For proto-persona, it’s weak because it’s fictional. To overcome the weaknesses, try to harness their validity by as much data as your can resource can afford to gather.
For persona that is built from only a small numbers of real people’s stories, it lies the risk of not being representative enough. Like above, increase the sample size to enhance the generalization of the persona.
For persona that is built from a huge piles of datas, it merges stories of many people who share similar characteristics. This has the risk of conflicted characteristics and goals. Combine too many and you may miss to serve a potential niches, combine too little then you will end up having too many different persona to serve and waste business resources.
Solution is real life test. If after a while, you can pinpoint that there’s a distinct niche which is big and profitable enough to serve, split the existing.
