DEV Community

loading...
Cover image for Dev Is Not Construction. It's Medicine.

Dev Is Not Construction. It's Medicine.

bytebodger profile image Adam Nathaniel Davis ・8 min read

We've all heard the analogies. Someone's trying to explain some aspect of software development and they inevitably compare the process to... building a house. You've probably done it. I've done it more times than I can count. And with certain examples, the analogy might be helpful. But overall, I've become convinced that this false equivalence is downright harmful. Why??? Because software development isn't much like construction at all. It actually has far more in common with healthcare.

This isn't just semantics. I believe that, if more people could apply the proper paradigms to software development, they'd get far better results from the process. Framing a software project in construction terms isn't just misguided. It could very well lead you to mismanage the project altogether - and anger those who are working so hard to make it a success.


Alt Text

Throw Out Your Blueprints

So much effort is wasted on chasing the fever dream of the "flawless spec" (blueprint). This (misguided) idea implies that if you could just craft the perfect set of requirements, you could then simply hand that blueprint to the dev team, and they could crank out the app exactly as you'd planned it, as though they're coding automatons. But this ideal is broken.

A detailed blueprint implies that the stakeholders know what they want. A bloated spec, chock full of reams of minute requirements, implies that the stakeholders absolutely positively know, beyond any shadow of a doubt, exactly what they want.

After slinging code for a quarter century, I can confidently tell you that I have never met a single stakeholder who truly, fully, completely knows what they want. Not one.

Of course, they have strong ideas about what they want. And they may have already done a great deal to define every aspect of the proposed app. But I can guarantee you that they don't 100% know what they want before we start writing the code.

Because, once they start seeing the screens in front of them, and they start playing with the features that have been completed up to this point, they inevitably start realizing that their original vision was, at best, incomplete. At worst, it may have been outright wrong.

You see, there's a name for software that can be 100% quantified. It's called a finished product. So unless you plan to perfectly copy a pre-existing system, there's no way that you've absolutely defined all the specs for your new project upfront.

Your home was probably built from pre-defined blueprints that were developed months or years ago. They've been used for other houses that look just like yours. And the process of building your house is truly part of a repeatable cycle. They print out another copy of the same blueprints. They may even hire the same workers. And in a predictable period of time, they can spit out another house just like the others that they've previously built.

But when you're building a brand new app from scratch, there is no such thing as a perfectly-defined, cookie-cutter development process. If you're paying a team to build custom software, it's because, on some level, you need something that's truly unique to your needs. And as long as that's the case, you won't ever be able to just hand the blueprints over to the dev team and know that they'll crank out your app 100% to-spec in a few months.

This "every-project-is-a-little-different" quality doesn't readily lend itself to construction. But it is a solid way to describe the process of providing healthcare.



Alt Text

You Can't "Dumb Down" Programming

In medicine, there's really no such thing as a "routine" procedure. There are just procedures that have extremely-high success rates with low likelihoods of complications - and there are procedures with much lower success rates and much higher likelihoods of complications.

Even a "routine" procedure can become quite risky if the patient has other complicating factors. It doesn't matter how many tomes are dedicated to medical procedures. There will never be a day when you can hand a set of pre-defined "specs" to a junior brain surgeon and be confident that he'll be able to successfully complete the surgery with no supervision.

I've noticed many trends in corporate America where there's an implied desire to make application development like some kind of assembly line. You stuff the detailed instructions in the front end. And eventually, your new app comes shooting out the back end. But once you realize the parallels with healthcare, you need to ask yourself: When YOU are going under the knife, do you want to put your faith in an instruction manual? Or do you prefer to trust the training and the experience of those who are performing the surgery??


Alt Text

Meet Your Devs

It may sound silly (or arrogant) to compare a programmer to a surgeon. For most of my career, I wouldn't have done it. Then a funny thing happened to me: I tried to teach someone else - a complete novice - to code. That's when I first started to appreciate the sheer volume of highly-specialized knowledge I'd packed into my brain.

The nature of my work may be nowhere near as important as a doctor's. But I'm pretty confident that the volume of technical info floating inside my head is comparable to a surgeon's.

I'm highly proficient in at least five different programming languages. I have reasonable experience in several others. I'm proficient-to-expert level in many other "supporting" technologies like HTML, CSS, SQL, REST, SOAP, GraphQL, etc., etc. I can develop in mobile, web, or console platforms. I can flip comfortably between frontend, backend, and middleware technologies.

The point here isn't to brag about my resume. The point is that when you start quantifying all of the technical details I know, across all of the platforms in which I can operate, it constitutes a metric crap-ton of knowledge. And I'm not just talking about myself.

Do a deep dive with nearly any dev who's been in this game for a decade-or-more and you're likely to find a comparable (or greater) level of knowledge. Work this career for long enough and you'll accumulate a staggering library of critical data.

It's not arrogant or boastful to state that many devs are as highly skilled in their field as surgeons are in theirs. And yet, it often feels to me like "the business" keeps looking for ways to relegate programmers to being assembly line workers.



Alt Text

Diagnosing

If I have a detailed set of specs that explain exactly how I'd like to have a cabinet built, I could probably take those specs to a skilled carpenter and, for the right price, receive a fully-built cabinet in a reasonable period of time. That's how many organizations try to approach application development. They want to just shove the specs in your face and say, "Here. Build this."

Now imagine that I schedule a first-time visit with a physician. When I walk into the office, he asks the reason for the appointment. I give him a big pile of medical documents that explain how to do a particular medical procedure.

I ask him how soon he can perform this procedure on me. I also tell him that I need a detailed estimate on the cost.

He wants to know why I want this particular procedure. He wants to know what symptoms I'm experiencing. He wants to know all sorts of pesky details about my medical history.

But I brush all of this off and I just keep pointing at the medical "specs" that I've given to him. I tell him that he just needs to deliver the medical care as I've outlined it in the specs. How well do you think that appointment will turn out??

The physician isn't there to serve up medical procedures as though you're ordering them off a menu. The physician is there to solve problems. Any skilled physician is a problem-solver first, and a medical practitioner second.

Even if we assume that my ridiculous demands are correct, it would still be irresponsible for any physician to perform the procedure without first understanding the problem. But too often, we hand specs to devs, expecting them to deliver a product without properly understanding the problem that product is designed to solve.

Think I'm being dramatic here? Consider that quite recently I had to deal with a client who kept changing specs on me. They kept making arbitrary requests in the 11th hour of the project that had very real technical consequences. And with every new request, I kept asking, "What is the problem that we're trying to solve with these changes?" But they didn't really want to talk to me about problem-solving. They just wanted to talk to me about the specific request.

If I understood the problem, I could properly consult on the best options to solve that problem. But if I'm routinely kept "in the dark", it hamstrings my ability to be a proper problem solver.


Alt Text

Emergency Techs

Doctors don't grill you for information because they want to make you uncomfortable. They try to gather all relevant data because it vastly reduces the likelihood of emergencies - and helps them respond better when those emergencies occur. Experienced devs are very similar.

No matter how glorious your specs may be, the simple fact is that, once I get nose-down on the project, I'm gonna run into... something. Something unexpected. Something that alters the original goals of the project. Something that alters the timeline. Something related to ugly legacy code. Or something related to an uncooperative vendor. Doesn't matter what that "something" is. On any sizeable project, there's always... something.

When organizations try to insulate programmers from the process of building applications, they leave themselves poorly equipped to handle these emergencies. They tie the developer's hands and force the triage process through committees and reviews - and the project suffers for it.

I've had many scenarios when I encounter some kinda unplanned hurdle during the project. More often than not, I can quickly assess what's wrong, and I usually know exactly what should be altered in the project to solve the unforeseen issue. But I can't actually fix it. Because, even after a quarter century in the career, I often find myself without the authority to make that call.

So I ping my PM and explain the issue. The PM calls a meeting with our client POC. The client POC calls a meeting with all the project stakeholders. Sometimes it even requires another meeting to specifically loop in the client's "leadership". Frequently, I'm pulled into every one of these meetings, to repeatedly explain the issue to the latest set of actors.

After weeks of back-and-forth (sometimes, months), the client finally makes an official decision on how to handle the issue. It's usually identical, or extremely similar, to my initial recommendation. And once everyone has signed off on the change, I can finally get back to work on the project.

If that project were a patient, the patient would have long-ago expired.

Discussion (16)

pic
Editor guide
Collapse
entomy profile image
Patrick Kelly

I was a bit confounded when you said you were going to make this comparison. Having read it now, I agree with the majority of your points, and think this is an apt comparison. The one major difference is urgency; medicine often has incredible urgency where programming can take a step back and think when needed.

To provide a concrete example towards something you said. We definitely don't know what we're getting ourselves into most of the time, and being flexible and reactive is a huge part of it. Even without medical knowledge, I'm sure the first thing you'd think for a patient coming into the ER with his "arm torn off" is a tourniquet. You'd be correct. So, I had everything ready ahead of time expecting that, with no complications, since that's very urgent. EMS neglected to mention this was a farming accident and the man didn't just have his arm torn off. He also had cow manure everywhere, including in the wound. The entire "playbook" went out the window, because now we have to carefully balance cleaning out the wound so he doesn't get sepsis, with stopping the bleeding so he doesn't bleed out. Curveballs, not necessarily this severe, are the norm, and it's expected any medical worker can think on their feet in response to new data. The "spec" is always the same, but what we think the "spec" is constantly changes.

A more technical example, is I applied this kind of thinking to some software I develop, since, it's what I know. I had some rough design requirements; a problem I was trying to solve. But that was it. How long would it take? Idk. How would it be implemented? Idk. It wasn't TDD because I didn't have a concrete spec; I couldn't. I just dove in, keeping the hard requirements in mind. What started as an uninspiring and "misguided" project many tried to talk me out of, turned into a major contender among parsing frameworks. I made constant little UX and perf tests, and refined everything until I eventually had a spec; the spec, or "diagnosis" was a late product. The design I have barely resembles what I thought it would be. No amount of design docs or meetings could have gotten me there. But I wound up with what seems to be the worlds first allocationless parsing framework, reliably in the top 3, and usually 1st for every performance benchmark I've seen (near 100 cases now), and with far less required knowledge than parser combinators need for anything non-trivial.

Looking at it both ways, I completely agree. I liken what's being done now, to the philosophers of old thinking if they thought long and well enough, they could know, upfront, how many teeth were in a sheeps mouth. You can't. And that you can't spawned sciences hypothesis-test-infer approach. This is Computer Science not Computer Philosophy right?

Collapse
bytebodger profile image
Adam Nathaniel Davis Author

Excellent points! And I love that you bring up the term Computer Science. I always thought that term was a bit "lofty" for what we do (unless you're, like, developing the next generation of quantum computers). But now I think I should embrace it more. Because science isn't about having all the answers. It's about following a rigorous process in an attempt to find those answers.

Collapse
entomy profile image
Patrick Kelly

Exactly. I had a hypothesis on how I could design a good parsing framework that solved the problems I laid out. I set up falsifiable and reproducible experiments (both benchmarks and UX tests) with a null hypothesis. The times I was right, I went forward with that approach. The times I was wrong, I rewrote that component with a new hypothesis.

Ironically, science tends to be more about finding all the wrong answers. We know a lot more about how things don't work than do. Less glamorous way to put it though.

Collapse
190245 profile image
Dave

The obvious caveat here, is anyone reading this that writes software for life support machines, or anyone on the Boeing team.

For those people, software bugs are somewhat more important than the rest of us.

Collapse
entomy profile image
Patrick Kelly

Absolutely. My original programming language was Ada, so you can guess the kind of stuff I was doing. When there's a physical product that's being controlled by the software, the whole process is drastically different. You know exactly what you need to do.

Collapse
190245 profile image
Dave

I have to admit... nail hit on the head.

A previous manager used to refer to me as "SpecOps Dev" and when I queried him on it, his response was "I can drop you in on a Production bug, no background, no tools, no prior knowledge of the system you're looking at, no-one to help you, and you always get the job done while giving someone else the credit. You have the strange ability to be able to just do absolutely anything I throw at you. Don't worry, it hasn't gone unseen."

I've also taken a complete novice (could barely use MS Office!) and within 6 weeks, he had written a complete application to have a RaspberryPi actuate a series of physical switches. I started him off by teaching how to think about the problem in UML.

It was that experience with the novice that changed what I tell people at social gatherings in response to "So, what do you do?"... my response now is something along the lines of "We know more about the universe than we do the seas... I basically head off into uncharted water with all the care & attention that airline pilots have... only I do it for a software company."

I'm spread across being a Lead Dev / Architect / Change Manager / PM / whatever other hat I need to wear, and today, I'm covering our Helpdesk Manager's day off while writing code... there's no way you could even start to explain all that to someone that doesn't already live & breathe commercial software development. So I stopped trying.

Collapse
bytebodger profile image
Adam Nathaniel Davis Author

Thank you - and well said!

Collapse
jonrandy profile image
Jon Randy

👍 Man, you should write a book

Collapse
bytebodger profile image
Adam Nathaniel Davis Author

Yeah... it'd be shorter. 😁

Collapse
simmol profile image
Pavlin Angelov

Great article!

I totally agree that developers should not be isolated from the decision process.

For me, that is the biggest takeaway from the whole Agile movement.
Is not just about splitting your work into sprints and have a daily standup is about having access to ask questions, clarify specs, or throw them out the window if they no longer work.

That said there are certain types of programming jobs that work more like a factory assembly line... and that is fine. To use your metaphor about Medical care, not all programmers should be brilliant neurosurgeons, we also need nurses and general doctors and whatnot.

The funny thing about the construction metaphor is a lot of people use it because they think that everyone understands construction, but actually, most people don't.
Most people have no idea how much effort is involved in building a house, what are the steps you need to take, and how much is going to cost them.

Same as in Software there is not a single big and complex construction project that does not stumble on some unforeseen problem or complication that stretch its budget and deadline.

Thank you for the great article, it made me think about a lot of things.

Collapse
bytebodger profile image
Adam Nathaniel Davis Author

Great feedback! And yes, it's absolutely true that most people don't really understand construction (myself included). If they did, they'd probably stop using that analogy.

I do believe that the process of designing software can be very similar to being an architect. But in the dev-is-construction metaphor we usually fall back on the idea that the actual construction phase is static - as though all the planning's already been done, and now those guys with the big machines and the power tools just need to "make it happen".

Collapse
dvddpl profile image
Collapse
cwraytech profile image
Christopher Wray

Hey Adam! Wow, just last night I was sharing with my bro that what I did was like building a house, but yeah, I think you are absolutely right! Thanks for another incredible article.

Collapse
ibrahimshamma99 profile image
Collapse
nebulabyte profile image
NebulaByte

Golden words man. Hats off