March 2012
Predictive analytics transforms Dell’s marketing strategy
June 2012
Case study of how a unique marketing strategy based on statistical analysis of
customer relationships delivered significant incremental Enterprise revenue for
DELL Europe
A paper by Elizabeth Press, Sayantika Bhaduri and Sumanth Suresh
Dell‟s Transformation Journey From Computer Hardware to IT Solutions
Dell was founded in 1984, during the height of what IT industry insiders call the “PC/Client Server Era”,
a time when units of hardware sold was the key indicator of success. The 2000‟s and the advent of cloud
computing and
virtualization has
heralded the
“Virtual Era” for
the IT industry.
Application of IT as
an enabler of
business has
become the core
value-add of IT in
the "Virtual Era."
Hardware has
become
commoditized. Thus
demand for IT is
linked heavily to the evolution of customers‟ industries. In order for IT manufacturers to positively
differentiate themselves, they need to be able to best address the infrastructure and application needs
of their customers.
Marketing for IT Solutions Deeper Customer Understanding is the key With the transformation from being a PC-manufacturer to becoming an IT solutions provider, Dell needed
to revise the go-to-market model in order to be successful in the “Virtual Era.” Business Intelligence
needed to create effective targeting methods for deepening relationships with existing customers and
winning new customers. Instead of relying on direct relationships with the customer as a strategic
advantage, Dell needed to gain a deep understanding of demand for the products and proactively
approach customers with solutions. This prompted a new approach to the way marketing campaigns
were carried out.
Analytics Enabled Marketing Campaigns Challenges Customer Insights are incorporated into Campaigns
Leveraging the power of analytics allowed us to take a holistic and solutions-based approach to customer
targeting in our marketing campaigns. Previously, we had used rules-based criteria to identify
prospective customers. We were able to replace the rules-based criteria by solutions-based factors. By
using statistical modeling, we analyzed a large number of explanatory factors and identified most
important indicators of purchase intent.
Customer Targeting for Marketing Campaigns Analytics and EMEA BI teams combine their expertise
Dell Analytics and EMEA BI
Marketing started to work on
the marketing transformation
initiative in November 2010.
Leveraging the team's
combined experience in
strategy consulting, risk
management and predictive
analysis, the Business
Intelligence and Analytics
teams jointly developed this method of targeting. Based on the team‟s collective experience in the
financial services and IT industries, this tailored methodology is comprehensive and deep in its analysis,
yet intuitive and easy to understand for the business.
Analytical Engine
Challenges
Providing Solution is necessary in the virtual era
We are looking through the prism of Client Server
Our direct business model was no longer a strategic business advantage
Outcomes
Dell Positioned itself as a global solutions provider
Expanded our view of the customers with advanced analytics Enhanced targeting in key multifaceted business areas is a strategic advantage
Approach to Customer Selection for Campaigns A deeper look at predictive analytics for customer selection
Our earlier targeting methodology was very successful in the “PC/Client Server Era,” where unit sales
were the main goal. The “Virtual
Era” however demanded a holistic
view of customers, as the goal was
not only unit sales, but deeper
engagement and stronger
relationships. It required us to
proactively approach our Public
and Large Enterprise customers
with solutions to their business
problems. Predictive analysis
enabled us to understand what
solutions customers would need
and when they would need those
solutions.
Predictive Analytics for Customer Selection
“Give me the best customer to call” is the universal ask of all marketers and sales makers
The selection of customers for campaigns in the new solutions-based approach needed to be centered on
addressing the customer‟s needs. Furthermore, the approach needed to clearly differentiate between
customers who might genuinely require a product against those who don‟t in order to enable the sales
teams to better understand the likely customer requirements and contact customers with the maximum
likelihood to purchase.
Among the various analytical approaches to customer selection, a logistic regression was proposed
considering the “Give me the best customer to call” ask of marketers. Logistic regression models the
likelihood of a customer purchasing a certain product. The final outcome was a list of customers with
their individual propensity scores (likelihood or chance of purchasing a specific product). Higher
propensity scores indicated a higher likelihood of response to sales calls.
Implementation of Logistic Regression
Preliminary preparation for Modeling
Once the objectives, outcomes and methods were fixed, we identified the
various data categories (financial, service quality, product related and
market related) for building the model.
Searching for trends and patterns in the data – Exploratory Analysis &
Hypotheses
Exploratory analysis of the data revealed patterns pertaining to purchase
cyclicality and buying trends which were further explored and validated with
the help of business managers and sales teams. We also identified certain
customer sub-segments within the population who were more inclined to buy
the target products. This enabled us to frame specific class and category
variables as inputs to the model.
Model Variables - Creation, Reduction & Selection
We began with a list of 500+ model variables for every product modeled. The variable creation
incorporated various features of the data such as customer RFM characteristics and customer
firmagraphics (such as employee size, industry type). An important feature in our variable creation
process was the involvement of business stakeholders. We had multiple rounds of discussions with the
product managers, solution specialists, sales and BI teams to build variables likely to impact purchase
decision. We were able to identify several significant product affinity variables using insights from
various stakeholders. By identifying product affinities using the inputs of Solution Specialists we were
able to bring a „solution‟ basis to model building. The overall list of 500+ variables was reduced to a
smaller set of 25+ modeling variables by applying multiple statistical, business and sense check filters.
The Logistic Regression Model – Model Selection & Customer Scoring
The 25+ modeling variables were tested in various combinations and different models were iterated.
Statistical tests and out-of-sample validations were used to identify the better performing models. The
iterations were also shared with stakeholders to seek their feedback regarding the business significance
of the statistically significant variables. We also used sign tests and checked individual variable weights
to avoid heavy loading on any single factor. The final model with 5-7 variables was selected based on
fulfillment of all the above criteria.
Customer Scoring
The customers were scored using the selected model and the end result was a purchase likelihood score
for customers to buy a specific product. The final customer selection was made after excluding the
customers with whom we have lost a deal in the recent past (last 6 months) or who are already in a
discussion with the Sales team.
Monitoring Performance and ROI Measurement Intuitive metrics of campaign effectiveness In order to effectively monitor and communicate the performance of predictive marketing methods
across the organization, we created intuitive and simplified metrics to measure the incremental revenue
impact. We tracked the incremental revenue over sales targeted activities by measuring two metrics,
conversion and average order value. Conversion measured the targeting efficiency of the campaigns
while average order value measured the revenue derived from converted customers.
Business Impact of Predictive Analytics ROI, Strategic Customer Insights & Ideation Framework
Usage of predictive analytics for campaign targeting went from an innovative idea to a strategy-changing
practice within a year. The implementation of predictive analytics has impacted the business in three
fundamental ways:
Dell grew incremental revenue and improved sales effectiveness
Sales specialists and management received strategic insights about customers
A continuous ideation framework which promoted a structured approach to incorporation of new
ideas
Delivered significant campaign ROI
Dell increased revenue and improved targeting effectiveness in strategic enterprise products such as
servers and storage. The sales specialist organization used the output of the statistical models for
identifying their targets, enabling an effective and transparent quota setting method for Sales
Specialists.
Research/ Expert Input, Feedback from
last iteration
Hypotheses Formulation,
Exploratory Analysis
Presentation of Hypotheses, Analysis
findings
Open Discussions with larger teams - Product, Sales, BI
Model Building and Extraction of Insights
Strategic Insights to Executive, ROI measurement
Provided Strategic Input for the Business
The insights gained during the modeling process
enabled framing of marketing strategy and
supported the Marketing and Sales teams in
understanding their customers better. One
example of business insight gained through
modeling was that the customer scores revealed a
“Pecking Order” which further validated the
„solution‟ approach used in the model building.
Additional insights that we have presented to
management have included:
Identification of seasonal trends
Firmographic niches for the different products.
Customer trend analysis.
These findings have been inputted into executive level strategic decision making and implemented in
transforming the go-to-market model.
Developed a continuous ideation
framework
We have created and documented a
process in our team to enable
continuous application of the insight
we gained through analysis into
strategy.
The purpose is to enable continuous
structured brainstorming and new ideas
in the organization. IT is a very
dynamic industry and impacting
innovation processes have created a
strategic advantage for Dell. Our
process takes a cross-functional and
thus interdisciplinary approach to
ensure a comprehensive look at the
issue to enable optimal insight and decision making.
High
• Solution Buyers: Customers who are highly likely to buy a specific product as part of an overall Solution
Medium
• Hardware Buyers: Customers who are highly likely to buy a specific product
Low
• Non Buyers: Customers who are unlikely to buy a specific product
“The Pecking Order”
Continuous
Ideation
Framework
And Dell‟s customers are more satisfied…
Our solutions- based approach has enabled us to sharpen our focus on the customer and providing them
with the Power to Do More. This approach can be applied by other companies in almost any industry.
Successful implementation of the principles of comprehensive analytics, collaborative team work with
sales and marketing, straightforward metrics and communication, coupled with a process of continuous
ideation and strategic insight can help companies build better relationships with their customers.
About the Authors:
Elizabeth Press ([email protected]) is a Research Sr. Advisor in the EMEA Public and Large Enterprise Business Intelligence Team and based out of Frankfurt, Germany. She has a BA in International Relations from Tufts University and an MSc in International Economics and Business from the Stockholm School of Economics. She has worked in strategy consulting and the finance & technology industries.
Sayantika Bhaduri ([email protected]) is an Advisor with the Marketing and Sales analytics team in Dell Global Analytics and based out of Bangalore, India. She holds a Masters in Mathematics from IIT Kanpur and has worked in Marketing Analytics for technology industry.
Sumanth Suresh ([email protected]) is a Sr. Analyst with the Marketing and Sales analytics team in Dell Global Analytics and based out of Bangalore, India. He has a Masters in Engineering from IIT Madras and has worked in consulting and analytics.
About Dell Global Analytics
Dell Global Analytics seeks to improve Dell‟s bottom line through the leveraged use of analytics touching all aspects of Dell‟s business operations. We offer a wide range of analytics services covering management reporting and dash boarding of key business metrics, forecasting and predictive customer response modeling and optimization of key business processes. Value for Dell is unlocked by the application of sophisticated data analysis, statistical and mathematical techniques under a Six-Sigma framework of business process improvement.
The range of supported Dell functions includes Dell Supply Chain, Pricing, E-Commerce, Contact Center Operations, Dell Financial Services, Marketing and Sales.
Office:
Dell International Services India Pvt. Ltd
Divyashree Greens,
Survey No 12/1, 12/2A, 13/1A, Challaghatta, Varthur Hobli,
Bangalore 560071, INDIA
About Customer Insight & Business Intelligence, EMEA, Public and Large Enterprise
Customer Insight & Business Intelligence drives the provision of live, relevant and timely business intelligence information and customer insight to the Public and Large Enterprise sales and marketing leadership teams throughout EMEA. We also provide targeting and strategic insight for EMEA-wide campaigns.
We are located at various locations throughout EMEA.