Title
Probability of Escalation (POE)
Objective
Determine those hand-raising customers who are most likely to go through the Mediation/Arbitration process
Industry
Automotive
Outline
The population of interest was first defined, followed by a data collection and cleansing process. New variables were created and models developed using logistic regression and decision tree models that predict the probability of a hand-raiser (complaint) ending up in the mediation process. The resulting models were highly accurate and identified a handful of key drivers such as severity, longevity, vehicle purchase history, geographic region and number of complaints amongst others. These results will allow the determination of the most appropriate response/settlement to return the customer to a satisfied state and maintain brand loyalty.
Title
Optimize Supply Chain of Parts Distribution
Objective
Determine the optimal configuration of the supply chain for the auto parts distribution in North America including the ports of entry for the parts, multimodal transportation mediums to be employed (rail, truck) and the location and capacity of Distribution Centers to support the dealerships network.
Industry
Automotive
Outline
Optimization and simulation models were developed to minimize the overall cost of the supply chain including transportation, distribution center, and inventory costs to support the parts needed for all the models that were sold in North America.
Title
New Truck Sales Potential
Objective
Develop a methodology to predict which businesses have select truck classes along with their distribution in states where data is not available
Industry
Automotive
Outline
There were three parts to this study the first being the development of logistic regression models to determine which businesses were most likely to have trucks of any type. The second part was the application of clustering algorithms to determine common groupings of truck classes and finally the development of discriminant models to predict which businesses had what combination of truck classes. The study utilized vehicle registration data from states where available and the Dunn & Bradstreet business database. The results were used to determine which business to target for new sales as part of campaign management and messaging.
Title
Market Segmentation for Vehicle X
Objective
Develop behavioral segments for Customer Relationship Management applications
Industry
Automotive
Outline
Principle component analysis, clustering algorithms, and descriptive measures were used to develop a series of primary and secondary market segments based on purchase patterns and demographic information for both specific brands and division line product families. These studies became a key element in the corporate marketing strategy by targeting smaller niche segments to correspond to customer purchase behaviors. The resulting behavioral segmentation for vehicle X increased profitability by an estimated $30 million in one year.
Title
Used to New
Objective
Develop a methodology to predict the likelihood of a current used vehicle owner to purchase a new ABC vehicle over the next year.
Industry
Automotive
Outline
Utilized logistic regression, decision trees, and clustering algorithms to develop a series of predictive models based on vehicle buying history, demographics, and behavioral segments. The modeling process was able to identified groups of used car buyers who were two to three times more likely to consider a new vehicle as their next purchase. These results were incorporated into campaign management programs that selectively targeted used car buyers.
Title
Customer Satisfaction
Objective
Develop models that identify the significant factors of customer satisfaction with dealer service
Industry
Automotive
Outline
Decision tree methods were used to determine the factors that most affected a customer’s perception of the quality received for automotive dealer service. This not only provided the ability to estimate the satisfaction level for customers who did not return surveys but also to understand the drivers of dissatisfaction. The results were utilized in the development of focused dealer programs to improve customer satisfaction levels.
Title
Product Transition
Objective
Develop predictive models to be used in customer communication strategies for determining the likelihood and timing of customers transitioning across product categories
Industry
Automotive
Outline
Applied decision tree methods and logistic regression to determining the likelihood of a current customer purchasing a vehicle in another segment over the next year. Of particular importance was the likelihood of moving upward to more profitable segments. The results were utilized in campaign management to target customers for cross-selling and up-selling
Title
Improve Performance of Dealership Network
Objective
Improve Customer Satisfaction for Sales and Service Areas across Central Dealership Network
Industry
Automotive (Mexico)
Outline
Customer Segmentation based on Sales and Service Information gathered via surveys and customer information. Data Mining project to improve Customer Satisfaction Index Across Dealership Network. Performed customer segmentation to root-cause poor performance in Service and Sales areas across top 10 dealers in the company’s Mexico Network, made recommendations to improve data mining and surveys
Title
Sales of Training/Educational Services
Objective
Develop a method to predictive which potential customers are most likely to purchase educational or training services
Industry
Education/Training
Outline
Developed logistic regression models based on a 50 question survey completed by potential customers. Identified the key questions and responses that indicated which potential customers were most likely to purchase services. The results were utilized in marketing programs to target new potential customers.
Title
New Customer Acquisition
Objective
Develop predictive models that identify potential new sales
Industry
Insurance
Outline
Utilize both logistic regression and decision tree models to identify potential new customers based on purchase history and demographic and geographic factors. The results of the models were the basis for targeting direct mail campaigns. The implementation of these models resulted in a 20% higher response and sales rate versus random mailings.
Title
Customer Lifetime Value
Objective
Determine the lifetime value of current customers
Industry
Insurance
Outline
Develop logistic regression models to predict customer retention based on purchase history and demographic and geographic factors. Applied the retention likelihoods along with profitability per policy over time to determine a present value for each customer. The results were used for multiple purposes to include potential additional product sales and economics of campaign spending.
Title
Fraud Detection
Objective
Identify claims that are potentially fraudulent
Industry
Insurance
Outline
Develop logistic regression models of past claim characteristics to identify those that are potentially fraudulent. Results used to identify new claims that require closer investigation for possible fraud.
Title
Call Volumes
Objective
Develop a method to predict call volumes for a large security based call center
Industry
Security
Outline
Developed regression models to predict monthly call volumes based on a series of factors to include marketing specials, season, month, holidays, new leads, appointments, new sales, current customer level, etc. The results were used to improve agent scheduling which reduced overall operating costs.
Title
Increase Service Level and Minimize Shortages
Objective
Increase service level and minimize inventory shortages
Industry
Energy (Middle East Region)
Outline
Developed predictive model using information from customer demand, inventory levels and costs. Predict feasibility of distribution center for the Middle East Area based on historical data of customer requirements, emergency shipments and inventory shortages. Developed predictive models and validated using simulation models to compare various alternatives and completed cost-benefit analysis.