Tuesday, February 12, 2008

How to deliver real value using Iterative and Incremental Development

The benefits of Agile development approaches in which developing software applications iteratively and incrementally is an important and inevitable principal are questionable if deliverables produced during the project do not introduce any business value.
If a development project doesn’t deliver business values incrementally iteration by iteration then it’s executed in a value-neutral setting in which

  • requirements are treated as equally important
  • code delivery is as important as the delivery of an usable system
  • implementing largest number of requirements possible become more important than delivering desired outcome of the project
In another word, the development team iteratively implements the requirements rather than iteratively solve problems. Although project members feel that they are iterating successfully, they never make the transition from iteratively addressing the technical risks to iteratively delivering usable versions of the system that provide realizable business benefit.
This will raise the question that how we exactly have to develop requirements iteratively to
secure delivery of the actual desired outcome of the project. To answer this question we need to define the best possible deliverable which assures the stakeholders that development is progressing toward creating a system that is aligned with their actual requirements. That deliverable is nothing but an executable that its features represent a collection of end-to-end behavioral threads of system usage. In a simpler word, the output of each iteration should be an executable that possesses three very important characteristics:
  • Should cover a set of the actual “behaviors” of the system that are most important to the stakeholders at that point of time
  • Should keep the confirmed “behaviors” that are implemented in the previous iterations
  • All its implemented features composing desired “behaviors” should be working integrated smoothly
You might’ve noticed that I put the world behavior in the quotes. What I am trying to emphasize here is that it’s not important how many features you implement. What’s important is that the implemented features form desired behaviors of the system and do something useful for the business. Another word, only those features that are really needed have to be implemented. Statistics data such as CHAOS report, as I’ve pointed out in another article at http://www.iasahome.org/c/portal/layout?p_l_id=PUB.1008.62, show that a high percentage of implemented features in many projects are never used, even in those projects that are declared successful.
Now that we have set the ground rules and defined what it is that our projects iteratively have to deliver, let’s take a look at the ways of doing that.
To achieve the value delivery, we need to have a way to work around the end-to-end behavioral threads (independent threads of system usage) or what is called scenarios. “Credit Check” and “User Registration” are examples of scenarios. They are independent threads of system usage that tell stories about how for example a user registers and what other things might happen along the way.

The obvious question here would be “How to find the best scenarios and in what order they have to be implemented?”
As I mentioned above, scenarios are stories and finding scenarios is about the requirement gathering skills that is far beyond the scope of our discussion here. But as the principals of Iterative development state, scenarios do not have same level of priority. With that in mind, at every iteration we need to find those scenarios that have the following characteristics:
  • They are addressing the most important goals of stakeholders at that point of time
  • They are addressing the most pernicious risks in the project
Project risks threaten the project's ability to deliver the desired outcomes in a timely and cost-effective manner. Technical risks can be mitigated by choosing scenarios that force the confrontation of the risks. By mapping the risks to scenarios that mitigate them, iterations can be planned to ensure that every iteration reduces project risk and increases the business value realized from the solution.
Relationship between Risks, Scenarios, and Desired Outcomes
So, for example if in our earlier example of user registration, the user information needs to be registered in a LDAP database, the development team need to find risks that are mapped to this scenario, for example limited knowledge of technical team in working with LDAP databases, and rate those risk. If the level of their threat is high then they have to be implemented before other less important risks.
I must mention that addressing risks is not less important than implementing important scenarios for stakeholders. What if a scenario which imposes a big risk is not important when you are in the third iteration and when it comes close to the end of the development a change in business strategy makes it an important scenario? This is a rare case of course. But it can become a counterproductive situation in a project. A smart architect keeps his/her eyes on those scenarios and related business scenarios and as the iterations pass by he/she tries to decide whether the stakeholders’ attention is turning toward those scenarios. Some architects assign a very small team (with One or Two members) to work around some of those risks simultaneously so if their related business scenarios become more important they don’t impose a big danger.
The widest used technique for modeling scenarios is Use Case Modeling. Use cases structure and group requirements by user goals and business value. A use case contains a set of flows, which describe the interaction between the system and its actors. The set of flows provide a map of how the system can be used to meet the goals of the use case and a structured way of describing the set of end-to-end scenarios needed to drive the iteration planning and the development of the software.
A use case captures a set of scenarios, each of which is described by one or more of its flows. All use cases contain a basic flow, which describes the set of normal, "happy day" scenarios. They also contain a number of alternative flows to describe variations, extensions, and exceptions to the basic flow. These can be combined with the basic flow to describe all the other possible scenarios. So, a scenario can be considered to be defined by a use case's basic flow plus zero or more of its alternative flows.
Besides, almost all of the case tools that I have seen out there have a tool for drawing and managing use cases. It seems that it has become a unified way of defining and managing scenarios.
Nothing but an executable code depicts the actual progress of a development project. But not every executable does that unless it implements end-to-end behavioral threads that deliver most important requirements of business. This doesn’t mean that there shouldn’t be any documentation. It has to be, but as much as it’s required.

-- This post has been published as an article on the ITtoolbox web site as well.