Reputation and Content Filtering: The Big Differences
By George Bilbrey
The August 6 issue of The New Yorker hit the newsstands with a piece about email spam: how it started, where it comes from today and what people from various businesses are doing to try and stop it. It was interesting to see this topic covered in such a lofty publication.
For the most part we thought the article was spot on. The writer discussed the matter at a fairly high level, but offered the lay reader a good sense of what the real issues are surrounding spam and filtering.
However, we thought the writer’s discussion of reputation metrics as a way to combat spam was a little off-base. He compared reputation to content filtering – essentially that spammers eventually learn to game the system.
We actually believe that reputation as a means of separating good mail from bad is fundamentally different from content filtering. With content filtering the differences between good email and bad email can be difficult to discern. An offer for free shipping from a retailer you frequent in an email message you elected to receive is welcomed. An offer for free shipping on body-part enhancing pills from a sketchy domain that you’ve never heard of is definitely not welcome. Machines find it difficult to tell the difference. Moreover, spammers learn the filter rules and find ways to thwart them – random words that lower the spam score, all image messages and more.
A well constructed reputation system is harder to game than content. The key in that sentence is “well constructed.” To our minds, a good reputation system that will really thwart spammers has the ability to:
- Pull from a lot of different receiving systems. It’s harder for spammers to fake good behavior across a number of systems. More importantly, at least some of those sources must remain unidentified.
- Look at a broad set of factors about the sender, including complaints, unknown users, spam traps, mailing patterns, and system configuration, to name a few.
- Take into account both short-term and long term behavior.
- Evolve – constantly and rapidly — to respond to the changes in spamming behavior
These are the sorts of things we are doing with our Sender Score reputation system and we believe that it has brought the chance of gaming down dramatically.
As our clients well know, building a good reputation in Sender Score is a process that involves pulling on a number of levers. The most well-respected senders have low complaint rates, low unknown user rates, few spam trap hits, a solid infrastructure and a stable sending identity.
Conversely, this is why a lot of marketers clung to the content idea for so long. Even though it felt unfair that the word “free” made your legitimate email look like spam, it also felt manageable. Just eliminate the word “free” and your email would sail through to the inbox. By contrast, managing reputation feels more difficult. However, it really isn’t that much more difficult and it has the added bonus of being more fair to legitimate senders and allows marketers to base content on response rates rather than arbitrary filtering rules.
About George Bilbrey
George Bilbrey is the founder of the industry’s first deliverability service provider, Assurance Systems, which merged with Return Path in 2003. He is a recognized expert on the subjects of email reputation and deliverability and is active in many industry organizations, including the Messaging Anti-Abuse Working Group (MAAWG) and the Online Trust Alliance (OTA). In his role as president of Return Path George is the driving force behind the ongoing innovation of our products and services. Prior to Return Path, George served as Director of Product Management at Worldprints.com and as a partner in the telecommunications group at Mercer Management Consulting. He holds a B.A. in economics from Duke University, and an MBA from the Kenan-Flagler School of Business, University of North Carolina.