A data engineering team is using Amazon Kinesis Data Streams to collect log data from various applications. They want to ensure that they can replay data for up to 7 days in case of an error or need for reprocessing. What feature should the team use to meet this requirement?