Data lakes exist because not all valuable data fits neatly into tables when it arrives.
They store raw files of any kind in cheap storage, often cloud storage built for huge numbers of files.
Nothing has to be designed before data is saved, which makes collection fast.
Files are usually organised in folders by source and date, such as one folder per day.
Open formats such as Parquet, JSON and CSV keep the files readable by many tools.
A catalogue records what each dataset is, where it lives and what its columns mean.
Without a catalogue, a lake quickly becomes a data swamp, full of files nobody can find or trust.
Lakes are common for logs, sensor readings, images and other data with no fixed use yet.
Data scientists use lakes for exploration and for work on large raw datasets.
Raw zones keep data exactly as received, while cleaned zones hold processed versions.
Storage is cheap, but scanning huge raw files is slow and can be costly.
Access control matters, since raw data often contains personal details that were never removed.
Deleting one person's data on request is harder in a lake, because it can be spread across many files.
In practice, lakes and warehouses are often used together, with the lake feeding the warehouse.
The usual mistake is treating the lake as a dumping ground with no catalogue, owner or rules.