JoveWhizz's Driver Analysis methodology uses advanced statistical techniques to identify and quantify the factors that most strongly influence customer satisfaction, brand loyalty, purchase intent, advocacy, and other key business outcomes. Our approach transforms complex survey data into clear, actionable prioritisation frameworks.
Driver Analysis establishes statistical relationships between independent variables (feature satisfaction, brand perceptions, service experience ratings) and dependent variables (overall satisfaction, NPS, loyalty, purchase intent). By quantifying each driver's relative contribution to the outcome, we identify the highest-impact areas for improvement investment.
Unlike simple gap analysis that prioritises based on low satisfaction scores alone, Driver Analysis reveals that not all attributes are equally important. Attributes with moderate satisfaction scores may be powerful drivers, while poor-scoring attributes may have minimal impact on the key outcome. This distinction prevents misallocation of improvement resources.
We deploy multiple statistical techniques depending on data characteristics and business questions. Linear regression provides interpretable coefficients for normally distributed outcomes. Ordered logit and probit models are used for ordinal dependent variables like Likert-scale satisfaction. Structural equation modelling (SEM) captures complex relationships between latent constructs and mediating effects.
For importance quantification, we use Shapley value analysis (Shapley Value Decomposition) which fairly allocates explanatory power among correlated predictors, providing more reliable importance estimates than standard regression coefficients when independent variables are intercorrelated — which they almost always are in customer experience data.
Our core output is the Importance-Satisfaction (I-S) Matrix, which plots each attribute on two axes: statistical importance (driver strength) and current satisfaction (performance). This creates four quadrants: Concentrate Here (high importance, low satisfaction — priority for improvement), Keep Up (high importance, high satisfaction — maintain performance), Low Priority (low importance, low satisfaction — deprioritise), and Possible Overkill (low importance, high satisfaction — potential resource reallocation).
We enhance this framework with impact simulation: "What would happen to overall satisfaction if we improved this attribute by X points?" This quantifies the business case for improvement initiatives and supports ROI-based prioritisation of resource allocation across competing improvement opportunities.
Driver Analysis is applied across customer satisfaction tracking, NPS driver analysis, brand equity modelling, employee engagement, customer experience measurement, and service quality improvement. Our methodology supports both point-in-time assessment and longitudinal tracking to monitor how driver importance evolves over time.
We conduct segment-level driver analysis to identify whether drivers differ across customer segments, enabling targeted improvement strategies. Driver Analysis can also be integrated with financial metrics to calculate the revenue impact of satisfaction improvements, building a compelling business case for customer experience investment.
We recommend a minimum of 150-200 respondents for stable driver models with up to 20 independent variables. Larger samples enable more granular segment-level analysis and more complex modelling approaches. The required sample also depends on the number of drivers being evaluated and the expected effect sizes.
There is no strict limit, but models with 10-30 drivers typically provide the best balance of comprehensiveness and interpretability. For studies with very large driver sets, we use dimension reduction techniques or hierarchical modelling to manage complexity while preserving analytical rigour.
Correlated drivers are common in customer experience data. We use Shapley value analysis which fairly distributes explanatory power among correlated variables, and we report variance inflation factors to diagnose multicollinearity. When necessary, we combine highly correlated variables into composite constructs.
Yes. Standard driver models assume linear relationships, but we also test for non-linear effects. Some attributes may have threshold effects (a minimum level must be reached before impacting satisfaction) or ceiling effects (diminishing returns beyond a point). Identifying these shapes improves prioritisation accuracy.
Driver importance can shift due to market changes, competitive actions, customer experience evolution, and changing customer expectations. We recommend periodic driver analysis to track these shifts and adjust improvement priorities accordingly. Longitudinal driver analysis is particularly valuable for mature tracking programmes.
A standard Driver Analysis study takes 3-4 weeks from data collection to final reporting, assuming survey data is available. Analysis typically takes 1-2 weeks depending on model complexity, number of segments, and the depth of simulation and scenario testing required.
Contact us to discuss how our Driver Analysis methodology can help you focus resources on what truly drives customer outcomes.
Contact Us