The main difference between a predictive model and a persona model is the actual meaning of the score.
A predictive model predicts the propensity for a certain action to happen. That can be lead to SQL, account to closed-won or any other action in the funnel. We feed the model with positive and negative examples relative to the use case, and the 0-100 score represents the propensity of that action to happen. The higher the score, the more likely this action will indeed occur.
A persona model measures the resemblance/fit to a defined persona (for example, demand gen marketer).
We feed the persona model with 3 inputs:
- Top-down approach (the customer definition of what they see as a demand gen marketer)
- Bottom-up data (examples of people who bought the product)
- Leadspace analyst research.
All three inputs are used to create a robust network of semantic keyword connections that capture the essence of a demand generation marketer.
The output is a score as well, which means 'how closely does this lead fit the definition of a demand gen marketer'. The higher the score, the more confident we are that this person matches the definition of the persona.
So bottom line, predictive scores provide a likelihood for an action to happen, while persona scores provide a numeric answer to the question 'does this person resemble my definition of persona X?'.
Both can be customized, predictive models by use case and which action we are predicting, and persona models by the different characteristics of each persona.