Use SQS ApproximateNumberOfMessages to scale out and in with auto scaling group?

Hi all,

I did an official practice exam via Pearson and there is a question along the line of architecture of handling high volume of SQS messages (backlog). It says a solution architect is tasked with improving the performance of the application while keeping cost low.

I selected run the message consumer instances in an auto scaling group configured to scale out and in based upon the ApproximateNumberOfMessages cloudwatch metric.

But there is another answer with converting the queue to a FIFO queue (which will increase the cost) and increase the visibility timeout.

Is my selection the correct answer?

Thanks

Arthur