Have been subscriber of Python Weekly for a long time and lately Pycoder's weekly. Both newsletters do a tremendous job. Have learned a lot from the links they share every week. Then why launch Yet Another Newsletter ?.
The problem with both these newsletters and pretty much every other newsletter is "It's a one size fits all" approach.
Assume a freshman learning Python and a Professionally with 5+ years of Python Development experience. They both have different needs. Links they click reflect their interest. Wouldn't it be great if a newsletter content could adapt to what they like and change accordingly ?. That's what ImportPython intends to do.
Import Python Newsletter launched a month back at Pycon India as part of the lightning talk. Approach was to learn till the first 10 issue how users behave and what links/articles they click/read. Based on the learning roll out a content recommendation engine. One that decides what content goes in which newsletter.
What are the parameters/document features that recommendation engine takes into account ?
- Click History. Terms that appear in the article that user reads.
- Length of the article.
- Preference to sections ? Articles ? Projects ? Videos ? Tweets ?.
- No of clicks per newsletter reflects consumption appetite.
How do we know if the recommendation engine is doing it's job ?
It's difficult to assess per user unless one can get feedback over a period of time. However click through rates will offer insight into how efficient the algorithm is ?. Also Average open rate being higher then other newsletters or industry standard is another benchmark to compare with.
Issue 10th is the last "One size fits all" version of importpython.com. Issue 11th to 20th will see rollout of the recommendation engine to a subset of subscribers. The subset of users are those who are subscribed from issue one. Every week the subset grows until 20th issue when all users as of today would have a recommended newsletter.