Deep dive into pair programming

Before I switched careers into software development, I didn’t have a clue of what a complex system looked like, how they are created and the effort spent building them. I had a misconception that programmers spent a lot of time writing code independently. When I started learning how to code and speaking to developers about their experiences, I was quite astonished at the collaborative culture within software development. In my job, I continually interact with other developers, the product owner, the UX/UI designers and sometimes the stakeholders of the product we are working to build. I am encouraged to not only share my thoughts on technical aspects, but on UX/UI prototypes and on the value of the features we are building for our users. My role as a software developer is much more than just coding. In this blog however, I want to focus on how developers collaborate through pair programming.

Pair programming

Pair programming is where two developers work together to solve a problem. Whilst there are advantages to pairing most of the time, I think there are some occasions when it is better to work individually.

When it is suitable to pair? Well, it depends on what the specific task entails. Is it a laborious task of deleting dead code or is it a new search feature? There are some tasks that are just not worth the time of two developers. One of the developers is likely to sit and stare at the screen watching the other developer continue with the task. I put together a non-exhaustive list of questions that I would ask myself to determine if its worth pairing or not:

Types of pair programming:

1. Driver-navigator

This is a classic pair programming style. One developer is the person on the wheel (i.e. keyboard) and their role is to type code. The other acts as the “observer” and their role is to review the code on-the-go and navigate the driver on what to code. This exercise should involve the navigator explaining how they plan to solve a problem and prompt the driver to codify that idea. The navigator should avoid dictating the code word for word to the driver. I think this pairing practice is useful when a more experienced developer is coaching another developer on how to code.

In my opinion, 20 minutes or less is a reasonable amount of time to write some useful code. These short sessions encourage developers to make small, iterative changes because you would ideally want to integrate that code often. Switching between the two roles is important because the perspective on the code is different. The drivers job is to think about the nitty gritty of the code at hand. The navigator can think more strategically about the bigger picture such as potential obstacles that might arise.

The key to pairing is to switch roles frequently.

In practice, these guidelines are not always strictly followed. There have been a number of times in my own experience where a developer and I have not switched roles for an hour or even more sometimes. Those sessions drained us and rendered us unmotivated. Ultimately, since the work still needs to be done, the solution is less thought through and important steps such as refactoring could be missed out of laziness. Switching roles increases the participation from each developer and thus the energy levels.

2. Ping pong

Ping pong pair programming is a fun way to program and it revolves around writing tests. In this style, developer X writes a test and developer Y writes the code to pass that test. Developer Y then writes the next test and developer X has to write the code to pass the test. This is a great technique because you are following Test Driven Development (TDD). TDD is a whole other topic in itself but, writing tests that reflect the desired behaviour is a good tool to design good, modular software. Read this post written by my colleague if you are interested in a convincing article about TDD.

This is my favourite pairing style. A major plus point is that the developer writing the test is focusing on writing a useful test without the solution in mind because the solution will be implemented by their pairing partner. This leads to clear, understandable tests and therefore a good chance of well-designed production code. I like how the code can evolve in an unexpected way when bouncing off each other’s code and tests.

The well-known TDD steps should be followed:

One of the important steps to remember is the refactoring stage because it is your moment to design well! You have reliable tests and working software so refactoring should only improve the quality of your code (if done correctly of course). Can you spot a pattern in your code? Can you extract some code to make it more readable? Are you injecting dependencies in a function or a class? Are you following SOLID principles?

3. Diverge-converge

This is not a very well-known technique but it is something I picked up whilst I was at a coding bootcamp. This technique involves two developers diverging out on their own to figure out how they would solve a problem and then converging back together to discuss and decide what they will implement. I think this technique useful when there is not a concrete plan on how to make something work. This method is suitable if the task requires a lot of research or reading through documentation which is more efficiently done solo rather than in a pair. The developers could perform a spike to see if their solution actually works (a spike is experimenting/exploring with some code). However, once you figure out something viable, it is important to bin all that code and start again using TDD.

Before and after pairing

There are different pair programming methods to code together. But when picking up a task, there are other important ceremonies before and after coding.


We talked about how developers can work together but I think it’s interesting to note why working in silos can damage overall output.

Benefits of pair programming

Overall, I think pair programming can be quite enjoyable if done correctly. Here are my top tips to pairing:

  1. Don’t hog the keyboard. Switch roles often and work in small steps. Maybe try to use the Pomodoro technique?
  2. Take breaks! Thinking, talking and typing can get exhausting so it’s refreshing to take frequent short breaks.
  3. Are you in a “pairing mood”? If not, be honest about it and maybe work on something that can be done by yourself for a while.
  4. Rotate the pairs. Anchor a story to one developer but change their pair partner everyday. This spreads knowledge and forces the developer anchored to the story to reevaluate and reinforce their decisions by explaining them again.

Sources: Pair programming illustration by Storyset