AWS CloudWatch works on the concept of metrics. A metric is a measurable value of an infrastructure component, sampled over a period of time. AWS provides predefined metrics, but you can choose to publish your own custom metrics.
For example, CPU utilization is a metric of an EC2 instance. The number of invocations is a metric of a Lambda function. These are predefined metrics provided by AWS. An example of a custom metric might be a stream of page loads done on an application's home page. These can be ingested by CloudWatch, as well.