Kurt McKee

lessons learned in production

Archive

Hey there! This article was written in 2007.

It might not have aged well for any number of reasons, so keep that in mind when reading (or clicking outgoing links!).

My Akismet analysis

Posted 1 May 2007 in accuracy, akismet, analysis, comment, and spam

Akismet is a service to help blog proprietors cope with blog comment spam. I installed it in late March and have kept meticulous track of its statistics for the month of April. I'm sure some people aren't familiar with some of the terminology of spam-fighting, so here are some quick definitions:

Ham

Messages from real people, i.e. not spam.

False positive

A message that is incorrectly flagged as spam. It set off the alarms when it shouldn't have. This is the regular guy with a replaced hip who sets off the metal detector.

False negative

A message that is incorrectly allowed through. This is the terrorist carrying explosives and ads for cialis and v1agr^a, but the metal detector didn't sound the alarm for whatever reason.

Here are the aggregate numbers:

Correctly marked as spam 252

Correctly marked as ham| 39

False positives (regular joes)| 7

False negatives (terrorists)| 1

Total number of comments| 299

Accuracy| 97%

I experienced a significant increase in the number of spam comments I received at the end of the month (and I'm currently in the middle of that spam storm). In the past few days I've received an average of 20 spam comments each day!

The problem with Akismet was well-highlighted by this increase: Akismet cannot distinguish between comments it's absolutely sure are spam, and comments it only thinks could be spam! That means that during the month of April I had to manually go through 259 comments to find the 7 that weren't spam.

I would never have considered this aspect (real comments getting lost in the mix of spam) as a significant issue were it not for Spam Karma 2, another way to fight spam comments. It has a slightly different interface that helps me ignore comments that are obviously spam. I'm going to test Spam Karma 2 exclusively for the month of May, so we'll see how the two compare.

All-in-all, I would give Akismet a thumbs-up. It's fairly accurate, assuming all things are equal. If, however, you give more weight to its mistakes, that 97% might drop.

☕ Like my work? I accept tips!