence the increases, don't blanket them
Once I know which campaigns have headroom, I don't raise them all at once either. I increase the highest-confidence campaign first — usually the one with strong budget-lost impression share and a clean conversion tracking setup — by 15-20%, then hold for a full learning cycle before touching anything else. This is especially true for anything on Smart Bidding, since a sudden spend jump resets the algorithm's confidence in its own bid strategy and you get a volatility spike that looks like a performance drop but is really just recalibration.
This is the same reason I'm cautious about wholesale account restructures — I've written before about how restructures that reset your learning phase often do more damage than the problem they were meant to fix, and the same principle applies here. A budget change is a smaller shock than a restructure, but it's still a shock. Sequencing lets you isolate which increase caused which result. Raise everything at once and you lose that signal entirely — if ROAS drops, you won't know if it was the Shopping campaign, the branded search bump, or the PMax budget increase that did it.
That's the pattern I'm looking for before every increase: does marginal cost hold roughly steady, or does it spike? A campaign that behaves like Campaign A can keep absorbing budget in stages. A campaign that behaves like Campaign B needs a different lever entirely — better creative, tighter targeting, or new audience expansion — before more money will do anything but inflate CPA.
Protect the parts of the account that make scaling look worse than it is
Two things quietly sabotage scaling decisions before you even get to the budget math. First, broken or partial conversion tracking will make a genuinely saturated campaign look like it still has room, because you're optimizing toward a signal that's missing chunks of real conversions — I go through this in detail in my conversion tracking audit piece, and it's worth ruling out before you scale anything. Second, if Smart Bidding has been quietly optimizing toward soft leads or junk form fills, added budget will just buy you more of the wrong conversions faster, making the CPA look stable while your sales team's close rate collapses. Fix the input quality before you turn up the volume dial — otherwise you're scaling the exact problem you're trying to outgrow.
- ✓︎Check budget-lost impression share per campaign, not just account-wide
- ✓︎Compare marginal CPA trend, not blended average CPA
- ✓︎Confirm conversion tracking is complete and accurate before trusting the data
- ✓︎Raise the highest-confidence campaign first, alone, and hold for a full learning cycle
- ✓︎Only move to the next campaign once the first increase has stabilized
- How much should I raise a campaign's budget at once?
- I generally cap single increases at 15-20% for Smart Bidding campaigns to avoid a full bid strategy reset. Larger jumps trigger a new learning period and short-term volatility that can look like a performance drop.
- How long should I wait between budget increases?
- At least one full learning cycle, typically 7-14 days depending on conversion volume, before judging results and moving to the next campaign.
- Can Performance Max campaigns be scaled the same way?
- The logic is the same but the diagnosis is harder since PMax hides search term and placement-level data. I watch conversion value trends and asset group performance rather than impression share alone.