How to build great software
1. Software must be designed in detail before development starts, so that a clear plan can be out-layed.
The more complex a design, the more like software the design itself is. By perfecting a design, then writing the software to that design, you're effectively writing the work twice. Instead, by doing just some simple design sketches and data modelling rather than a book-like design, a good development team can create a shell for the software and efficiently refine it towards the finished product. This process of refinement creates natural prototypes, allows easy adaptation when issues that would be unforseen by a design arise (or brought up as fresh concerns by a client), and the total process takes significantly less time. To pull this off requires a close team, skill, and experience, but it is by far the best option for the majority of situations.
2. There are programmers, designers, analysts, and users.
By structuring development so that all developers get some exposure to each part of the development process, skills may be shared and greater insight may be gained. If developers are encouraged to actually use the software then they can use that expertise to think of improvements that otherwise would not come to light.
3. A happy team is a productive team.
A team of people with a wide variety of natural skills, experience and concern, that criticises each other and argues vehemently over the smallest details, will bring up and resolve issues that otherwise would never be tackled. A furnace of relentless argument is the best way to forge understanding and reach perfection.
4. It's important we understand our direction and don't compromise with it.
Life is compromise, and compromise is not a weakness. There will always be issues (such as efficiency, budget, ease-of-use, power, scope, and the need for easy internationalisation) that cannot be simultaneously met without such compromise.
5. We know what the client wants, we know what the issues are.
Without constant re-evaluation, it is easy to lose track of the objective. Developers are often faced with problems to solve that they consider the issues, when those are in fact separated from the actual market goals and can become totally irrelevant. Developers must always understand the market goals and be able to adapt when other things change, or even the goals themselves change.
Features can easily confuse users, and their actual value should always be considered against the cost of confusion. In some cases it is sensible to actually remove working features due to such concerns.
- The customer is always right.
Most customers try hard not to look ignorant in front of software developers, and hence phrase their suggestions in a technical way. The effect is that often suggestions aren't really appropriate, because they're not founded on a solid understanding of technical issues.
- The customer is often wrong.
Although customers needs are often not best met by doing literally what they say, they always know what they want and why they want it - and usually for very good reason. Understand them and adapt what they say, discuss with them, but never ignore them.
A bad workman blames his tools. Whilst some development tools aid development substantially, a good developer can do great results in most things served to them. There are a few exceptions, like Microsoft Access, or assembly language, but generally speaking the difference in quality results is much more due to the skills of the developers than the quality of their tools.
The interface doesn't just need to be easy-to-use, it needs to be navigatable without an overall systems understanding. Screens need to be self-describing.