Hi! Analytics has been a struggle to implement due to the E2EE meaning we have to download _all_ responses client-side. However, we're very happy with where the implementation is right now.
To start with, you get pretty graphs to summarise each question, with the exact graphic depending on the question type (e.g. a heatmap for the "address" input). This is a quick, simple summary and anyone can understand these. They're also easy to download and share.
Then there's correlation analysis, showing which pairs of questions have the most strongly related answers. We summarise this using plain English to ensure people without stats degrees can understand it! You can also view the regression graph for each question pair, which can help determine other patterns.
Despite the complexity, we've done some initial benchmarking (which we're hoping to formalise and publish soon) that shows these analyses perform very well at scale, hardly exhibiting lag even with tens of thousands of responses. Overall, it's a great groundwork that'll be nice to build upon in the future as we add more sophisticated analysis without compromising on the simplicity aspect.
Hope that helps, and thanks for your question :)