StrataScratch, built by Nathan Rosidi, sits in a specific and useful corner of the learning market. It is not a course and it does not pretend to be one, it is a large and well organised bank of real data science interview questions that you work through in the browser. The value comes from the sourcing. These are questions that have actually been asked at companies people want to work for, so instead of grinding through invented puzzles you are rehearsing the exact style of problem you will face, which does a lot to calm the nerves on the day.
The SQL and Python coding sections are the strongest part, and the inclusion of open ended analytical and product sense questions is genuinely useful, because those are the questions that trip up candidates who have only ever practised writing code. Being able to filter by a specific company is the feature I found myself using most, since prepping for one employer is far more efficient than practising blind. The honest limitations are worth stating plainly. This platform will not teach you a subject you do not already understand, the explanations exist to justify an answer rather than to build knowledge from the ground up, so a genuine beginner will feel lost and should start with something instructional first.
The editor and general interface are functional but a little tired, and because many solutions are community contributed you sometimes get an answer that works but is not the model of good practice. Machine learning and deep learning interview content is present but noticeably lighter than the analytics and SQL material, so heavy ML roles will need to supplement elsewhere. My overall view is that StrataScratch earns its place as a rehearsal tool. If you already have the theory and you simply need to walk into the interview having seen the shape of the questions many times over, the paid tier is inexpensive and pays for itself the moment it helps you feel composed.
Treat it as a gym, not a school.