I was editing error messages for a terminal emulator program that had an error message that said: “Error 226: You should never get this message on a remote terminal.” So I asked the developer to explain what the error really meant. It seems there was a bug that caused error messages that should have displayed on the system operator’s terminal on random terminals connected to the system.

Then I asked him, “If you were working in your lab at 10 pm, and got this message, with the given explanation, would you do anything about it?” Of course, he said, “No”. Next, I asked, “To get rid of the bug, what do you need the user who gets the message to do?” “Call the sysop and explain what he was doing when the error occurred.” So we changed the message to read, “Error 226. If you get this error code, call the system operator and report what you were running or transmitting when it occurred. It’s the result of a bug and the sysop needs more information to track it down.”

[Reply]

Ben Reply:
March 8th, 2010 at 4:24 pm

That’s funny. The ironic thing is that instead of fix the bug, the developer coded an error message to show up when the bug is encountered. But maybe he did some homework and found that fixing the bug was complicated and costly, and the project manager made the decision to go the cheaper route at least at first. Thanks for the example.

[Reply]

Although all of these are good, I think that #1 is the most important.

Here’s another one: I’ve been taught to avoid language that sounds accusatory. For example, rather than “You entered the wrong text,” write “[program name] cannot recognize the text.”

I’d like to hear your views, Ben, as well as those of other readers, on whether this advice is still sound or whether a between-the-eyes approach might be more effective sometimes.

[Reply]

Ben Reply:
March 8th, 2010 at 4:34 pm

Avoiding accusation is still good advice. Problems aren’t always the user’s fault; there can be any number of reasons the user doesn’t do the exactly right thing. I think in the example you gave, we still don’t know exactly what’s wrong, so I would suggest telling the user what the system accepts. For example, “Only alphanumeric characters are accepted” or “Type only alphanumeric characters in the Favorite Cookie box.” A common one in this category is date formats. But I’d like to see any examples you have of when “between the eyes” is better.

[Reply]

Nice list! Here are a couple more:
* avoid ‘please’
* delete exclamation marks (they shout at the user and make them feel as though it’s their fault something went wrong or make them feel stupid)

I have written several blog posts on error messages, but those most relevant to this discussion are:
* http://cybertext.wordpress.com/2009/03/07/using-please-in-error-messages/ (or http://bit.ly/clXQ2A)
* http://cybertext.wordpress.com/2008/04/03/writing-useful-error-messages/ (or http://bit.ly/ajmN64)

[Reply]

Ben Reply:
March 8th, 2010 at 4:39 pm

I agree with both of these items, and I saw plenty of pleases and exclamation marks in the messages I went through. I left “please” in there a couple of times because for some reason it seemed appropriate. Maybe I was wrong. I seem to remember seeing your “please” post, which you posted only a couple of days after I posted on a similar topic, empathetic user assistance.

[Reply]