Thursday, July 26, 2012

documented Requirements as an essential artefact for systems development

Let’s start with a clear statement:  I view documented Requirements as an essential artefact for systems development.

Given that, I don't want to imply that Requirements are 'sacred' in some way or worse yet, an end in them-selves. They are a vital but intermediate deliverable along the way to delivering the solution that your business needs to succeed.

This brings me to the often-cited problem in Requirements work most commonly known as 'analysis paralysis'. Requirements gathering, analysis, and documentation cannot be some endless task that defers delivering the Requirements because of uncertainties or changes in the details of the Requirements. It does require sufficient time, but historically I have seen the effort spent on Requirements average around 25% of a total systems project. If you are spending more than 25% on a regular basis, you might have some paralysis going on.

(Sidebar: the other 75%? Typically I see 50% on design, build, and unit testing, and 25% on the rest of Testing: System, User, Acceptance, Volume, Regression, and probably more I am forgetting.)

How can you avoid paralysis?
- Define a clear scope of the Business process or function that is to be supported by the solution. If you have or can quickly put together a process map (start by defining all the external events the business must respond to), then it can be shown what you are gathering Requirements for, and what you are not. Having that Process Map also helps identify units of work in process steps that can be documented as Use Cases.
- If nailing scope down is difficult, then time-box the work; take what you do have defined as likely in scope, and spend 2 to 3 weeks of focussed work with business involvement, and you will be surprised the volume of Requirements you will produce. Stop at the end of the time-box, and re-group/review with your Sponsor. You may have enough to start a round of development/delivery, or a better view of how much time you need to finish a complete set of Requirements.

It is also important to recognize the Requirements will change, but not necessarily for the reasons you might think. For example, you may document a portion of the requirements and review it with the business staff involved in the gathering, and you just may have written it down wrong or misunderstood what they told you; reviews/walkthroughs are essential for any quality product. If you can succeed in getting agreement that the documented Requirements are correct from the business, the amount of change after that point will be minimal.

Agile methods proponents will tell you the business changes too much and too often for documented requirements to remain valid; don't believe it. Core information systems requirements about collecting, storing and using data do not change much over time; what does change the most is process/procedural activities, the main reason why an Information System should not automate business process, especially now that BPM tools are available that are designed to accommodate procedure changes. As author Steve McConnell once said, the main reasons requirements change during agile development is that the requirements were not fully investigated before development started.

The last advice I would offer to fellow BA's is that you do document the Requirements as your work, but you do not own them; they belong to the business. So, don't agonize over what may happen to Requirements after you have delivered them. There is a quote about working whose source I must track down, as it guides me every day; it is:…

"Show up on time, do your best; don't get attached to the results".

Wise words. Besides, there is always another project up next that needs its Requirements documented, so get to it!… 

No comments:

About Me

Ontario, Canada
I have been an IT Business Analyst for 25 years, so I must have learned something. Also been on a lot of projects, which I have distilled into the book "Cascade": follow the link to the right to see more.