Have you heard of the Scunthorpe Problem? It highlights the side effects associated with trying to filter out profanity.

Where it affects me and the clients I work with most often is with personalization. They have to be careful -- creating an email with "Dear Firstname / Lastname" as the opening text can cause a problem if either of those fields contains something obscene or otherwise insulting. I've found that if it is paying customers who provided credit card information at the point of data collection, this isn't very likely to be a problem. Those customers aren't going to enter a swear word instead of their name. But if it's a free registration -- especially if it's in exchange for providing a free service or download -- those fields are sure to be filled with unbelievably profane garbage.

So it seems reasonable to try to filter it the profanity, right? You don't want to send out an email that says "Dear D*ckhead" -- that customer is unlikely to be happy about receiving that. Even if they entered that to begin with. So you add "d*ckhead" to your profanity filter. But what do you do when the next guy enters "D*kkhead"? How many variations can you include in the filter? It can seem never ending.

And then get ready for the false positives -- that's the Scunthorpe Problem.

So what do you do? Is there an easy answer? I don't think so. But if it was my call, I would:
  • Not directly filter names of people who provided credit card info. This segment is very unlikely to contain profanity.
  • Provide basic profanity filtering on free registrations, even if it can occasionally miss something, or cause a false positive rejection.
  • Don't personalize based on first name / last name, for subscriber list segments that are from free registrations.
  • I would possibly enforce a double opt-in on the free registrations. That way you'll know that if somebody types their name in as "D*ckhead," they themselves did it, and there's nobody else to blame.
Just about every one of these options carries risk, though. Doesn't it? If you block somebody unintentionally, then they get mad and complain about it on social media (providing new examples for that Scunthorpe Problem wikipedia page). Or if you send out an email with profanity in it, you're just as likely to get yelled at, and potentially have a PR issue on your hands, even if it's not your fault that the subscriber provided bad data.
Post a Comment