Post by rahim on Jan 31, 2024 9:59:47 GMT
Data and other customer data from the same people are available and the personas can be described using the transaction data / CRM data. The problem with survey participants in external sources is that they are anonymous and any overlaps with their own customers are unknown. In addition, the amount of overlap in features present in both sources is small. This makes reverse mapping significantly more challenging and possibly inaccurate or even impossible. In this initial situation.
an intermediate step must be taken by developing DB to Data function using the information available in both data sources (e.g. regional, socio-demographic and behavioral characteristics). Methods – scoring / differentiating rules / artificial intelligence (AI)On the reverse mapping approach: Application of scoring models/rules for personasUsing various (machine learning) algorithms, an attempt is made to reliably differentiate between the personas found or to describe them in comparison to the customer base. For this purpose, the 'classic' methods such as decision trees, regression analyzes and 'newer' algorithms such as self-learning neural networks are used. (Re)mapping.
after internal differentiation of personas via decision tree:Comparison of personas customers with each otherIn the best case, you get n sets of rules only based on the enriched characteristics that reliably differentiate the n personas (with a certainty of > 70% each).These sets of rules are then applied to the customer base; This results in a persona assignment and an assignment probability for each customer The challenge is the quality of the differentiating set of rules: if it is too low, this approach is ruled out(Re)mapping using regression.
an intermediate step must be taken by developing DB to Data function using the information available in both data sources (e.g. regional, socio-demographic and behavioral characteristics). Methods – scoring / differentiating rules / artificial intelligence (AI)On the reverse mapping approach: Application of scoring models/rules for personasUsing various (machine learning) algorithms, an attempt is made to reliably differentiate between the personas found or to describe them in comparison to the customer base. For this purpose, the 'classic' methods such as decision trees, regression analyzes and 'newer' algorithms such as self-learning neural networks are used. (Re)mapping.
after internal differentiation of personas via decision tree:Comparison of personas customers with each otherIn the best case, you get n sets of rules only based on the enriched characteristics that reliably differentiate the n personas (with a certainty of > 70% each).These sets of rules are then applied to the customer base; This results in a persona assignment and an assignment probability for each customer The challenge is the quality of the differentiating set of rules: if it is too low, this approach is ruled out(Re)mapping using regression.