Pros:

• Can be confident that outliers are not overrepresented
• Know what percentage of your total user base each persona represents
• Can reverse-engineer the persona clustering (using discriminant analysis) to figure out which survey questions most strongly predict which persona someone was clustered into, which is useful for recruiting in future studies and ensuring representation

Cons:

• Expensive, time-consuming, and requires expertise in statistical analysis
• You could end up doing all the work of both qualitative research and statistical analysis only to get the same results as you would from just creating qualitative personas
• It’s massive overkill in most situations and can leave a mess if not done carefully