Since going full-time on my own startup 6 months ago, I’ve spent a lot of my time recruiting, evaluating, and working with a lot of different developers. My startup, MarkedUp, is an analytics provider for Windows 8 apps (and eventually, other platforms.)
Our technology is sophisticated and difficult even for large companies to master, so the bar is pretty high for developers who join our team. We’re selective about who we work with, but even with careful recruiting practices it’s still difficult to find developers talented enough to pull their own weight on a complex product like ours.
So I want to help average developers who want to improve by calling out 10 performance-killing behaviors that stagnate the careers of most developers.
Most developers stagnate both intellectually and productively after 4-5 years in industry; they adopt some tools, pick up some patterns, learn a language or two, and maybe they’re even able to work at a successful company and contribute to some important products. Great!
But what happens when you hand a developer a blank sheet of paper and the opportunity for them to own a product?
Most of the time: chaos and failure.
Many developers fantasize about being able to own a product and make big technical decisions, but the truth of the matter is most of developers can’t handle anything more complicated than building formulaic CRUD Rails websites using the same 30 gems each and every time.
So what are the behaviors that limit the growth of average developers?
You’re too afraid of failure to learn new tools / languages / frameworks.
During my first quarter at Microsoft I gave a talk to the LA .NET User’s Group on ASP.NET MVC3, which was in beta at the time. For the most part the attendees asked me thoughtful questions and showed a legitimate interest in what the ASP.NET team was cooking up in Redmond.
However, there was a trio of attendees who kept asking me a very suspicious set of questions:
“What are the weaknesses of ASP.NET MVC3?”
“Isn’t ASP.NET MVC too unstable for use in the enterprise?”
“But doesn’t ASP.NET WebForms already do a better job?”
I realized shortly after the Q&A was finished, when I had a chance to speak to them privately, that they were attending my talk to actively look for reasons why ASP.NET MVC shouldn’t replace WebForms at their place of work!
These developers were so afraid of trying something new and possibly failing that they actually went out of their way to research just the talking points they needed to not have to learn it!
This is an extreme example, but the point is the same: average developers cling to their favorite tools and stop learning new ones because they worry that leaving their pet languages and tools behind invalidates their years of experience.
Strong developers embrace the challenge and opportunity to learn new ways of doing things.
You won’t push your commits until the feature is “done.” (It’s never done!)
We call this “commit hoarding” at MarkedUp. This happens with developers who don’t have the confidence to tolerate criticism or scrutiny from the rest of the development team try to hide their work until it’s in a “finished” state.
Ultimately these developers hurt the productivity of the rest of the team, because we lose visibility into their daily output and they tend not to ask for help in the course of regular development. They’re responsible both for delays in shipping and for lots of last-minute bugs.
Strong developers recognize that they are not their code, and thus will put code out in front of the rest of the team more frequently and take criticism and suggestions with dignity and confidence.
You know just enough to be dangerous.
In an age where it’s become a lot easier for developers to work with threads, concurrent programming, distributed databases, and so forth you run into a lot of devs who can talk a good game. They can tell you all about thread pools, a hot new design pattern, or why a headless database is better than a masterful one in some scenarios.
They are able to know what something is and its potential benefits, but they don’t understand it.
I’ll give you a recent example: in C# 5.0 Microsoft recently introduced the async and await keywords, which make it radically easier to create and manage asynchronous callbacks. These keywords have even lead to the rise of some exciting new concurrent programming patterns like the TAP pattern.
However, do you know what this keyword has also lead to? A ton of average developers who suddenly think they can harness the awesome power of multithreading simply by wrapping every I/O call into a Task object in C#. They do this blindly, without any thought as to the costs of context-switching, multi-thread access to shared resources, and more.
And this creates dangerous, unpredictable, and very-difficult-to-test code in the hands of average developers who blindly uses this new toy everywhere.
Good developers not only know what something does, but they seek to understand why it was done this way and under what conditions it should be used.
Sometimes you’ll hear a member of our development team say “dude, stop Gandalfing the documentation and start coding something,” referring to the scene from The Lord of the Rings depicted above.
Analysis paralysis is a classic problem that affects lots of average developers; it’s often a problem of over-analysis, but in my experience the root cause is usually fear of making a bad decision.
Research is easy – implementations are hard! As a developer, you don’t want to make a mistake when it comes time to do the actual work. So just research the issue until the heat death of the universe and you’ll make a perfect decision for which you won’t ever be criticized!
Average developers worry about looking bad or making a mistake – they want to get it right on the first try.
Strong developers who are in unfamiliar territory don’t care – they’ll write shit code, unit test it, and throw it away in 45 minutes if they don’t think it can get the job done. Strong developers actively limit the amount of time they spend researching, because they recognize that it’s a trap – it feels like work but often isn’t.
You don’t invest in your tools or development process.
Average developers will spend a lot of time learning how some fancy new language feature or API works. They’ll invest a ton of time mastering the minutia of the Windows runtime or some new D3-based charting framework, but at the end of the day that doesn’t change the fact that they’re still average developers who churn out mediocre work at a glacial pace.
If you want to be a talented developer, you have to invest time into improving your skills and knowledge, and the number one factor that separates the men from the boys in most software markets is the ability to turn out production-grade code quickly. You can have both good code and speed – but you have to invest in your process you use for building it first!
Average developers don’t invest in their tools, their process, or their environment.
Think about it – when was the last time you took a step back and thought about the way you test your code or how you look for bugs? Have you tried experimenting with different types of abstractions for the same problems and seen which one produces more maintainable code? Have you tried picking up a new IDE or text editor to see if it can help speed up common tasks? What about investing in build or deployment automation?
Often times, the biggest improvement you can get as a developer isn’t focusing on improving the code you write – it’s optimizing your process for writing it.
Free protip for anyone who follows my blog regularly: the biggest improvement I’ve made in my own development process easily is decreasing my usage of interface-driven design and dependency injection. I do a lot more work using static classes and compositions these days and it results in radically less code to test and maintain.
You’re too embarrassed to ask for help.
When we started digging into the guts of Hive for some of our work at MarkedUp, we got a little stuck on how to tackle some problems we had specific to our use case. So what did we do? We reached out to one of the guys who wrote the O’Reilly book on Hive and asked him over LinkedIn. And he replied back!
We try to be self-sufficient developers to the furthest extent possible, but we also know when it’s more economical to simply ask someone for a favor versus beating our heads against our desks for hours on end.
Average developers get embarrassed and don’t want to reveal their ignorance, so they pretend they know what’s up until they commit some sort of horrible fuck-up in the codebase itself.
There is nothing wrong with saying “I don’t know how to do this.” Strong developers know this, so they go and ask for help when they’re stuck.
You don’t know how to make it easy for other programmers to work with your code.
Like an only child growing up, you never learned how to share your toys with anyone else. The same goes for your code – you developed some bad habits and an over-abstracted style that is impossible for other developers to follow.
An important part of working on any technical team is human parallelism – the ability for multiple people to be able to work simultaneously on the same codebase. But it’s also important for teams to be able to work asynchronously too – I should be able to make changes to your code when you’re away, and vice versa.
Average developers don’t think about their code this way – they simply set out to do a task and write it through to completion with the assumption being that they will always own that piece of code. Strong developers understand technical debt and try to limit it by designing code that is as maintainable and self-explainable as possible.
Writing readable code really requires developers to change their outlook – your code should last longer in the organization that you.
You don’t know how to read someone else’s code (or don’t want to.)
It’s an interesting paradox – average developers can’t handle a blank sheet of paper very well, but they can’t write the last 30 pages of a novel in progress either. Average developers are really hesitant to try to learn an existing codebase and often have to be handled all the way through it until they’re able to produce something.
A drastic example is when an average developer is brought in on a codebase written in a language / framework that they aren’t familiar with and will immediately want to rewrite it without a second thought to business value or time to market. It’s a type of pain avoidance – having to live with the precipitation of some other programmer.
Strong developers accept that the business costs in a rewrite are usually unacceptable and should be avoided, and they’ll diligently try to understand, learn, and modify the existing codebase sitting in front of them.
Reading code is harder than writing code, but strong developers invest time into learning how to excel at it.
You can’t code from the end-user’s point of view (you think too small.)
Average developers who manage to make it a lot closer to the “great programmer” end of the bell curve usually trip up here. You may have mastered some of the technical details and you may know how to work with other developers, but you’re still missing the most important piece: being able to see your code from the point of view of the real-world.
As a programmer, you job isn’t really to solve technical problems – you solve technical problems in order to solve business problems.
An average-to-bad developer will chase technical problems down deep rabbit holes without remembering why they were solving the problem in the first place. You’re not upgrading the version of MongoDB you’re using because there’s inherent value in upgrading – you’re upgrading because your team agreed that it was worth taking the risk to upgrade because of potential business upside due to new features / performance / reliability / whatever.
More offensively, average developers struggle and fail to produce anything of business value from scratch. When asked to design a new feature based off of a simple customer story average developers will take a rigid, literal interpretation of the story or spec and ship something that is barely useable by a human.
Average developers don’t anticipate other related use cases; don’t think about the end-user’s experience; and basically have to be managed heavily when working on anything user-facing. This is why so many of them get stuck writing line of business apps instead of products.
Great developers look at their code from the perspective of the end-user. How do I make this easier to solve my user’s problem? What are some other facets outside the literal content of the story that might make this feature even better for an end user?
If you’re a developer and don’t ask yourself these questions, then you’re not a great developer. You’re passable at best.
You aren’t able to judge the business value of any programming task.
Related to the previous issue – many technically strong developers fail to realize their potential due because they aren’t able to take a step back and look at their work from the perspective of the business or organization itself.
Strong developers are able to self-manage and make good business decisions about how they choose to invest their time. They’ll ask questions like:
Is this the most valuable thing I should be working on right now? How much time should I invest into doing this? Given the deliverables I have due in two weeks, is there something I can invest in right now that will make it a lot easier for me to hit that target?
Average developers don’t – they’ll take a spec and blindly implement it until it’s finished, without accounting for how their work relates to the company’s business goals and impacts other members of the team / other business groups.
The highest cost of any engineering team is development time, and average developers waste lots of it on minutia and technical tasks that are ultimately low business value.
If you want to be a better developer, it starts with changing the way you look at your code and how you program.
You have to understand and appreciate the business costs behind each line of code you write. You have to be able to step outside of your text editor or IDE and look at your work from the perspective of a customer or end-user.
You have to accept that your code will outlast you in any organization your work with, so design everything to be inherited by other developers one day.
Most importantly, never be afraid to take on new challenges and ask for help along the way. You won’t improve working alone in the corner on technologies you’re already familiar with. Software development is inherently social – learn how to bring other people into the experience.
If you enjoyed this post, make sure you subscribe to my RSS feed!
Last week our startup, MarkedUp, hit the first important milestone for an early stage technology: we shipped the first version of our analytics product and put it into the hands of actual end-users.
For those of you who don’t read my blog regularly (pretty much everybody, I suspect,) MarkedUp provides in-app analytics for desktop application developers. We’re focusing our company’s initial efforts on supporting Windows 8 and WinRT developers specifically.
Now, we still haven’t completely opened the doors to everybody – we’re capping registration for MarkedUp through the use of invite codes while we gather important UX feedback and reports on bugs we may have missed.
Nonetheless, we hit that important milestone and have live users! Yay!
However, here are some interesting facts about the timeline leading up to our ship date:
- We were feature complete, with high test coverage and few known issues, a full two weeks before we were able to ship;
- We didn’t have a design for the marketing parts of our site, like the homepage, until 48 hours before our launch;
- We actually cut features from the product the night before our launch;
- We started the process of gathering user feedback when we were about 50% finished with the first version of the product – by then we had demoable charts and tables (features of our analytics product) we could show to target users.
We weren’t able to ship for two weeks, two weeks (!) after we were feature complete – and we cut features the night before we shipped? Are we a bunch of crazy incompetent bastards? No!
Shipping Software: More Than Just Code
This wasn’t my first time participating in the ship process for a piece of software, but this was my first time running it.
I suspect my opinion on this will change the more often I do it. In my estimation, the time it takes to design, plan, and code the core product and its features constitutes only 40% of the actual work required to deliver a software product to market.
So, what were the things we spent our time doing leading up to launch?
Great question, I’ll break it down by category:
- Feature development and functional testing – 40% of time;
- Pre-launch UX testing and refinement – 20% of time;
- Creating a demo application – 5% of time;
- “Marketing engineering” – time spent coding features responsible for driving users towards conversion goals – 10% of time;
- Graphical design / branding (largely outsourced) – 5% of time;
- Writing documentation, tutorials, and setting up support infrastructure – 10% of time;
- Writing and proof-reading marketing copy – 5% of time;
- Planning and executing roll-out marketing – 2.5% of time;
- Planning and executing roll-out software release – 2.5% of time.
One thing to bear in mind about these numbers – this was for a soft beta launch, not a top-down, TechCrunch-style launch. The amount of time spent on things like roll-out marketing is astronomically higher for a massive top-down launch.
What’s the trend here? Most of the the time we spent working on the product was spent on usability issues.
Like Jeff Atwood says, your app is merely a collection of details. There’s really two classes of details that matter:
- Details that affect how the user engages with your product and
- Details that affect how the user understands your product.
Marketing engineering, demo apps, copy writing, and graphical design are all investments we made into helping make our product easier to understand.
Documentation, support, and UX are all investments we made into helping make our product easier to use.
The bottom line here is that the auxiliary stuff needed to support your product, like documentation and support infrastructure, take a lot of time to do well. They’re as much a part of your user’s experience as the software is itself.
Shipping the Right Features at the Right Time
So we cut features right before we shipped, literally 24 hours before I pushed to production – why?
The answer is simple: we have specific goals for what we want to learn about our users with this version of our product. Any feature that we considered to be too unpolished, too complex, or too distracting was cut – even if we had a reasonably good implementation of it ready to deploy.
Every software release should be purposeful and careful – if you overwhelm your users with too many features and options during the first release, your user feedback gets diluted across that feature set and you never get a chance to learn what you are fundamentally doing right or wrong!
If you ship a bunch of poorly implemented features, you might churn through all of your early users way faster than expected and have to subsequently work harder to acquire enough users for the next iteration. If you ship a bunch of “distracting” features, you detract from the feedback that really matters to your business.
Ship software with the intent to learn and test very specific things in that release. Don’t ship for shipping’s sake.
If you enjoyed this post, make sure you subscribe to my RSS feed!