Why metadata is key in data lakes

Directing your data pipelines to a data lake is great - but how can you keep up with the dataset context and their interdependencies?