Software as knowledge repository

Summary Software is the most relevant representation of knowledge about a business. It is crucial that this knowledge be accessible to decision-makers. When managers resent refactoring as "not adding business value", they overlook the value of software as a knowledge repository.

Software is the most relevant representation of knowledge about a business. Although it only implements automated workflows, those form an ever-growing part of business activities. Source code is the only 100% accurate and up-to-date documentation of such workflows and often reveals a lot about manual processes as well.

More importantly, software is executable knowledge. It doesn't simply describe automated processes - software becomes the process once running in a production environment.

The problem of access

It follows that a decision-maker trying to truly master the processes in a business needs fast, flexible and efficient access to software-bound knowledge. What's more, she needs read-write access in order to augment, extend and deepen the knowledge. It's a non-trivial requirement:

  • In software, business-specific knowledge is intertwined with lots of other concerns (infrastructure, ergonomics, security, governance, ...).
  • Source code addresses several audiences at once (end user, other programmers, compiler).
  • Software tends to be heterogeneous even in modestly-sized businesses (different authors, different programming languages and platforms, ...).
  • Software has to be precise enough for a computer to execute (i.e. extremely precise by human measure).

The way software is usually managed presents additional difficulties. Business knowledge gets "baked in" at various stages of a complicated dance involving decision-makers, analysts, architects, programmers, testers - oh, and hopefully also end users. The transformation progresses via specification documents, UML diagrams, Scrum stories, bug reports or what have you. Given sufficient skill and will on all sides, there is reasonable certainty that the software does what decision-makers intend it to do. The process is neither flexible, fast nor efficient, however. Improving this situation surely means competitive advantage.

Distilling knowledge

Many approaches have been tried to bring stakeholders closer to software (and many more will be tried as software grows more important). They often start by isolating business-specific parts of a program (a.k.a. "business logic") from the rest. The idea is sensible enough yet fraught with pitfalls:

  • The boundary between generic and business-specific code is often blurred.
  • Business-specific algorithms can exist at many levels of abstraction.
  • Local peculiarities creep into unforeseen places (masked as validation criteria etc.).
  • Isolating parameters is much easier than isolating an algorithm.

The last point is especially relevant. Many businesses end up with an agreed set of easy-to-tweak configuration settings, complemented by a change-request process that supposedly balances flexibility with maintainability. Such a set-up makes it all too easy to make the wrong trade-offs under schedule pressure. Another possible outcome is programming by configuration, i.e. spiking a program with so many parameters as to make both the code and the configuration extremely complex (this is the shadowy domain of SAP consultants).

Matters of presentation

When it comes to giving decision-makers useful access, extracting algorithms as well as parameters is clearly essential but won't suffice unless the CEO is fluent in the language in which the code is written. Making business logic palatable to non-technical folks is therefore a major area of activity featuring its own array of methods and tools, with Business Process Management (BPM) apparently in vogue these days, complemented by Business Rule Mangement (BRM).

Such tools are definitely useful but they only go part of the way. Even with declarative approaches such as BRM, understanding software requires algorithmic thinking. It's a hurdle no programming language nor diagramming technique will overcome. Algorithmic thinking is a skill and can be learned, however. The recent debate regarding whether everyone should learn to code is, at its core, about letting people comprehend the world around them as software pervades it. The concern is especially relevant for business decision-makers.

It will take a generation shift (maybe several) for algorithmic thinking to become common among non-programmers. Mediation between decision-makers and software is therefore poised to remain a feature of the business world for the foreseeable future. As mentioned, the process tends to be rather chaotic in practice and varies immensely between businesses. What's universal is the ever-stronger imperative to make it fast, flexible and efficient.

Perils of mediation

This is where agile methodologies enter the picture. The Agile Manifesto reads like a sigh of frustration with the communication barriers between programmers and other stakeholders. Yet it tackles precisely the interface between decision-makers and the software they pay for - which, as is hopefully clear by now, is a really tough nut to crack.

Agile has endured some backlash over the years but it represents great progress nevertheless. It clarifies that there are aspects to developing software that don't have technical solutions. That's always been obvious to managers but putting it in words programmers can understand was certainly commendable. Overall, agile practices do result in improvements when applied properly.

The most common cause of failure in agile projects seems to be misunderstanding and mistrust of the agile principles on the part of key stakeholders. The principles are thus not applied consistently and the project ends up a half-hearted faux-agile mess.

Two agile values are particularly difficult yet vital from the perspective of software as knowledge repository:

  • Working software tends to be misunderstood by non-programmers - they don't appreciate how fragile working software is and how easy it is to introduce bugs. Hence the ignorance of test coverage and the general disregard for software quality in the presence of other priorities.
  • Responding to change tends to be misinterpreted by programmers and non-programmers alike as giving in to schedule pressure. Requirement analysis becomes sloppy, necessary maintenance is skipped.

Both pitfalls lead to low-quality code that makes business-specific knowledge catastrophically opaque. The problem can arise surprisingly quickly and the consequences are serious. Requirements are not met, changes take lots of time and introduce bizarre bugs, developers quitting cause undue disruption - a typical project in disarray.

Refactoring to the rescue

The agile practice of refactoring aims to achieve high software quality without slowing down the pace of development, even increasing it. Briefly, refactoring means changing software in small incremental steps and confirming the changes through automated test runs (the test runs ensure the behavior of the software stays the same as its structure improves).

Refactoring is intended to be used early and often, preventing the build-up of hard-to-understand code. The proposition is counter-intuitive because it sounds like extra work. As a result, many teams never adopt the practice. It's a shame: refactoring does speed up coding by slashing the cognitive load imposed on the programmer. In other words, there are far fewer things to worry about when modifications are modest and unit tests are robust.

Why no-one does it

What happens in practice is that software quality is left to atrophy until a developer comes to a decision-maker and says "Hey, we really need to refactor this." Time is alotted but it's not enough so big chunks of the code are cleaned up haphazardly with no test coverage. A flurry of new bugs appears and refactoring gets another blemish on its reputation even though it was never brought to bear. Overcoming this anti-pattern requires adaptation on both sides:

  • Software developers must get serious practical exposure to true refactoring as part of their training. The impact has to be experienced to be appreciated.
  • Decision-makers must admit to themselves how critical sotware is to their organization and stop treating it as a "cost center" or an obstacle to achieving business objectives.

Why you should do it

A well-maintained body of code is valuable whether or not it's directly readable by non-technical decision-makers. Knowledge is at hand, answers are quick, changes are smooth. The company becomes nimble in a way it never could without automation. At a software startup, this notion forms the core of its competitive strategy - that's why working at a startup is so alluring to programmers. It also drives startups' disruptive potential. Decision-makers at conventional companies face the tough task of integrating the value of software into their cultures even if they don't compete with startups. Once they do it may be too late.

Proudly powered by Pelican, which takes great advantage of Python.