Recommendations in news products are terrible. | News Product Made with Humane Club

Recommendations in news products are terrible.

Published May 24, 2021
Updated Jun 14, 2021

News websites recommend articles based on topic, tags, author, section, and possibly past browsing behavior. But why this approach? And is it the right method?

The first mainstream websites to crack recommendations were e-commerce companies like Amazon. And then the rest, including news companies, seemed to have built on those principles.

But a base assumption was ignored — a user cannot buy 20% of a book or consume 2% of a teapot. The entire product needs to be purchased. The same is not valid for digital content.

For topics one is deeply interested in, one can easily read an 8000-word article. For topics one is not, brief quick updates are enough.

Print newspaper editors did this by giving all the news that fit by changing the length of articles based on his/her judgment of what is essential and what isn’t.

Given that, an ideal recommendation system for news websites would allow users to create a newspaper of their own that gives all the news in the day but with varying lengths, based on their interest.

For example, suppose I am not interested in politics but am interested in the economy. In that case, I should get political news but in 1-2 lines (as tweets) but economy-related news in detail.