For popularity-based item recommendation, you just recommend the most popular items (Items that people like the most, buy the most, etc), there is no fancy algorithm for that.
There is no simple answer to your question, as the definition of popular depends on your specific case and the insight you can extract from on your data, and metadata...
User filtering or product filtering or evaluation metrics will depend too on your specific case. If you provide more details, I'm happy to help you.