Archives

Archives / 2022 / July
  • Thinking About the Costs of a Software Feature

    (This article was inspired by parts of my talk “Things I wish I had learned earlier as a developer” at the Developer Week 2022 in Nuremberg, Germany)

    More than just coding

    The total cost of a software feature comprises more than just the time spent on writing the code. When you think about the steps from the first idea to the release of the product, various contributing factors come to mind:

    • The work necessary for turning a rough sketch of an idea into a viable concept
      • What exactly does the feature do? And what not?
      • What is the actual benefit for the users?
      • How are users supposed to use it? In what context?
    • Time spent on planning the work
      • Who is available and has the skill set to implement the feature?
      • Does the feature influence / depend on other features? Does this force a certain order?
      • Is this a risky feature that better should not be started shortly before the coding freeze prior to a software release?
    • Effort that goes into preparing the coding part of the development
      • Which technology / framework / library should be used?
      • “Buy vs. Build”: What are advantages/disadvantages?
      • Are the licenses of all required components known and, most importantly, compatible?
    • Time spent on writing the code
      • “Known knowns”: Writing the code you know you need for the planned feature.
      • “Known unknowns”: Finding the best solution by trying out which of the different imaginable approaches is best.
      • “Unknown unknowns”: Dealing with unexpected issues regarding performance, reliability, usability, or user experience – sometimes you only find out when “the real thing” is running under realistic conditions. Also: running into bugs in tools / frameworks / libraries beyond your control.
    • Testing the software
      • Do you write unit tests? And/or integration tests?
      • How do you test the UI?
      • What are the economics regarding automatic/manual tests?
    • Documentation
      • How much external and/or internal documentation do you need?
      • Can you get away with not writing documentation?
      • Who can write the documentation? In what quality?

    Wait, there’s more

    A feature can also create costs after the release. Some features more, others less.

    • Fixing bugs
      • You need to reproduce the bug and preferably write a failing test.
      • For critical bugs, you may have to stop your current work immediately for a hotfix.
      • For a long-existing bug, fixing the bug may break expected behavior that other people have worked around.
      • Worst of all: There is always a chance > 0% that the bug is caused by something beyond your control, and that the bug either may take a long time or may not be possible at all.
    • Dealing with feedback
      • You need to determine whether you are dealing with valid concerns or the opinion of a loud minority.
      • Does the feedback lead to extending or changing the feature?
    • Managing change
      • Changes need to be communicated to customers – there is a high probability that, no matter how much effort you put into communication in the past, certain people will be surprised by the change.
      • Does the change involve some kind of migration? Manual migration may require documentation with step-by-step instructions. An automatic migration, e.g., of configuration data, may have to deal with a variety of different settings as the result of repeated upgrades or manual changes. 
    • Dealing with roadmap issues
      • Users of your software view existing features, including the feature you just added, as “the way things are intended”. For example, in terms of how data is processed or what the user interface looks like. If future features follow a different philosophy, this needs to be communicated. And you may have to deal with habits your users have formed because of your feature.
      • A quick decision while developing a feature (e.g., what a single click or a specific hotkey does) may be short-sighted and block future development. This either requires a workaround or managing change (see above).
    • Handling end-of-life
      • Do you simply remove the feature? Do you add a configuration switch and keep it switched off by default?
      • Are existing users still able to perform their specific workflow? If switching to a new workflow is too complicated, or maybe not even possible, customers may remember the other shortcomings of your software and switch to competing products.
      • Always remember that no matter how specific and little-known a feature is, there is a probability that somebody uses and relies on this feature. You may never hear any feedback, even after announcing the removal (who reads release notes, after all). But once the feature is removed, chances are that somebody is upset and lets you know.

    Where to go from here: The great filter

    The cheapest feature is the one that is not implemented without anybody missing it. If you feel a feature is important enough to start work on it, you should first answer one simple question:

    What’s the real-life scenario?

    When being asked to explain the purpose of a proposed feature, are you able to go beyond “one can do X with it”? Are you able to tell a believable story with a realistic protagonist in a certain context, who encounters a problem that can be solved by using the feature? And does this story make sense when telling it to other people?

    If not, that feature may be a candidate for not spending budget on it.

    Want more?

    If you liked this non-technical article, you may like the other articles of the “Thoughts” category, e.g.,