
Business Problem: Evaluating the effectiveness of a premium membership program and the possible reasons for decline in revenue
Data: Customer and sales data from the past and current year
Analysis: A series of statistical hypothesis testing was conducted
Actionable Results: Premium program shown to be effective at helping to up sell customers, while high pricing was a driving factor in declining revenue in the loss of customers
Business Problem: Identifying best customer demographics to target for selling optometry and ophthalmology equipment
Data: Survey data gathered from 200+ medical eye doctor businesses with 40+ questions ranging from demographic/firmographic to attitudes and firm planning
Analysis: Principle Component Analysis with centering and scaling was used to pre-process the survey data followed by clustering analysis. Cluster averages were used to identify key characteristics of each identified cluster
Actionable Results: 6 customer segments were identified and the top 2 were used for targeting best potential new customer demographics
Business Problem: An auction site for rare collectibles needing estimated selling prices for their unique items
Data: A sampling of characteristics and prices of past items sold on auction site
Analysis: A combination of variable transformation, similarity analysis, and regression was used to create an algorithm for estimated selling prices
Actionable Results: Algorithm that could be implemented in a variety of environments to estimate profit and sale price on each item

Check out
Mentor Analytics'
past projects and the type of work we can do for you!

Business Problem: Food manufacturer needs methodical solution to minimize waste of product based on individual product weights
Data: A sampling of weights of individual products coming off a particular assembly line
Analysis: Monte Carlo simulations were run on a variety of bagging/grouping algorithms. Parameters were used to determine the best solution after having received the typical weight distributions of products and numbers in each bag
Actionable Results: The best algorithm was selected with adjustable parameters based on weight distributions. This Algorithm could be programmed and implemented into manufacturing assembly line to minimize product waste
Business Problem: Company had a wages dispute with former employees and needed more concrete information
Data: Geolocation of company trucks logged over previous time periods which is of interest to the employees
Analysis: Parsing geolocation data into working and non-working time periods based on stoppage time of vehicles. After that, the data was aggregated together displaying the working time for each driver
Actionable Results: Settlement for much lower cost to company while avoiding costly court and lawyer fees
Business Problem: Price matching algorithms for provider-client marketplace which has similar model to Uber. Client having trouble finding good matches in various circumstances
Data: Historical data on successfully completed jobs, and on the bids of both successful and unsuccessful matched jobs
Analysis: A combination of data aggregation, key metric creation, regression analysis and hypothesis testing
Actionable Results: Analysis returned top 5 contributing factors of unsuccessful matches to allow client to find places to introduce surge pricing
Business Problem: A food subscription company is churning customers on regular basis and wants to find out the reasons why
Data: Demographic information, customer profiles, and order history for past and present customers
Analysis: A combination of data aggregation, variable transformation, multivariate and logistic regression was used to find high risk customers and identify high risk factors
Actionable Results: Top 3 reasons for customer churn was identified to allow company to focus efforts on mitigating those reasons for all customers. A set of high-risk customers were also identified to allow targeted incentives to be put in place to help retain those customers
Business Problem: An E-learning company needs better information on its users and their habits
Data: A running log of user actions taken on the
E-learning site regarding the action taken, the user identification, and the time of the action
Analysis: A combination of data aggregation followed by cluster and segmentation analysis was conducted. Clusters were then analyzed and assigned profiles based on the cluster characteristics and habits
Actionable Results: 3 types of typical users were identified and allowed the E-learning company to customize the user experience based on projected user type

​Business Problem: Health officials in Maryland and Baltimore area need to monitor lead levels in children
Data: Records of BLL (blood lead levels) in children by census tract number and year
Analysis: A heat map (with higher levels being brighter red) was creating using the data provided. An integration with census tract geography boundaries and Google Maps were used for easier visualization
Actionable Results: Heat map can be used in areas of high concentrations of high BLLs and allow health officials to target resources appropriately.

Business Problem: Need to identify lawyers who would be successful in a career transition
Data: A database of lawyers, their employment history, general demographics and schooling
Analysis: Created an algorithm that takes into account certain characteristics of a lawyer. It takes their employment history and then outputs a probability of moving law firms
Actionable Results: Algorithm implemented in a software version that allows law firms to target lawyers most likely to transition in their career. This aides in spending less time finding the right candidates
