Scaling is the point where every weakness in an account becomes expensive at once. A structure that worked at €50 a day falls apart at €500 — not because Google changed, but because the flaws were always there and the budget was too small to expose them.
Before you scale anything
- Tracking verified against real orders. Scaling on double-counted conversions is how stores lose serious money quickly.
- Break-even known. You cannot judge whether a higher CPA is acceptable if you never worked out the ceiling. Calculate it.
- Brand separated, so you can see whether growth is real or just more of the demand you already had.
- Fulfilment that holds. Doubling orders with a supplier who cannot ship them is not growth, it is a refund queue.
The "limited by budget" signal
This status means your campaign could spend more than you are allowing. On a profitable campaign it is the clearest scaling signal Google gives you — free information that most accounts ignore for months.
On an unprofitable one it means the opposite: you are capping a loss rather than fixing it.
How much to raise at once
There is no published threshold for what resets learning, and anyone who quotes you an exact percentage is repeating a forum post. What is true is that large jumps move the campaign's data far enough that the bidding has to re-learn, and during that window performance is unreliable.
The practical approach most operators use: raise in modest steps, a few days apart, and change one thing at a time. Raising budget and tightening target ROAS in the same afternoon makes the result unreadable for a fortnight.
Change everything, see bad numbers during learning, panic, change everything again. The account is now permanently learning and never performing.
Where extra budget should actually go
Not evenly. In order:
- Winners with headroom — campaigns already above break-even and limited by budget.
- New search coverage — terms converting in the search terms report that you have no dedicated campaign for.
- New channels — Demand Gen or YouTube when Search has genuinely run out of room, not before.
- New markets — last, and only when shipping, returns and support actually work there.
When ROAS drops as you scale
It usually will, a little, and that is not automatically a failure. Cheap demand gets harvested first; the next euro is always slightly more expensive than the last. The question is whether you are still above break-even and whether total profit is rising.
A 5x ROAS on €3,000 is a worse business than a 3x on €20,000, assuming both clear your break-even. Scale against profit, not against a ratio.