When do you consider a data lake for your analytics needs?

hey everyone, i've been thinking about data storage strategies for analytics lately. we're hitting a point where our current relational database is getting cumbersome for complex ad-hoc queries and holding raw, unstructured data. i've seen a lot about data lakes, but im curious about the practical tipping point. when do you folks actually decide its time to move from a more structured approach, like a data warehouse or just a heavily indexed DB, to a full-blown data lake? what were the signs in your projects?