Genii Weblog

Directions on the box

Sun 22 Jun 2003, 11:37 PM

by Ben Langhinrichs
We like to play board games at my house, lots of different kinds of board games, and I have noticed that while some have directions on the box, many others do not.  This brings to mind a connection with software.  When the software is first installed, the directions are readily at hand, and can be followed easily, but all too often, software is being used by someone who doesn't have the directions with them.  With board games, this amounts to having lost the enclosed sheet of paper which detailed the instructions.  If the board game has the directions printed on the box, they are always there and easily accessible.  If not, you better hope that the person who used it before was organized.

With software, we developers often rail on people to RTFM, as if that is enough.  It may be a reasonable expectation  for developers (or programmers or consultants), but for users, it just isn't.  They may not have a manual, and asking them to search through a large, complex help file is a little like asking a second grader who wants to know the spelling of a word to "look it up in the dictionary".  Somebody older might guesstimate, but a second grader is likely to say, "I can't look it up because I don't know how to spell it".

So what is the answer.  Typically, context sensitive help is the closest to printing the directions on the box, but it is often very difficult to know the context well enough.  You almost need to know what the user did before they got wherever they are to give proper assistance.  I am not sure what the best answer to this, so I'll give two.  

Artificial Intelligence/Expert Systems - It seems that if a user follows a certain sequence of steps, then looks for help, a lot can be guessed about what they wanted to find.  It might be possible to build a system that followed the steps a user took, and "learned" what answers worked for them (I'm not sure whether they would have to click on a button to say "this answer helped after looking through several, or whether you should look at the last Help document they found or some other heuristic).  This is  pretty complex bit of programming, and requires storing a lot of information, but could be a useful approach for a product like Lotus Notes.

Über-Related Topics - Too often, related help topics are created by the developer, who is thinking of related code.  What I think might be useful is something more akin to's "People who bought this book also bought...".  What if the related topics for help documents were dynamic (or one set of them were), and reflected the idea that people who read this help also read these others?  Could that be used, like Baynesian filters in a way, to self determine in an organization what help topics might be useful for a person looking at this topic, even if the two do not appear connected to the developer?

Copyright © 2003 Genii Software Ltd.

What has been said:

19.1. Richard Schwartz
(07/21/2003 11:49 PM)

Reminds me of a line I used to use a lot. I've used it in training classes for senior admins. I've also used it when discussing the merits of centralized versus decentralized architectures with clients. The line is "Way too often, the administrator of the Domino server in the branch office is the person who now holds the job held by the person who inherited the job from the person who went to the two day training course. That's why, if we go with a distributed architecture, we're going to need to document standard operating procedures at a level that assumes zero prior knowledge". Best of all would be a combination like "The following documents that appeared in searches that also contained this document were most frequently read by the users who did the searches and who read this document.