Agile Adoption – Crossing the Chasm
Why is it so difficult to switch to agile process from waterfall?
Everybody acknowledges that IT has to plays a key role in any new service or product design. Therefore, IT needs to align with business and be flexible to changing business needs. It is a question of how to be agile rather than should we be agile.
Then, why is there such a gulf between the people who evangelize agile processes and those who look at them with great suspicion?
Any application is built to satisfy a business goal. The process has two major steps where the step one is to define the application behavior which will help in meeting the business goal and the step two is to translate the application behavior to a working application. The definition of the application behavior is the requirement specification and it is the link between business and IT.
In the waterfall methodology, development starts after the requirement is frozen. It is assumed that if business users can unambiguously and comprehensively specify the application behavior, IT can build the application satisfying the stated behavior. All the management practices based on waterfall methodology is build around this assumption.
However, in most real life project we find that one of the following things have happen.
- The stated behavior as understood by the business is different from the understanding of IT
- The stated behavior is a sub-optimal solution for the given business goal
- The business goal has changed
As a result the final application does not meet the business goal. Then it becomes a blame game between business and IT, where business claims that “this is not what I wanted” and IT claims that “I have delivered as per your signed-off requirement”. Since there is a gap between desired application and what is stated in the requirement, it is concluded that the root cause of the failure is improper requirement definition. In the next project more rigorous process is applied to make the requirement definition more comprehensive and as early in the lifecycle as possible. In spite of all these rigor, one of the most frequently cited reason for project failure is improper requirement management.
In current business context, is it possible to freeze requirement?
If we go back in time, most IT applications were written to automate internal processes. The emphasis was on adopting industry best practices. Vast majority of the user were internal to the organization and they had to adopt to the application interface rather than the application interface being tailored for the user needs.
Times have changed and many of the IT applications are an integral part of the product or service offering. Significant part of the users are external to the organization. The focus has shifted from following the best practices to having innovative offering. The new motto is usability and user centric design. Though there are guidelines and principle available on how to design such application, in practice it involves many iterations where many idea needs to be tried before a workable solution is arrived at. Sometimes, applications may have to be modified very late in the development cycle.
In such cases, though the requirement is driven by a business goal, business users are not always sure about how the system should behave. It is necessary to try out the application before taking decision on its suitability. To this we can add the changing business goal arising out of change in the environment.
If all these points are added up, it becomes clear that requirements will change.
This has led to the realization that we need agile processes which works when requirements are constantly changing rather than spending time and energy on freezing requirement. As a result agile processes were born. The key principle behind any agile processes is to have a mechanism where the users can try out the application as it is being built and give feedback so that:
- Gap in understanding is reduced
- Sub-optimal solution can be improved
- Application can be realigned to any change in business goal
It is easier to adopt agile methodology in letter than in spirit. An iterative development process becomes ineffective, unless user feedback is used to change and fine tune the application behavior. This may involve introduction of new features, discarding features and reprioritization of features. It is not always easy to reorient an organization entrenched in waterfall development.
Not only it is necessity to change mind-set of the people involved, it is also imperative to modify some the key management practices.
Funding: How is project sanctioned?
Normal process of budgeting involves calculating return on investment, which in turn, requires both investment and the return on it to be quantified. Even when the requirement is clearly defined, arriving at an accurate effort estimate is an error prone exercise. When requirement is expected to change during the life of the project, it becomes almost impossible to have an accurate estimate up front. This becomes a hazardous exercise and in practice the estimate gets tuned to make the ROI look good. Once sanctioned, the budget becomes frozen and it acts as a great disincentive to allow flexibility in changing requirement mid way.
The way out of this situation is to have a method of incremental funding of project.
Initially funds can be made available for first few iterations and this sanction can be based on a reasonable accurate estimation. The scope of the project and the fund required can get progressively fine tuned. With incremental funding, it also becomes easier to detect unviable project much earlier in the lifecycle.
Interaction: Do stakeholders trust each other?
Change in requirement involves rework and it implies discarding work already done. IT has to trust that a specific change suggested by business is needed and could not be specified earlier.
Similarly, a feature may look simple from outside but may involve considerable amount of programming effort. Business has to trust the effort estimate made IT.
Both IT and business has to believe that understanding is progressive and any gap in understanding is neither deliberate nor because of incompetence or carelessness of the other party.
It can be achieved by having transparency from both side. IT needs to establish a method of estimation and make it transparent. Similarly, business needs to share the business plan and explain the need for any change.
Any change will involve unplanned effort, more effort than what is feasible within a given time frame. Feature will have to be re-prioritized and and the revised plan should have joint ownership.
Defect: What is a defect?
Standard method of measuring defect is to:
- Define and freeze the requirement
- Derive test cases from the requirement
- Use the test cases to test the software
- Measure the number of test cases that has failed
Zero defect software means all the identified test cases are passed. This does not in any way indicate that the software meet the business needs. Since, test cases can only be identified for the explicitly documented requirement, we need to move beyond that and reorient how defect is defined.
Therefore, regardless of presence of any explicit test case, a defect is when the application …
- … crashes
- … behaves in an illogical manner
- … does not behave in the manner expected by the user
The responsibility of identifying and eliminating first and second category of defect is with developers. The third category can not be done by the developers alone and needs an active involvement of the user.
Success: How is it measured?
Traditionally, the goal of successful project management has been to the deliver the application on time, within budget and without defect. All project managers are brought up on the belief that the key to achieve this is to freeze requirement and minimize rework.
However, when the planned application is a part of a major product or service release, the cost of failed application can be many time to cost of the application. Therefore it is necessary to shift the paradigm and measure success in terms of satisfying business needs. It can only be done by an agile methodology which interactively refines requirement by allowing the user to work with and feel the application on a regular basis.
When we look at some of the most successful software company we find that it is better to deliver the right application even though it may be late, over budget or have some bugs than to deliver the wrong application within budget, on time and without defect. The likes of Microsoft, Apple and even Google has slipped on their promised delivery and the delivered software are also not free form problem. That has not stopped them from becoming the more successful than anybody else in what they choose to do.