Every time we apply tROAS, it stops the algorithm’s learning process and prevents us from infusing more spend into the campaign. While the tROAS setting is applied, the algorithm "muscle" atrophies.
Often times, over-pacing to force the algorithm to learn and grow, and then pulling back on the spend to allow the ROAS to shoot up, is better than simply applying a tROAS goal - it’s like telling the algorithm to “only use the knowledge you already have to try to spend the budget and hit the goal.”
Doing this causes problems because the knowledge it has doesn't update with the ever-shifting market of new users, which causes the algorithm to not be able to spend the full budget, and ultimately causes us to fall behind on spend, and lose valuable time in the learning process.
We only apply it for short periods of time, and we have to sync it up during a portion of the month when we want to push the algorithm in that particular direction (high return), but it's not a long-term solution to achieve high performance.