Filters are stupid

Sometimes it bears remembering that many content filters aren’t score-based, but rather string-based. Because of this, you may have to more carefully craft your messages to deal with the fact that the filter cannot take context into account.

And so, today I wish to share with you three different strings actually seen to have caused rejections for bad content.

• “spic”: Often used as a racial pejorative, this string is used fairly frequently to block mail due to racial insensitivity. The problem? It’s also part of the name of a common household cleaner (Spic And Span).
• “cialis”: This word is often found in spam for body part enlargement as an alternative to Viagra. It seems to make sense to block on that string until you realize that just looking for those letters strung together in that order also manages to catch other words, like “speCIALISt.”
• ” cum”: This rather ingenious string was intended to catch sexually-oriented spam. It suffers from the same problem as or previous example, even though there was an attempt to limit the damage by including a leading space. It manages to catch unintended words like “cumbersome.” Maybe they should have included a trailing space, too, eh?

Feel free to add your own examples to the comments. The point I’m making here, though, isn’t just to make fun of filters for catching unintended words, but to make us think about things:

1. Are the filter strings I’m creating catching just the stuff I want them to catch, or are they catching other things that I wouldn’t mind actually receiving?
2. Is the copy I’m creating for my campaign likely to get caught by filters where a different choice of words would not?

Remember, at the end of the day, a string-based filter is stupid in that it can’t tell the difference between an off-color, stupid, insensitive (and likely not even funny) joke, and the announcement of a sale on cleaning products. So, is it better to block that string and have my janitorial staff miss a sale on products we actually use or not block that string and potentially face a lawsuit?