We have developers who have cleared the technical round of interviewa and the technical interviewers mentioned that the developers appear to have good communication skills.
Since bringing good people into a project is difficult, what information can I use to predict positive future performance when evaluating candidates for developer positions where good communication is key?
If good communications is key I think the best that you can do is to have a variety of interviewers. If you base your analysis only on the technical interviewers you will bias the assessment towards someone who is good at communicating technical topics with other experts but may not be able to translate that to good communication with lay people.
Assuming you go with a variety of interviewers, it is important to have what they are evaluating overlap. For example, a developer and a business analyst could both assess the candidate's understanding and communication of functional and non-functional requirements. If you silo what individuals are evaluating you will only get a disjointed picture of the candidate.
There are no uniformly valid predictive indicators for successful hires. However, there are certainly valid interview practices for assessing communications skills and potential project fit.
First, communication skills are often self-evident during an interview—assuming the interviewer has sufficient communications skills to evaluate them in others. Some ways to validate candidate communications skills include:
None of these things will guarantee that the person will communicate effectively if hired, and people with disabilities may perform poorly in some of these areas despite otherwise good communications skills. Still, if effective communication is a basic job requirement, the only way to test for communications skills is to actually attempt to communicate.
From a project management perspective, your main concern in the interview process ought to be whether the candidate is a potential fit for your project management framework and project culture. It is generally someone else's responsibility to vet the person for the technical skills necessary to complete deliverables; even if you happen to be a technical subject matter expert, your role is managing the project, not vetting someone's source code, so focus on how that person is likely to fit within your core areas of responsibility: project framework practices and project controls.
Investment companies will be the first to tell you that past performance is not a guarantee of future results, and most interviews are really just hearsay anyway. Nevertheless, there is a lot of value in inquiring after a candidate's experience with your particular project management framework (especially when using agile frameworks) and then listening to how the candidate answers.
If you're a Scrum Master, there may be value in hiring a candidate who already has experience with a relevant Scrum role. A candidate who understands (and can articulate!) the purpose of a standup meeting or how to play Planning Poker is probably a more likely to be an effective self-organizing team member than one who is used to needing written authorization to blow his nose.
If you're a waterfall practitioner, there's probably more value in hiring someone who understands the planning and reporting requirements you need than someone who prefers to "get 'er done" without all that tedious project management stuff.
That doesn't mean you can't hire someone who isn't a project management expert, or who doesn't have experience with your particular framework. However, all other things being equal, an understanding of the team's project management framework and practices (and, perhaps more importantly, a shared appreciation of those practices) is going to be more valuable to you as the project manager than anything else you could ask.
There aren't any guarantees in hiring. Even if someone is a perfect fit, they may get a better offer, learn to detest their teammates, suffer a health crisis, or decide to enter a monastery as a career change. Life is unpredictable.
All you can really do is identify what's important to you in your specific organizational role, and then do your best to select for the characteristics that make your job at least possible. If you get it wrong (and it's even odds that you will) then do your best to honestly assess whether you selected for the wrong things or whether it was "just one of those things." If the former, fix your selection process; if the latter, increase your statistical sample until you achieve the results you want.
Employee selection is a huge topic and very complex. There are a good bit of studies out there analyzing the validity of performance predictors and future job success. I read previously, but I cannot find the source, that some believe we have a 50% chance of selecting correctly. If this is true, this means you can dispense with the expensive interview process and simply flip a coin. I wouldn't recommend that not because of potentially disastrous results but because it would be politically unacceptable.
One of the first things you should do on this topic is trash what you think is true on this topic. Education, where someone went to school, training certificates, PMP, experience, results of interviews, all of these things rank quite low as valid predictors. Hunter & Hunter in 1984 found that these things have less than 0.20 validity. They also found that cognitive tests and job tryouts are the best indicators, with a validity of 0.53 and 0.44 respectively.
The penalty with these findings is, cognitive testing and job tryouts are very expensive. The tests and job simulation have to have validity itself, which means they need to be developed and tested over time. Companies that do this charge for its use.
But to answer you question about your specific scenario, I would suggest first that the results from the first set of technical interviews are, at best, questionable, 50/50 they got it right, and if you have a good size population of candidates to begin with. If you can't go back and do something else, then the best thing you can look for is, 'are they smart and can they get things done' (Joel Spolsky). And leave your biases at the door.
