Why a High Target ROAS Reduces Google Ads Conversion Volume

Question 1

Why did conversions fall after target ROAS was raised from 450% to 650%?

Accepted Answer

According to the article, after top campaigns were raised from 450% to 650% the algorithm pulled back and stopped bidding into auctions where it could not confidently predict meeting the 650% threshold. Several high-spend campaigns showed "Learning (limited)" while others showed "Eligible" but were barely spending, and fewer impressions led to fewer clicks, fewer conversions, and even less learning data. The article describes this as a reinforcing downward loop.

Question 2

How does a too-high target ROAS starve smart bidding of conversions?

Accepted Answer

The article states that Google's smart bidding including target ROAS needs a consistent flow of conversion data, and when the target cannot be reliably hit it restricts bids, avoids auctions it deems risky, and narrows the audience it will show ads to. The result is campaigns go quiet, not because demand dried up, but because the system stopped competing for it. The article notes the algorithm is doing what it was told, but that instruction is incompatible with generating enough volume to sustain itself.

Question 3

How much conversion volume does target ROAS need per campaign?

Accepted Answer

According to groas, Google generally needs at least 15 conversions per campaign over a 30-day period for target ROAS to function well, and more is better. Several campaigns were generating 8 to 12 conversions per month even before the increase, and the article states a campaign with 10 monthly conversions cannot meaningfully optimize toward a 650% ROAS target. With that little signal, the model cannot reliably distinguish between a $90 conversion and a $180 conversion.

Question 4

What does Learning (limited) mean when the ROAS target is unreachable?

Accepted Answer

The article states that "Learning (limited)" means the bid strategy does not have enough data to optimize. When the cause is a structurally unreachable target, waiting does not resolve it and each day in that state loses auction momentum and generates even less data. The article notes the learning phase is only temporary when the conditions for learning are actually met.

Question 5

How fast did conversion volume recover after removing the ROAS target?

Accepted Answer

The article reports that within the first two weeks of removing the tROAS constraint, conversion volume recovered to roughly 80% of its pre-stall level. By week four, it exceeded the previous baseline. The campaigns that had been in "Learning (limited)" for weeks exited that state within days of the bid strategy switch.

Question 6

How should target ROAS be reintroduced after volume is restored?

Accepted Answer

The article recommends waiting until conversion volume returned to pre-stall levels after roughly three weeks on Maximize Conversions, then setting initial targets at 380%, just below the actual observed ROAS of 4x. From there it recommends raising the target incrementally in steps of 20 to 30 percentage points, only when the previous target had been consistently met for at least two weeks. Over the following six weeks this approach allowed efficiency to rebuild without repeating the stall.