The most common flaw I see in the first three minutes of an interview is a candidate's first impulse being to write code. Jumping straight to implementation is almost always a negative signal; just like in real life, you want to spend time sussing out the domain and making sure you understand all the ramifications of the prompt.
A really easy tactic for this: restate the problem to your interviewer and ask them to confirm your understanding! This takes, like, thirty seconds, demonstrates that the two of you are on the same page, and gives you confidence that you understand the problem being asked.
If your interviewer is nice, they'll give you a couple example invocations of the method or class you're implementing, like such:
most_common_word("dog dog cat") == "dog"
most_common_word("The lazy cat jumped over the lazier cat.") == "cat"
If they do this, add more example inputs and outputs. Pay attention to edge cases, like:
If they don't do this: especially add inputs and outputs. Again, it shows that you're thinking about the problem systematically, that you've considered scenarios besides the most obvious, and it gives you a great framework to approach the solution. 1
As an interviewer, my job is to collect as much positive signal about a candidate as I can. 2 Specifically, this means I'm looking for things that you do well:
You know the one time it's impossible to get anything along these lines? When you're silent, thinking very hard about a problem or an issue. I understand the temptation to just kind of sit back and think, but it's awful for me in terms of getting actionable feedback. Try and talk through what you're thinking or write down your thoughts on a notepad/whiteboard.
Seriously. Everyone loves comments. It might feel off-brand to write a bunch of comments when you're on the clock and in a tunnel vision of "oh god gotta solve this quick", but it is worth your time. Some genres of comment that I always love to see from candidates:
// TODO: handle this edge case
// This is inefficient, but I'm going with this approach because <reason>
// Abstract this out later.
All three of these do a great job of conveying I understand this isn't ideal, and know that given more time I can improve this, but in the interest of expediency I'm ignoring it for now.
I don't mean this glibly! I have in my entire life not met a single engineer who likes the way a conventional tech interview works. I think my current employer does a lot of things very well (no whiteboard coding, no algorithmic shibboleths, no manhole-cover questions) and I still think there are tremendous flaws.
Being good at technical interviews is a different skill than being good at software development. This is shitty, but it's true. So many folks I've talked to have said something along the lines of I'm a great programmer, I just don't interview well and that's because those are activities whose Venn diagrams of competence overlap much less than folks assume.
The antidote to this is practice. Go through Cracking the Coding Interview; whiteboard some architectures with friends or fellow classmates; it gets more comfortable with time and experience.
Email me. I'd be more than happy to spend an hour interviewing you!