In reply to Moley:
Make sure you/the panel ask the same set of questions to each candidate (e.g. a series of competency questions relevant to the role, tell about a time when you did x, explain how you would solve problem y) then score each one 1 (poor) 5 (exceptional).
This enables you to achieve predictive validity in terms of job performance (did any of the candidates meet the standard/score) and also comparative validity (which one was best)
Unstructured interviews where you ask different/random questions to people have been show to have the same predictive validity as analysing their hand writting style (i.e. low!)
Clearly the intro and end can be a bit more free flow but still aim for similar questions (tell me what you know about the organisation, outline why you want the role, what are your strengths/weaknesses)
If you all score using the same 1 - 5 methodology and agree in advance what it takes to get a specific score you should be able to easily differentiate between candidates.
If you then have a "tie" you can hold a 2nd interview/work trial to work out "fit" which is harder to assess in formal interview situations.
Post edited at 17:16