Requirements gathering, the act of trying to understand a problem by talking to a selection of actual and potential users, is common place in nearly all good IT projects. Traditional waterfall projects require that a problem is fully understood, and documented, before beginning to build the solution. Agile projects stipulate that only a “broad brush” understanding of the problem is required to start work, with the gaps in knowledge being filled in as the project progresses. But generally speaking, any type of project, large or small, agile or waterfall, will have some form of requirements gathering component to it.

And that is a problem. You see, all requirements gathering activities are flawed. The end result of any requirements gathering phase is just not very good. Don’t believe me? Let me take you through a typical scenario.

A Typical Requirements Gathering Process

Our consultant is going to talk to a client about a new Intranet. The client already has an Intranet, but wants a new one. It works well in the main, but the technology is old and the client would like some new features. Over the course of several days our consultant quizzes the client about all areas of the system. He is a good consultant, so he does the following:

  • Structures the meeting into particular topic areas (review of old system, interview with users, potential new features etc.) and works through them with the appropriate people.
  • Takes notes from all his meetings.
  • Spends some time himself working through the system, forming his own impressions.
  • Writes up a report at the end of the requirements phase, with his recommendations.

Our consultant reports back to his project team that he understands the problem, works with a developer to devise a solution, and sits back with a rosy feeling as the solution is built. What he has forgotten is his requirements gathering is flawed, and his results are not very good (do you see a recurring theme yet?). However, it is not his fault; he did a good job, or at worst, an average one. You see, it is just the nature of the beast. Here is why:

The Process is Flawed

Our consultant sat in lots of meetings, being told what was wrong with the old system. Different stakeholders showed him various problems. Some discussed what they would like to see instead. Some even drew pictures. Our consultant listened, processed and made notes. Let’s breakdown these activities:

  • Client thinks about a requirement
  • Client communicates a requirement
  • Consultant hears requirement
  • Consultant documents requirement

But let us look again:

  • The client has a particular angle on a requirement based on their own experiences with the system. These experiences might be representative, or they might not. The requirement might be valid, or it might not.
  • The client communicates the requirement as best they can to our consultant. They might be good at this, or they might need some coaching from the consultant. Let’s say they get 80% of their version of the requirement over to the consultant.
  • Our consultant listens, understands the problem as best they can, and naturally puts their own interpretation on what they have heard. We are now quite possibly down to 70% of the original (possibly itself biased) requirement.
  • Our 70% requirement is documented. I think most of us will agree note taking is an art in itself, so our requirement is watered down further to 60% of its former self. Maybe our notes aren’t clear, don’t make sense after the meeting, or are just wrong.

So now when our consultant sits down with the development team to discuss the problem, we understand that in fact they are discussing a 60% watered down version of the original problem. Depending on who they spoke to originally, this problem might not be all that accurate either. This is the inherent danger in requirements gathering. This is not to say requirements gathering is a pointless exercise, but it must be approached carefully.

8 Tips to Capture Better Requirements

So how do we mitigate against these factors, if they are oh so inevitable in requirements gathering? Good question. Try the following tips. They won’t magically solve the problem, but they will all help you keep the percentage of the original requirement you record as high as possible:

  1. Firstly, accept you will never hit 100% accuracy. Bear it in mind at all times and it will drive you to capture better requirements.
  2. Spend as long as you possibly can on this phase and talk to as many people as possible. Talk to people in a wide range of roles, but ask them all the same questions. You will be surprised how good an insight staff “on the shop floor” have about so called “management level” questions (and vice versa).
  3. Ask “Why?” Ask it a lot. Never accept an answer in the requirements gathering phase without asking “why” at least five times. No matter the project, the technology, the company or the interviewee -- “Why?” is the best question.
  4. Requirements gathering has to be conducted by as few people as possible. Imagine a different consultant carrying out each meeting in the scenario described above. Our percentage would plummet. If more than one person is really needed (and avoid it if you can), then the process has to be headed up and coordinated by a single person. Some single person must hold the whole problem in their head.
  5. Always have a dedicated scribe. In meetings and workshops, the person conducting the session should never write the notes. If they are doing a good job (see the first point), then they just won’t have time.
  6. Better still, use “dialogue mapping” techniques to document meetings. There is not enough room to go into this topic here, but this involves a dedicated scribe documenting decisions and questions using diagrams on a big screen that everyone can see.
  7. Deliver outputs of requirements gathering to the client as often as possible. Always send them a copy of your raw meeting notes. Then follow up meetings with your findings. If you need to produce a formal report, go over your thinking and draft copies with the client as early as possible.
  8. Your final report will not be perfect, so when development starts, keep talking to the client. If possible, keep showing them the solution as you build it to check your theories and ideas.

Editor's Note: You may also be interested in reading: