James Serra explains why you might want to use a data lake:
To refresh, a data lake is a landing zone, usually in Hadoop, for disparate sources of data in their native format. Data is not structured or governed on its way into the data lake. This eliminates the upfront costs of data ingestion, especially transformation. Once data is in the lake, the data is available to everyone. You don’t need a priority understanding of how data is related when it is ingested, rather, it relies on the end-user to define those relationships as they consume it. Data governorship happens on the way out instead of on the way in. This makes a data lake very efficient in processing huge volumes of data. Another benefit is the data lake allows for data exploration and discovery, to find out if data is useful or to create a one-time report.
I’m still working on a “data swamp” metaphor, in which people toss their used mattresses and we expect to get something valuable if only we dredge a little more. Nevertheless, read James’s article; data lakes are going to move from novel to normal over the next few years.