Why big jumps break the bidding
Smart bidding builds its strategy around a budget level: which auctions to enter, how much to pay, which users are worth chasing. A sudden budget double invalidates that strategy and pushes the campaign back into learning, where performance gets erratic exactly when the stakes doubled.
There is a second mechanism: the best traffic is finite. At any moment there are only so many high-intent buyers searching your products. A budget the market can absorb buys those buyers; a budget beyond it forces the algorithm to fill the gap with cheaper, lower-intent clicks, because spending the money is the instruction it was given.
The controlled scaling process
Scaling that holds works in measured steps:
- Increase budgets 15 to 20 percent at a time, then hold one to two weeks and read the results before the next step.
- Scale what works: expand proven campaigns and products before launching speculative ones.
- Loosen ROAS targets gradually if volume is capped, because a too-tight target throttles spend regardless of budget.
- Watch impression share to see how much headroom your market has left.
Each step either confirms the account absorbed the spend or shows the limit, before the next increase raises the cost of being wrong.
Marginal ROAS, and when the account is not the problem
Average ROAS hides what scaling decisions need: the return on the last euro added, not the blend of all euros spent. Marginal ROAS declines as spend grows, because the cheapest conversions get bought first. That decline is normal; the question is whether the added revenue still clears your margin. Holding the average flat while scaling is the wrong goal, and chasing it leaves profitable growth on the table.
If ROAS collapses with every increase, the constraint is usually demand or the website, not the ad account. More spend cannot fix a market that is fully tapped or a site that leaks the traffic it gets.