Blueleopard,
I have seen that you record IP addresses and e-mail details to check that users are not adding fake reviews etc., but how many fake or malicious reviews can be left before the system recognises them?
If fake or malicious reviews do get through, how will you account for this either negative or positive feedback in the agents overall score?
do you validate that the property the reviewer has entered has actually been sold or let by the agent?
Would it be feasible to notify the agent that a review is being left and to ask them to validate that they were actually responsible for the sale or let?
How will you stop people posting a review for the wrong agent by mistake, say if more than one agent have very similar names?
How long do the reviews count as part of the agent score for?
An agent could have been really poor 8 months ago, but could have made changes, and now be brilliant (or vice versa) - how would this be reflected?
Just a few thoughts.