Computers & IT (Non-Engg)

There is a significant overlap between a data analyst & a data scientist but here's what I see as the main responsibilities of each:
Data scientist: Mainly looking at estimating the unknown, e.g.
Building statistical models that make decisions based on data. Each decision can be hard, e.g. block a page from rendering, or soft, e.g. assign a score for the maliciousness of a page, that is used by downward systems or humans.
Conducting causality experiments that attempt to attribute the root cause of an observed phenomenon. This can be done by designing A/B experiments or if A/B experiment is not possible apply epidemiological approach to the problem, e.g. @Rubin causal model
Identifying new products or features that come from unlocking the value of data; being a thought leader on the value of data. A good example of that is the product recommendations feature that Amazon first made available to a mass audience.
Data analyst: Mainly looking at the known, I.e. historical data, from new perspectives, e.g.
Writing custom queries to answer complex business questions.
Conceiving and implementing new metrics on capturing previously poorly understood parts of the business / product.
Addressing data quality issues, such as data gaps or biases in data acquisition.
Working with the rest of engineering to instrument incremental new data acquisition.
Of course, there is a significant overlap between the two roles. A data scientist always needs to write custom queries and a data analyst may need to build a decision-making module either by simple rules or applying machine learning principles.