If I understand correctly you wonder how many experts should you use. I don't think there are rules regarding the minimum number of experts. One expert gives you one pairwise comparison matrix. If you have several experts you derive one pairwise comparison matrix for each expert and then you can choose the best matrix which has the minimum consistency ratio value. Or maybe you can derive a single matrix using the mean relative importance of factors indicated by your experts.

Thank you for your answer, what if i used the average of the of the weights for each expert?

You can do this as well, provided the experts don't have very different oppinions. For instance, if one expert gives a high weight to a factor, and another a low weight, on an average you get a third situation

This is the case for me, and the average gives a new look that's a consensus

Thank you for your good discussion. There is a question:

How could I convert Likert scale to pairwise and doing AHP ?

Elaborate the likert scale

I don't know if it is feasible to ask an expert to complete (even if indirectly) a matrix with 13 variables. Maybe a conjoint approach will be more reliable. However when you have more variables the optimal number of experts is given by their willingness, time... Beside that there is an article which says that optimum number of variables is 7/(7+2). Taking into account this problem and the characteristics of experts and of their answers is necessary to check step by step recursive bias of weights

hey bogdan, can you refer me to the article i have a look at it

now i found some suggestion : Kardi Teknomo

ANALYTIC HIERARCHY

PROCESS (AHP) TUTORIAL