Hacker News new | past | comments | ask | show | jobs | submit login

Yep, Gmail's spam filters work based on the collective judgment of Gmail users. The core of your and linus' problem is a lot of people use the 'mark as spam' button as unsubscribe button for mailing lists.



based on the collective judgment of Gmail users

I find this trend of "follow the majority" quite disturbing - it's as if they're implicitly saying that everyone should think the same way and punishing those who don't follow. What's spam to me may not be spam to you, and vice-versa.

Then again, having a personalised spam filter for each user would probably consume a huge amount of resources...


Not sure why you are down-voted. Perhaps because everyone (that run their own mail) generally runs individual filtering per account. Typically spam assassin will score an email, but filtering (based on that, and other criteria) is up to the individual user (eg: by having a white-list, choosing spam score to treat as spam etc).

As mentioned further up, some scoring works well for many users, but not for all, such as marking eg: Russian/Chinese/Not-spoken-here-by-most language as spam.

I really see no reason for why Google should be so bad at classifying email as they apparently are.


That idea sounds compelling at first, but the data doesn't support it. There are plenty of email marketers who are focused on a non-technical audience (who presumably use 'mark as spam' to unsubscribe frequently) and which have no problems with spam folder placement.

There's a spectrum, and if a given sender looks considerably worse than average, they're more likely to get filtered.

If anything, if a newsletter is getting filtered, it's more likely to be the marketing manager's fault - perhaps they don't adequately monitor deliverability, or they don't test their content, or they don't use activity segmentation... etc.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: