The minimum dataset for a predictive lead scoring model is 40 qualified leads and 40 disqualified leads. Microsoft's Dynamics 365 Sales documentation states that an organization must have created and closed at least 40 qualified and 40 disqualified leads within the selected training period to train the model. The broader lead and opportunity scoring documentation confirms the same threshold and distinguishes lead scoring from opportunity scoring: lead scoring requires qualified and disqualified leads, while opportunity scoring requires won and lost opportunities.

This distinction is critical. Predictive lead scoring predicts the likelihood that a lead will qualify or convert based on historical lead outcomes, so the model needs examples of both successful and unsuccessful lead qualification results. Won and lost opportunities are not the correct training records for lead scoring; they apply to predictive opportunity scoring. The question also states that the company has historical sales records for three years, but the model requirement is based on the minimum qualifying lead dataset, not merely the total age of available data. Therefore, the required minimum is 40 qualified leads and 40 disqualified leads.

References/topics: Predictive lead scoring; lead scoring model prerequisites; qualified leads; disqualified leads; AI-driven lead prioritization.