We transitioned 3  Amazon ads campaigns from “Dynamic Bids – down only” to “Rule-based bidding”.  Rule-based bidding is a setting where you set a target ROAS and the Amazon algorithm will adjust bids to keep the ROAS at or above the target.

The piece of information that as of February 2023 Amazon did not disclose is that the algorithm takes 21 days to adjust and gather data.  During this time, the ROAS can go below the target.  In our case, it can do DRASTICALLY below the target.

One of our campaigns had a ROAS of 14.17.  We set a target ROAS of 10 with the idea that the algorithm would bid to increase sales while keeping the ROAS above 10.  By day 5 of our experiment, the ROAS on this campaign had plummeted to 3.5!  This low ROAS continued until 17 days in when we reverted the campaign to “Dynamic bids – down only” because we were losing so much money.  It wasn’t until later that we talked to the rule-based bidding product team that we were told that the first 21 days were a trial period when the algorithm basically plays around with bids to get the data it needs.  Presumably after 21 days the algorithm has the data it needs and the ROAS will generally settle at or above the target although Amazon doesn’t guarantee that that guardrail will be hit.

Budgets are also an issue for rule-based bidding experiments.  Amazon recommends that the campaigns chosen to be transitioned to rule-based bidding remain in budget for those first 21 days.  However, since the algorithm now sets bids, and the bids can be VERY high, it’s impossible to guess how much the campaign will spend per day.  In our case, we had a budget of $117/day before the switch and we raised it to $160/day to make sure it stayed in budget.  What we found was that the campaign absolutely blew through the budget!  CPC rose dramatically as did spend.  It felt like we were just giving Amazon to simply spend as much as possible regardless of the results and that’s exactly what happened.

The idea of rule-based bidding is excellent and has the potential to save a lot of time for Amazon ads customers.  However, for now the implementation is very sub-par, cost us thousands of dollars, and the documentation does not accurately describe the potential risks.