Why more spend always lowers ROAS (at first)
The expectation feels reasonable: if $1,000 returned a 5x ROAS, then $3,000 should return 5x too, just bigger. But that's not how ad auctions work. Your first dollars buy the cheapest, most obvious buyers, the people most likely to convert. As you spend more, you exhaust that easy demand and start reaching people who are less ready, less familiar with you, more expensive to convince. They still convert, just at a higher cost, so your average ROAS slides down as volume goes up. This isn't a failure. It's the predictable shape of scaling.
A few other forces pull in the same direction. If you raised the budget in one big jump, you may have knocked the campaign back into a learning phase, which causes a temporary dip while the platform re-optimizes. Seasonality moves the baseline: the same campaign performs differently in a quiet month than in a peak one. More advertisers competing for the same audience raises auction prices. And creative fatigue, the same ad shown too many times, quietly erodes performance no matter what you spend.
The number that actually matters
Here's the part that reframes the whole worry: ROAS is an efficiency ratio, not a profit total. A 5x ROAS on $1,000 returns $5,000. A 3.5x ROAS on $4,000 returns $14,000. The second one has a "worse" ROAS and makes far more money, and as long as that 3.5x is still comfortably above your break-even (the point where margin, shipping, and fees are covered), the drop is the cost of growth, not a problem to fix.
So when ROAS falls as you scale, don't reflexively cut spend. Check three things: did you jump the budget too fast (slow down the increments), is the campaign still above break-even (then it's fine), and is the lower ROAS buying you more total profit (then it's working). Cut only when you cross below break-even, because that's the line that actually matters.