> I wonder if there are studies on the effectiveness of pair programming vs code review and bugs caught.
Whenever folks ask me this, I want to turn it around: are there studies on the effectiveness of working solo vs pair programming?
> Even when I have pair, I like to do an alone-time code review and things almost always pop up.
Agree - even with code that's been mostly paired on it can be helpful to let it sit and get additional review.
In my experience, the killer combo for pairing is to combine it with test-first programming. The tests help create a shared construct for discussion and understanding.
I freely admit I have a very strong pro-pairing and pro-test-first bias. But in my defense it's based on almost two decades of experience.
>Whenever folks ask me this, I want to turn it around: are there studies on the effectiveness of working solo vs pair programming?
Pair programming inherently halves your potential productivity, as you have 2 people doing 1 work.
Additionally, you introduce additional friction because of communication between the two programmers, and finally, you still (usually) need a review of the final code, even if we assume it's better then in 'solo' scenario.
So for the pair programming to be objectively effective, it would need to provide over 100% of productivity boost minimum, because thats how much it takes away.
That's a huge ask, and it's on advocates of it to prove that.
Normal process in comparison, because of having inherently twice the working power, can waste up to 50% of person's productivity and still be ahead of pair programming starting position.
The 'math' is simplified ofcourse, but the burden of proof is on pro-pair people by default, not the other way around.
Whenever folks ask me this, I want to turn it around: are there studies on the effectiveness of working solo vs pair programming?
> Even when I have pair, I like to do an alone-time code review and things almost always pop up.
Agree - even with code that's been mostly paired on it can be helpful to let it sit and get additional review.
In my experience, the killer combo for pairing is to combine it with test-first programming. The tests help create a shared construct for discussion and understanding.
I freely admit I have a very strong pro-pairing and pro-test-first bias. But in my defense it's based on almost two decades of experience.