Thursday, June 28, 2012

How to Write Crummy Requirements

Here's an object lesson in bad requirements writing.

"Good" is defined as a nice simple and intuitive GUI interface. I would be able to just pick symbol from a pallette and put it somewhere and the software would automatically adjust the spacing.
Some problems.
  1. Noise words.  Phrases like "'Good' is defined as" don't provide any meaning.  The word "just" and "automatically" are approximately useless.  Here is the two-step test for noise words.  1.  Remove the word and see if the meaning changed.  2.  Put the opposite and see if the meaning changed.  If you can't find a simple opposite, it's noise of some kind.  Often it's an empty tautology, but sometimes it's platitudinous buzzwords.
  2. Untestable requirements.  "nice, simple and intuitive" are unqualified and possible untestable.  If it's untestable, then everything meets the criteria (or nothing does.)  Elegant.  State of the art.  Again, apply the reverse test:  try "horrid, complex and counter-intuitive" and see if you can find that component.  No?  Then it's untestable and has no place.
  3. Silliness.  "GUI".  It's 2012.  What non-GUI interfaces are left?  Oh right.  The GNU/Linux command line.  Apply the reverse test: try "non-GUI" and see if you can even locate a product.  Can't find the opposite?  Don't waste time writing it down.
What's left?  
pick symbol from a palette ... the software would ... adjust the spacing.
That's it.  That's the requirement.  35 words that mean "Drag-n-Drop equation editing".

I have other issues with requirements this poorly done.  One of my standard complaints is that no one has actually talked to actual users about their actual use cases.  In this case, I happen to know that the user did provide input.

Which brings up another important suggestion.
  • Don't listen to the users.
By that I mean "Don't passively listen to the users and blindly write down all the words they use.  They're often uninformed."  It's important to understand what they're talking about.  The best way to do this is to actually do their job briefly.  It's also important to provide demos, samples, mock-ups, prototypes or concrete examples.  It's 2012.  These things are inexpensive nowadays. 

In the olden days we used to carefully write down all the users words because it would take months to locate a module, negotiate a contract, take delivery, install, customize, integrate, configure and debug.  With that kind of overhead, all we could do was write down the words and hope we had a mutual understanding of the use case.  [That's a big reason for Agile methods, BTW:  writing down all the user's words and hoping just doesn't work.]

In 2012, you should be able to download, install and play with candidate modules in less time than it takes to write down all the user's words.  Often much less time.  In some cases, you can install something that works before you can get the users to schedule a meeting.

And that leads to another important suggestion.
  • Don't fantasize.
Some "Drag-n-Drop" requirements are simple fantasies that ignore the underlying (and complex) semantic issues.  In this specific example, equations aren't random piles of mathematical symbols.  They're fairly complex and have an important semantic structure.  Dragging a ∑ or a √ from a palette will be unsatisfying because the symbol's semantics are essential to how it's placed in the final typeset equation.

I've worked recently with some folks that are starting to look at Hypervideo.  This is often unpleasantly difficult to write requirements around because it seems like simple graphic tools would be all that's required.  A lesson learned from Hypertext editors (even good ones like XXE) is that "WYSIWYG" doesn't apply to semantically rich markup.  There are nesting and association relationships that are no fun to attempt to show visually.  At some point, you just want to edit the XML and be done with it.

Math typesetting is has deep semantics.  LaTeX captures that semantic richness.  

It's often best to use something like LaTeXiT rather than waste time struggling with finding a Drag-n-Drop tool that has appropriate visual cues for the semantics.  The textual rules for LaTeX are simple and—most importantly—fit the mathematical meanings nicely.  It was invented by mathematicians for mathematicians.


  1. "invented by mathematicians for mathematicians"
    So what are non-mathematicians suppose to do?

    "WYSIWYG doesn't apply to semantically rich markup"
    Very well put!

  2. "So what are non-mathematicians suppose to do?" Either learn some math or stop trying to write equations.