I recently read a thought-provoking thread on the IIBA LinkedIN forum, relating how Business Analysts should state requirements. The conversation centred around one main topic: Is it advisable to define a system by stating what it shouldn’t do, or should these “negatively framed” requirements be avoided.
There’s no hard and fast answer on this, and there are certainly varying views on what constitutes a well-written requirement. However, I firmly believe that it is highly preferable to define systems in terms of what they can do (rather than what they can’t).
I’ll illustrate this with a real life—but non IT—example. I live in Portsmouth, on the south coast of the UK. Near the seafront, there’s a really steep hill. Given the number of pedestrians that are in the area, cyclists are banned from using the path downhill (as it’s likely they’d lose control and collide with an unsuspecting passer-by). A picture of the start of the hill is shown below. Trust me, it’s much steeper than it looks in the picture!
Think about this picture with your “BA hat on”. There’s a real-world requirement there (expressed as a rule) that says “No cycling, no skateboarding”. It’s painted in bright white paint on the path. That’s a clear requirement, right?
Wrong. In isolation this requirement is extremely ambiguous. It tells us two forms of transport that aren’t allowed – but it doesn’t really confirm which are allowed. In fact, on reading this requirement, some people may assume that it implies that:
- Scooters are allowed
- Rollerblades are allowed
- Aircraft are allowed to taxi down this path
- Tanks are allowed
- Zorbing is allowed
- Etc, etc
Now, in this instance most of us would apply common sense and say “Well, if skateboarding isn’t allowed, then rollerblading down there probably isn’t a good idea either”. However, in the world of analysis, that kind of ambiguity causes issues when a solution designer comes to architect a solution. What about a sledge in winter? Or a Segway? Are these allowed? Extrapolate this into the world of projects, and you can imagine the challenge and cost that imprecise requirements can cause.
My view is that it is always better – wherever possible – to define requirements in terms of the required behaviour, using positive language. Taking the example above, perhaps the real rule/requirement is “Pedestrians only”.
I also believe that relying purely on verbalised requirement statements is risky; it’s far better to have a range of requirement artefacts, including some models. A good quality logical data model (or a ‘class model’ if you’re using UML) will help a great deal. This can help implicitly show negative requirements (Theoretical example: if a data model shows that “one student can attend only one class” then it is implied, but also perfectly clear, that they can’t attend more than one).
What are your views? I’d love to hear them. Please add a comment below.