To explain our choice of methodology a little history is required. HMW Computing was founded in 1981 a time-sharing services company primarily working in APL, an interpreted mathematically focused programming language.
Our early work was on mainframes and dial-up services in the management accounting and insurance business areas. Usually the team were small - 1 to 3 developers working directly with the users of the system who usually supplied requirements in a vague informal manner.
After some years we commenced a long period of development and support work in currency trading systems. This involved a larger team, up to 14 company and client staff, and a more rigorous to development, but didn't improve the informal way requirements were provided by the users of the systems. During this period we devised tools (code repositories change management software) and techniques (software release documentation, process and data definitions) based on existing methodologies (Arthur Anderson's Method One ) James Martins Information Engineering and ...L
In the early 1990s we adopted DSDM, as it seemed to be a formalisation and vindication of our methods. DSDM did have 5 phases, with two sequential steps & 3 iterative, repeating ones. It seemed sensible to have a project initiation followed by prototyping phases, culminating in implementation of the solution.
The problems are that this still tends to impose a linear approach to the development, it doesn't sit well with mixing design of a part of the system and manufacture of the code in the same step (although DSDM does say this is possible), the continuous release of methods such as extreme programming and the latest release of DSDM called Atern adds pre- an post-project phase, the idea of three focus streams (which tend to fragment the method), plus a host of other required deliverables. This runs counter to the principles of an agile methodology and is too rigid.
We no longer feel DSDM is an agile method, it has become more of a tool to constrain the "wild men" of agile development in a harness that management can understand.