Sunday, January 26, 2020

The pros and cons of Agile

Agile methodology is currently the popular way of attempting to develop commodity software in a factory-like manner.  It's a little hard to define what exactly is Agile and what is not.  I've worked in several companies and with several groups in companies that all claim to be Agile, yet they did things in very different ways.  But they all seemed to agree on a few things, and this I suppose could be the spirit of Agile, if not the letter.

The basic characteristic of Agile is that you break down the software into smaller and smaller tasks until you reach those tasks that can be comfortably handled by a small (2-8 person) team of engineers to complete in a small (1-2 week) time frame.  Then at the end of the time frame, you supposedly have some software that “works” in some sense of the word.  It exhibits basic functionality and the general gist of what the customer wanted, if not satisfying all the requirements for the entire project. Then over the next several time periods of development, the software is iteratively improved by fleshing out remaining requirements, adding needed functionality, hooking components together, etc. During this time, the customer is involved to make sure that what is being delivered is actually what is needed.

Some projects I worked on did this formally by a book and followed strict guidelines for the methodology, others just winged it, but all had the basic characteristics above.

One of the primary advantages of Agile is its use to management.  By having the large problem broken down into team-size, biweekly pieces, management can easily allocate and track resources usage and progress.  They can treat a team as a black box of software development and assign tasks as they arise.  Management can attempt to measure the performance of a team and see whether it is increasing, decreasing, or remaining steady.  Teams are what managers like to manage.

Another advantage is frequent feedback from the customer.  Since after each time period, a somewhat working version of some fragment of the product is available for demonstration, the customer can give feedback as to how and whether this seems to meet his needs.  He can offer suggestions about what might be improved, what features he needs to make the product at least minimally useful to him, and prevent development from getting off track.

But Agile is not a panacea.  There is still a significant amount of software produced with the traditional “waterfall” methodology with specification completed before coding begins and integration done as a final step in coding and only then releasing to the customer.  There is also a fair amount of software written “artistically”. I would define artistic software as that written by a single developer working alone over a period of several months. Frequently, such a project never gets beyond the hobbyist stage, and as such it is a risky approach to writing production code. But on occasion, an artistic project can turn into something novel and useful. It more often exhibits a unity of vision and coherence that is harder to find in software written by groups of people. (Which isn't to say that small groups cannot write software with unity of vision and coherence, it's just harder. Or they'll have one particular person in the group that has more insight than the others.)

Managers aren't as useful to artistic developers. Artistic developers tend to manage themselves. And you cannot swap out one developer for another without swapping out the entire project with him. A manager can work with an artistic developer as a peer, and help manage the project, but cannot manage the developer.

Frequently involving customers has its pros and cons as well. Often customers have difficulty imagining anything beyond incremental improvements to the current ways of doing things. They'll want a UI widget that will make some task slightly easier, but not think of automating the task altogether. They'll want to automate a series of inefficient tasks when a different viewpoint of the end result might make those intermediate tasks irrelevant. Customers are not impressed with changes to the code that don't produce visible effects. You may have spent a week refactoring in order to make it trivial to add new commands and new outputs, but customers don't care. Customers don't care about potential use cases, they care about their specific use case to the exclusion of everything else. This can be discouraging to developers.

Because Agile is so useful to managers, big and intermediate sized companies will continue to use it to develop commodity software in a factory-like style. It isn't going to be replaced any time soon. But there is still ample room in the market for small companies and individuals with vision to carve out niches that Agile methodologies will overlook and find tricky to fill.

(But I'm a romantic at heart, and I like the image of the lone hacker crafting software on his home computer in his room late at night. If only it were easy to make a living that way.)

No comments:

Post a Comment