Saturday, April 23, 2011

The Generalize Agile Development and Implementation Process for Software and Hardware

In 2000, I created a Rapid Application Development (RAD) process, based on eXtreme Programming (XP) that was CMMI Level 3 conformat. Conformance means an outside audit was performed and the process conforms to the requirements of the CMMI level 3 key practices (note that CMMI levels 4 and 5 are organizational practices rather than process practices, so the Level 3 is as high as a process can get).

Advantages of the RAD Process
This RAD process, as I created it, has several advantages over traditional software development efforts, like "waterfall process".  The following advantages derive from XP.
  1. Works in monthly to 6 week cycles, enabling the customer to start using an Initial Operating Capability (IOC) version of the system, while the system in under development.  This meant that the customer was gaining some value even while the application was being implemented.  In the case of one small but unusual asset management, accounting, and accounts payable system, the customer, in the process of loading data into the IOC version found enough errors in the accounts to recover more than twice the cost of the effort to implement the application, that is, $100K+ errors during the implementation of a $50K system.
  2. Enables the customer to get their highest priority functions.  The RAD process does this by allowing the customer to add requirements during each development cycle and reprioritize the requirements between cycles.  This produces a much more satisfied customers, than "Big Bang" processes like "the waterfall" process.
  3. The development process is agile because it allows the customer to reprioritize their requirements.  In a book by Dr. Ralph Young on requirements, he cites a statistic that only 49% of the real customer requirements are known at the start of the development effort.  This RAD process allows the 51% to be incorporated, while keeping in budget using this repriortization function.  Since the product always meets the customer's highest priority requirements, they are much more satisfied with the result.
  4. Focuses the developer on developing, gathering than documenting--this is what the software developer is good at, while documenting isn't.  Consequently, the developer's morale improved (which meant their effectiveness and cost efficiency improved) so both the quantity and quality of the products improved.
  5. Minimizes the number of intermediate artifacts, that is, it eliminates the need to create documents like the Program Management Review (PMR) reports, status reports, design documents, Engineering Change Orders and so on.  This greatly reduces the need for developers and the program manager to create much of the documentation of other formal development processes.  In fact, as I created it, this RAD process eliminated the need for PMRs, because the customer was involved with the development team on a weekly basis.  And since the process assumed "that the program/project doesn't have all of the requirements at the start and enables the customer to add and reprioritize the requirements throughout the process, the resource requirements are kept level.  Therefore, from a programmatic perspective, this is a Level of Effort.  This means the same amount of money is spent each period; which reduces management of cost to nearly zero.  And since the implementation schedule is based on an unknown set of requirements, there is no way for a program manager to really create a schedule at the start of the effort and keep to it.  Instead, there is a monthly agreed set of requirements that must be fulfilled in development and rolled out--which by the way works, but which makes transaction managers and finance engineers paranoid--because they have so little CONTROL.  However, this process produced products of both a higher quality, and much greater quantity, and much more satisfied customers than the previous CMMI Level 3 processes.
  6. Like XP, I used Use Cases to gather requirements.  A Use Case is One way for one type of actor (either a person in a particular role, or another application) to use the application under development.  In my experience and in the experience of many of the 50+ systems engineers that I coached or mentored, or otherwise trained that Use Cases were the best way to gather requirements because they could be documented in ways the meant something to the customer (they can be documented to use the customer's own words); most other ways, like "shall statements" communicated much less because they are a transliteration of the customer's requirements.  Additionally, in general, they communicate as much, if not more to the developer.  Since the goal of requirements identification and management is to "communicate the customer's requirements to the developer and to ensure that these requirements are met", using the Use Cases, as recommended by XP proved both highly effective and cost efficient.
The net of all this is more satisfied customers, happier developers, and much more effective and cost efficient processes and products--in this case a real live "Win-Win" situation.  One consequence was that by 2005, literally, hundreds of small and medium (and even a few large) effort adopted this RAD process--and it proved highly successful within its domain (software development).

Limitations of Rapid Application Development (RAD) Processes
The key limitation of the RAD category of formal processes (processes that meet CMMI Level 3 practices) is that they are strictly for software development.  Formally, it is not possible to use this process for software or hardware integration, and for linkage to systems external to the organization, those activities that are most likely to occur in today's SOA and cloud-based environments.  Since most organizations purchase software and integrate it into systems rather than develop their own, a short cycle, agile implementation and transformation process is needed for such efforts.  This process should be the equivalent  for integration of the RAD process  that I designed and documented for software development.  In fact, it should have the same advantages as the RAD process, but applied to integration and tailoring of systems.
The Generalized Agile Development and  Implementation Process
Starting in about 2005, I gave thought to this Generalized Agile Implementation and Development (GAID) process would work.   This is a notional design and the challenges in designing such a process.  The notional design consists of four loops.
  1. The Requirements Identification and Validation Loop (Approximately 3 Months).  The purpose of this loop is to work with the customer to identify their requirements (the Systems Engineering Role), to decompose and derive the functions, to structure the functions, to allocate functions to components, to determine the acquisition method for components through trade off studies (make/buy/rent) (all in the System Architect role), and to validation that the product meets the customer's system requirements for this loop and all prior loops.  Additionally, each three month cycle allows for the acquisition and installation of any needed software, hardware, and networking components.
  2. The System Assembly/Implementation and Verification Loop (Approximately 1 Month).  The purpose of this loop is to assemble the various sub-assemblies (either those developed or those acquired from a supplier) and to verify that they meet their allocated functional and design constraint requirements.  Additionally, this loop enables the customer to see the overall progress of the product and suggest minor changes that will enhance the utility of the product for the customer. For example, this might include the placement of various data windows on a particular display.
  3. The Sub-system Design and/Assembly and Verification Loop (Approximately 1 Week).  The purpose of this loop is to ensure that components perform the functions specified in the component requirements to the level of dependability required and that the interfaces among components match for integration.  Additionally, this loop give the customer a first loop at the components, as they are being implemented, which means they can communicate changes they to like to see to the designers before the product start full up assembly and testing.  This greatly reduces the cost of the changes.
  4. The Component Construction and Verification Loop (Daily).  This is the loop where all of the detailed design, development, and COTS product configuration takes place.  While this loop might strike the reader as silly, all loops including this consist of engineering functions with a single program management meeting at the start/completion of a loop.  In this loop, it is a 15 minute long "stand up" meeting.  The reason for the meeting is to identify risks and issues, and to have the implementers set aside time to reduce the risk or close the issue before the next stand up meeting.
If this works like the RAD process I developed worked (and it should), after the initial period of learning the process, both the customer and the implementers will find it much more straightforward, simpler, and easy to use, while producing much more effective and cost efficient products.

No comments:

Post a Comment