TL;DR: A big challenge for data-driven personas is that their users rarely participate in their creation. This evokes the question of how to develop techniques that enable participatory design of data-driven personas.
There is a line of thought within persona research that argues something like this: personas are less useful than their creation process.
The intuition is that the persona creation process teaches about customers, which is always valuable, but the resulting personas (especially if static) might be less valuable. Especially when building the personas together, while basing these activities on actual customer data (qualitative & quantitative), team members engage in a participatory process that helps internalize mental models about different customer types (i.e., the personas). Discussing with others about these mental models can created shared forms of communication about the customers and associated product decisions.
An analogy is business plan —- the plan itself might not be that useful, but the exercise of planning is because it forces you to consider different aspects of the business you’re trying to build (customers, pricing, distribution, and so on).
If this is true, and the creation process is more important than the outcome, data-driven personas are at a natural disadvantage. This is because they are usually created in isolation, independent of the eventual persona users. For example, a survey person creates a questionnaire to collect feedback from customers. A data scientist then analyzes the data and does some “factor analysis” or “principal component analysis” — processes that are murky, non-transparent, and hard to comprehend. But, through such process, personas then emerge. The problem is that such personas might be missing the mental linkage to stakeholders, unless stakeholders were a part of their creation.
Hence, the vital question is: How to encourage persona users to participate (and how to create techniques for their participation) in a data-driven persona creation process?