This is an open-ended question, which is why the following answer is an opinion.
The focus of QA (quality assurance) in the DevOps field is to monitor the deployment metrics to ensure that faults are caught before they can manifest. The most direct metrics of application health include application availability and errors. These can be directly extracted from error logs and automated monitoring of served web pages (that status code 200 or the login form are always returned).
The next level allows you to anticipate errors, bottlenecks and resource usage. This is very deployment specific. In some cases you may have a micro service that receives a high load or in others a single page may make several database queries. Monitoring your servers' CPU load and memory usage can also be used as metrics. These may act as key indicators to your applications' health.
Other aspects delve into feature and device usage which enable you to make informed decisions about depreciating these in order to limit your application's complexity. Since complexity is directly linked to bugs and development velocity, it makes sense to limit it where possible.