-
tCPA vs tROAS experiment
I am looking to setup a campaign experiment testing tCPA vs tROAS.
I understand that tROAS preforms better when it is not limited by budget.
Any recommendations of the best way to configure this type of experiment.
-
3 Replies
-
tCPA runs into issues where it won’t or can’t spend when your target and budget are too close to one another. Otherwise, I don’t know how it would differ from any other bid strategy when it comes to limited by budget status.
Depending on whether CPA or ROAS is a more important metric for you, I’d build a campaign with the corresponding bid strategy. If you still want to test the other then set up an experiment.
If you don’t have much conversion data to feed then consider starting with maximize conversions or manual bidding until you have enough data to properly allow tCPA/ROAS to operate.
-
Imo tCPA is better if you have one conversion value. tROAS is for when you have multiple conversion actions worth different amounts.
-
The best way to set up the experiment is with Drafts and Experiments (testing bid strategies is exactly what it was designed for).
Both strategies perform better when not limited by budget – in fact both somewhat break when limited by budget, ie the logic behind them is designed on not being limited by budget.
In terms of targets, you should start both out with similar targets to what you’re getting currently (NOT what your eventual goal is) and then incrementally move the targets to what you want them to be (~10% every 2 weeks or so, depending on how much traffic you get). This is because the bidding strategy wants to use your current data as a starting point, so if you tell it to try to get a ROAS double what you’re currently achieving it has no idea how to do that.
Log in to reply.