From a developer’s perspective, modern engineering practices have a lot of advantages. You can spend less time debugging when you have a suite of automated unit tests to pinpoint problems. Most developers I’ve met who’ve made the switch from conventional coding practices to modern engineering practices have had an almost religious experience with how much it has helped them technically, professionally and even personally.
But that’s not the topic for today. Today’s topic is why modern engineering practices make business sense. Why are they worth investing in? Why will they benefit the organization?
Working at Pillar Technology the past eight years, I’ve had the chance to see several dozen teams using these practices, including pair programming, SOLID principles, test-driven development (TDD), continuous integration and DevOps. Some of the practices seem counter-intuitive, at best, and downright nonsensical, at worst. Especially to managers and businesspeople. But I’ve seen the benefits firsthand.*
Let’s outline the five business benefits I’ve observed.
Where do IT shops spend most of their money? On the new development efforts? Nope. Almost all technology departments spend an overwhelming amount of their budgets on what is called “keeping the lights on.” That means paying for the on-going operations of the data center, help desk, setting up new users, upgrading versions of software and, a very big item, maintaining existing code. Every time you build a new app and finish it, that app goes into the pile of existing apps and has to be maintained by someone. Over time, that pile gets larger and larger and eventually eats the IT budget. Instead of spending so much time optimizing our development process (often 20% of the budget) how about we optimize the maintenance efforts (80%)?
Using modern engineering practices leads directly to reducing the big line item of on-going software maintenance. Developers are thinking ahead to how easy (or hard) this software will be to maintain later. Things like variable naming are a big deal with the new engineering practices. I remember one developer named a switch “Deja Vu” to show that “you’ve been here before. Cute, to be sure, but he really didn’t seem to be worried about someone knowing what he was trying to accomplish five years later.
But engineering practices go far beyond variable naming. Developers look at the
lengths of methods, keeping them short and simple and only trying to accomplish
one thing. They try not to have embedded
for loops, because that
always creates confusion when you’re trying to debug.
In fact, if we can just reduce debugging time after an app is in production, what a giant relief that would be (and money saver)! Modern engineering practices focus in on exactly that. Reducing debugging.
Which leads to probably the number one most productive engineering practice. Testing. No, we are not talking about some separate testing group running special system-wide tests against thousands of lines of code to see if the business functionality is working. We are talking about unit tests written by developers. Lots of unit tests. Great developers create tests as part of writing code, in a process called test-driven development. The results, in terms of reducing debugging time, are just this side of magical. The whole mindset of writing tests immediately while coding creates a situation that I say reduces the “legacy of legacy applications.” A legacy application is the one that people are afraid to touch.
“Here, fix this in the Antiquated-78 system, please.”
“No way! If I make one fix, I might bring the whole house of cards come crashing down! Pick some other victim, okay?”
But with lots of tests (high-quality tests, that is) you can make a fix and then run the tests to see what else you broke. If the tests are small in scope, you can see within a few lines of code right where the problem is.
So we can attack that giant “keep the lights on” budget line item with engineering practices. What else?
Just like we can decrease debugging time post-production, we can do the same for the development lifecycle. People can accomplish so much more business value if they can easily spot their errors with code that is easy to read and has tests wrapped around it.
Pairing is an engineering practice where two developers work side-by-side, two monitors, one computer, one piece of code. One person is called the driver and is writing the code, the other is the navigator and is watching what the driver is doing and trying to help. This practice helps productivity go through the roof. I think of it as “instant QA” because we’re essentially trying to find errors before they are even compiled. Plus, the old problem I remember as a developer of banging my head against the wall for hours because I missed one small thing is often reduced to seconds, because the other person can bring their unique perspective and can point out the error in no time.
Pairing also helps with ramping up a new team member. Remember how intimidating it was to have to jump in to a new application and figure out what was happening? With pairing, you can benefit from an experienced team member’s viewpoint, and they can give you a guided tour of the unfamiliar code. Pairs don’t have to be one junior and one senior. There are lots of ways to pair and ways to learn.
These practices also keep the team from having siloed knowledge. It is difficult to be the only expert in a particular piece of code when people are pairing and using tests and good clean coding principles. Then when that person quits or has to move to another team, there isn’t the huge gap that there might have been. Siloed knowledge costs money.
This isn’t a direct result of modern engineering practices, but it is correlated. Developers using these practices seem to greatly prefer open source development tools over proprietary tools, like those from IBM or Oracle. There is something about being able to look inside the code of the tool you’re using to figure out the best way to use it.
The developers' preferences, in this case, are the company’s financial gain, because open source tools cost exactly nothing in license fees, although there is often a need to pay for support, but the total bill is reduced by half or more from what you’d pay for the proprietary stuff. And the open source tools are updated very often and include the latest features and ideas, because there are so many people contributing to them around the world. The tricky thing can be to pick the “it tool” among the has-beens. There is nothing more frustrating than to start using an open source development tool only to find that everyone else has moved on to the next thing, so the support community is non-existent.
Again, it is hard to show the exact connection, but I’ve certainly seen a correlation between these practices and reduced developer turnover. In our consulting firm, turnover is ridiculously low. I attribute it to our company culture, of course, but also to the practices. The thing is, this type of engineering goes beyond practices, after all. It is more about a mindset, a worldview. It is a line of thinking that developers should write code they feel proud of. Developers should not allow themselves to be in situations where they are rushing to create the next chunk of functionality, abandoning the quality they know they should be building in. As a manager, working with this type of developer is a very different world from regular code-slinging programmers. It is more frustrating for the manager, to be sure, but the results are also a lot better in terms of everything we’ve described in this article.
Although the modern engineering practices and worldview may seem very odd to non-technical executives and managers, there is a significant business case to be made for trying it. If you want an experiment to reduce the heavy costs you’re paying each year for maintaining code, slow development cycles and heavy turnover, consider hiring a firm like Test Double or Pillar Technology to help you get started with a high-quality software engineering approach.
* No, I don’t have scientific proof of these benefits. In order to have “proof” we would have to run identical teams side-by-side where one group uses modern engineering practices and one does not. They’d have to be building the same application and have the same team members on both teams (some developers are more productive than others, after all). Do you see the problem? Yes, these are anecdotal, but my observations are based on several dozen “anecdotes” which hopefully helps my case.
Daryl Kulak is an executive consultant at Pillar Technology. His latest book is called “The Journey to Enterprise Agility: Systems Thinking and Organizational Legacy” (Springer, 2017). It is available in hard cover, PDF as well as audiobook on Audible.