Unblocking The Main Thread Solving ANRs and Frozen Frames
Case Studies - Customer & Marketing Analytics for Retail
1. CUSTOMER & MARKETING INTELLIGENCE SERVICES
Customer Targeting Strategy Development
Case Studies
2. Customer Intelligence
Identifying Most Valuable Customer
Case Study
Client: A US based high end luxury retailer in the space of apparels and accessories
Business Context & Client Problem:
The retailer has several high end luxury retail stores in key cities across the US and sell luxury apparels and related
accessories. They want to identify the most valuable customers within their existing customers based on relationship
and purchase patterns for proactive relationship management purpose. They also wanted to assess the profile of such
best customers and catch them young in their lifecycle.
Impact on Business:
The study not only identified the sweet spot of the business , but also created detailed understanding of best customer
profile and characteristics and helped the retailer to create effective CRM program
Solution: Customer segmentation based on relationship quotient and identification of demographic and
behavioural sweet spot for most valuable customers
Relationship Segmentation Segment Drill down Identify Look Alikes
• Analysing various transaction
patterns, cycles, mix etc.
• Design segmentation scheme
based on Recency, Frequency
and Monetary factors
• Identify the Most Valuable
Customer segment covering
80% of the business value with
20% customers
Understanding behavioral and
demographic characteristics of the best
customers against rest
Create scoring model to identify
customers who look like the most
valuable customers, however not yet
reaching the status
3. Customer Intelligence
Targeted Cross Sell
Client: A leading global technology company
Business Context & Client Problem:
A global technology giant wanted to cross-sell profitable docking stations to some of their existing SMB (Small &
Medium Business) customers. Due to budget constraints, the company wanted to be focused on reaching out to the
right set of accounts based on their propensity of buying a docking station in the near future. This could be derived
through the relationship quotient, product purchase sequence as well as the estimated need of the company.
Impact on Business:
A targeted cross-sell campaign based on relationship, product purchased so far and industry dynamics helped the sales
people to not only focus on high RoI accounts, but also helped them customise their communication
Solution: A scoring model to prioritise the set of accounts based on R-F-M segmentation as well as natural product
association between docking station and other products
Relationship Segmentation Product Association Analysis Scoring Model
• Analysing requency, frequency
and monetary aspects of
relationship with the accounts
• Design segmentation scheme
based on RFM characteristics
• Identify sweet-spot for best
customers
Identifying association between various
products based on how often they are
bought by same customer
Weighted Model incorporating
relationship quotient, installed base
association and industry dynamics
Case Study
4. Customer Intelligence
Prediction of New Product Sales Trajectory Leveraging Social
Media Buzz & Sentiments
Client: Global marketing organisation of a leading manufacturer of personal computers
Business Context & Client Problem:
In a market where product life-cycles are a few months long and competition is heavy, waiting for and relying solely on
point-of-sales data was less predictive and constraining in terms of quick course corrections. The PC manufacturer
wanted to utilise the market buzz and indications obtained from social media on early days of launch to predict
potential growth path of the product.
Impact on Business: The solution provided initial insights on key social media indices to track for assessing performance
of a product and react quickly to potential corrective actions. Such solution is expected to be technology enabled and
operationalised across various products
Solution: Crawled data from social media sources like Twitter, Amazon, Google etc to create predictive indices
around market buzz and consumer sentiments on key features to correlate with potential sales trajectory of
launched product.
Creation of Social Indices Build Predictive Model Operationalisation of Solution
• Crawling of mentions, reviews, comments from
various sources like Twitter, Amazon & Google
reviews, CNET for 9-10 products launched in
last 2 years
• Advanced text mining to identify key features
and scoring sentiments displayed
• Creation of social indices around mentions,
promotion, average reviews, sentiments across
key features for each product lifecycle
• Standardisation of growth trajectory of
various similar products through parametric
curves
• Creation of an advanced panel regression
model to relate the social indices and trends
with the growth observed over time for
various products
• Assessing most predictive factors for relating
with growth trajectory and build a scoring
model involving various social indices
• Developed set of indices which are
highly predictive about product
performance
• Operationalising the technology
solution
Case Study
5. Marketing Effectiveness
Measuring Impact of Trade Discount & Promotion
Client: An India based Consumer Packaged Goods Giant
Business Context & Client Problem:
The client is a conglomerate of diverse business lines with a significant focus on Consumer Packaged Goods, especially
food. In this scenario, the client wanted to measure effectiveness of various trade & consumer promotion on trade
revenue with wholesalers, convenience stores and retailers segregating the impact of price
change, promotion, competitive actions and cross SKU cannibalisation & Halo effects.
.
Impact on Business:
The analysis not only revealed hidden patterns of true effectiveness of different promotional spends and cross-category
interactions, but also provided enough insights for differentiated promotion strategy across segments
Solution:
Segment Data, Create Indices Modeling Decomposition of Impact Analyse Scenarios
• Segmentation of outlets based on similar
responsiveness and product assortment
• Creation of Price Indices, Promo calendar,
competition indices and cross-category
interaction indices
• Treatment of data for trend, seasonality, outlier
etc.
• Analyse underlying patterns based on first week
or last week of the month, start of year ,
significant changes etc.
• (Mixed Effect) Regression modeling of
volume sold against price, promotion,
competition and interaction indices
• Decomposition of volume realised into
base volume, promo net impact,
cannibalisation & competition impact
etc.
• Analyzing RoI of promotion spends
based on incremental value
• Identify optimal promotion & price
for each channel & segments
Case Study
6. Marketing Effectiveness
Optimising Marketing Mix
Client: An online education company based in the US, which offers associate degree programs and other
certifications based on tie-ups with universities and self generated content
Business Context & Client Problem:
The client organisation deploys a variety of marketing vehicles to generate awareness and demand for their course
offerings. These are both online like Display Advertising, Cost-per-action (CPA),Pay-per-click (PPC) arrangements and
offline activities like Branding initiatives. There is a need to understand the relative effectiveness and ROI from each of
the marketing vehicles, so that the marketing mix can be optimised to get the best return on total marketing spend.
.
Impact on Business:
Based on the model and tool’s suggested marketing mix, there is an estimated lift of 10% in ROI which translates to
approximately $1.3 Million on an annualised basis for the current marketing budget
Solution: A market mix model that provides estimated ROI for each of the marketing vehicles and an Optimiser tool
that uses the ROI estimates to suggest the ideal marketing mix for a given marketing budget
“Base” Sales & Incremental
Effect
RoI Estimation for Mktg Vehicles
Optimiser Tool for the
Ideal Mktg Mix in a Given
Budget
Estimate “Base” sales and incremental effect
due to marketing vehicles
Average and marginal RoI for marketing
spend in each vehicle
Directional suggestion of marketing
mix changes and incremental RoI
Case Study
7. Text Mining of Qualitative Inputs in Surveys with
our Cloud Based App.
Impact on Business:
• A cost effective way to analyse unstructured text data fast and derive actionable insights from it
• Increase speed of drawing insights
• Simplification of analytics in the hand of business managers
Solution: A comprehensive survey data analysis which support advanced requirements in predictive analytics and
text mining with an easy-to-use interface designed for business managers
Input Data Interactive Outputs
Business Context:
Responses to open-ended questions in market research and customer surveys typically go unanalysed albeit there are
immense insights in them. Our team set out to frame the text analysis problem and create an application that can
address all needs of survey analytics in one place. While the application has a broader survey analysis flavour, it has a
key textual analysis module that addresses most of the textual analytics needs.
Intended Clients: Market Research teams , Customer service , Brand management, Loyalty management, etc
• Any kind of textual data with
as many split-by variables
• Capability to handle BIG Data
• CSV, Excel formats
• Capability to fetch data
automatically from any
database
Word Cloud
Thematic
Analysis
Sentiment
Analysis
Interactive
Charts
Case Study
8. Personalised Recommendations
Objective
To identify upto top 5 recommendations
for cross-sell program for a retailer
Results
The improvement in precision was 8x using demographic
information of customers and 7x using only purchase history
Summary of Retail
Data
Precision = Percentage of correct recommendations in all the recommendations
# of Customers 8,419
# of Products 1,559
# of Transactions 58,308
Avg # of Products 5
Time Period (years) 2
Random
Recommendations
# of Hits Expected # of Hits Precision % Improvement # of Hits Precision % Improvement
1 4,668 14 96 2.06% 686% 115 2.46% 821%
2 4,668 28 185 1.98% 661% 204 2.19% 728%
3 4,668 42 260 1.86% 619% 266 1.90% 633%
5 4,668 70 362 1.55% 517% 374 1.60% 534%
Based on Purchase Data only Based on Purchase and Demographic
DataNumber of
Recommendations
Customers in
Validation Dataset
Random vs Our Recommendation Solution
The approach includes identification of significant
demographic attributes influencing customer preferences
Similar techniques could be used to predict customer
ratings (e.g. movie, CDs, games, etc.)
The approach could be used to recommend online
customers and could used in conjunction with click-
stream data
Case Study