Automatically predicting that a proposed electronic message is flagged based on a predicted hard bounce rate

摘要

Certain embodiments involve stopping potential flagged email, such spam, before it becomes flagged email. For instance, certain embodiments can make a prediction (e.g., spam or other flagged email) about a “proposed” email message to a list of email addresses of proposed recipients. The flagged email prediction is generated based on the list of email addresses of recipients. The flagged email prediction can involve using stored data about historic email addresses, historic lists of email addresses, historic email transactions, or some combination thereof to calculate historic hard bounce rates. The historic hard bounce rates can be used to train processes to calculate the hard bounce rates for the email addresses on the list and the list of email addresses.