Data Science Services, Data Science Consulting Services, USA

Adopt data-driven machine learning to improve future outcomes

By doing data science not statistical modelling

GrayMatter practices an exploratory approach to discover relationships between variables.
This is opposed to statistical modelling wherein an assumed hypothesis is confirmed or rejected based on statistical significance of the assumed relationships. What this means is that we let the data free to provide all possible patterns instead of trying to force fit the data into one of the assumed hypotheses
Validation of patterns with the help of domain experts to confirm meaningful relationships
What this means is that all the possible relationships are validated with the help of GrayMatter’s domain experts, that ensures that the model is relevant

Quick Enquiry

Business Problem

Company provides life term policies with monthly premium schedule. Non-payment of premium for 3 consecutive months is considered Lapse. Current lapse rates were 20%. Company wanted to understand factors responsible for the policies to lapse and predict likelihood of a customer/policy to lapse in next 3 months period

Solution Approach:

Collected data of all the policies and their events since inception of business. Data included

Customer Demographics (captured as of inception date) – Gender, Age, Race, Income, Nationality, Marital Status
Customer Interaction mode and frequency with company – Email, Phone, others (fax, letters)
Number and type of insurance products customers have bought from the company
Policy details – Agent, Sum Insured, Premium, Term
Each event for the policy – Inception, Lapse, Claim, Reinstatement, Cancel, Surrender, Mature

Significant time was spent in cleaning data and correlating with business processes. Ensemble of Decision trees, Logistics regression and Naïve Bayes models were used

Key Findings

Age, Customer Interaction and Agents the key contributors for the Lapse.
Agents were cross selling other insurance policies to their own clients after initial cool-off period.
Model provided Precision of 82% and Recall of 74% on a test dataset with 50000 policies.

Employee attrition management with the help of segmentation model. Identification of high performers and high potential employees who are likely to leave and adopting retention strategies. Identification of underperformers and early warning to them for course correction

GrayMatter’s demand driven dynamic pricing system for car parks in an Airport or even in Malls or Parking Zones. GrayMatter’s Car Park Revenue Management System predicts and recommends future price every day for few days ahead based on historical data analytics, algorithms and business rules. In the image below you can see the projected % occupancy depicted using red-amber-green indicators, along with flags for price change. In the table below, the actual booking, actual capacity and projected demand are shown along with projected price

Large supermarket store customer segmentation to identify customer segments like premium customers, price-sensitive customers etc. The demographics and behavior of every segment is analyzed in conjunction with the % revenue contribution of that segment to decide upon actions for the key segments like loyalty program for premium customers, product positioning specific to price sensitive customers and so on.

Segment

1 (0.2%)

Segment

2 (2.2%)

Segment

3 (1.2%)

Segment

4 (22.6%)

Segment

5 (33.4%)

Segment

6 (26.3%)

Segment

7 (12.4%)

Segment

8 (0.5%)

Segment 3 Characteristics

Demographics

Average Age: 40.2 Yrs – Maximum among all segments
Highest population of Females: 57%
Income: Highest estimated Household income

Behavior

Highest Fresh Food proportion among all segments
Average Member duration: 2.54 Years
(Highest)

Regression analysis for a leading alcoholic beverages company, analyzing country-wise beer consumption data to find out correlation between sales, consumption volumes & brand recall and derive predictors for sales thereof. In the charts below you can see that sales and volume (consumption) are highly correlated for almost all the countries whereas sales & brand recall as well as volume & brand recall are highly correlated for 8 countries.

Video analytics to enable accurate estimation of count of people in a queue, by taking into account congestion due to surrounding crowd. Applications of this include airports, taxi companies etc. wherein resultant actions include advance intimation of wait time, opening of new counters and optimal resource allocation (security, AC, cleaning staff etc.). The chart below clearly shows the corrected count vis-à-vis the manual count by factoring for congestion..

Analysis of consumer sentiment towards a beverage brand, on social media. The purpose of the analysis was to gauge pre-launch consumer response to the brand, right up to date of launch of the brand. As you can see in the outputs below, the percentage share of responses/comments for every brand is depicted. The share of negative and positive sentiments for every brand are also shown.

WordPress Tabs Plugin

Data Science Services

GrayMatter delivers unique, path-breaking data science solutions and services to global clientele. GrayMatter’s offerings range from data science consulting to actual data preparation and modelling, followed by ongoing optimization and performance tuning of the models. We also migrate algorithms, models across advanced analytics platforms. We also provide training services to increase business user adoption of data science offerings. GrayMatter’s Center of Excellence for Data Science ensures that the latest, cutting-edge offerings are offered to global clientele and the in-house competencies in Data Science are constantly enhanced, to help retain and grow GrayMatter’s presence as outstanding data science service providers, differentiated amongst the several data science consulting and implementation companies.

Consulting

Scope identification, feasibility assessment, choice of tools and algorithms

Data Preparation

Enriching the data set by missing value replacement, outlier analysis, categorical variable definition

Model generation

Generating the model, testing and refining on basis of validity of output

Performance Tuning

Enhancing model and scripts on ongoing basis, over and above model self-learning

Migration

Migration of algorithms, models from one platform to the other, e.g., SAS to R

Training

Training of business users as well as technical teams on SAS, Python and R

Expertise across Technology Stacks

GrayMatter has a dedicated Center of Excellence for Data Science (CoE-DS) with expert Data Scientists. GrayMatter’s Data Scientists have skills in a variety of technologies & tools including SAP PA, MS Cortana, R, SAS, Python, SPSS, Weka, MATLAB and others. The span of offerings include data science solutions like car park revenue optimization, queuing analysis, insurance lapse prediction analysis, social media sentiment analysis and others. The CoE also enables custom Data Science services for customers across verticals.