Predictive Analytics As The Proverbial
Early Bird
BY DON MURPHY, PEOPLESOFT
Why does the early bird get the worm? The behavior of living organisms
results in part from data gathered in past efforts for survival.
Similarly, expertly leveraged data are a strong factor in successful
initiatives within the corporate 'ecosystem.'
Customer relationship management (CRM) analytics are known for powerful
synthesis of historical customer data. However, the data's potential
often goes untapped. With the right tools and a little intuition,
organizations can increase profitability by leveraging customer
information to anticipate customers' needs and influence their behavior.
By consolidating customer information from within and beyond an
enterprise, organizations establish the groundwork for modeling and
predicting customer behavior. Quality information gathered from numerous
customer-facing and third-party systems can help companies define and
select customer populations for building predictive models. Companies can
then use these models to segment and target customers for marketing
campaigns, service programs and customer loyalty and churn analyses.
Descriptive And Predictive Analytics
Comprehensive CRM analytics include both descriptive analytics and
predictive analytics. Descriptive analytics ' also known as performance
analytics, operational analytics or effectiveness analytics ' can
describe everything about a company's customers, activities and
performance up to the most recent moment. Predictive analytics leverage
this wealth of customer information to generate models targeting the
likelihood of future behavior in a given segment. Proper interaction
between descriptive and predictive analytics can provide practical tools
to help organizations understand their customers, anticipate customers'
needs and influence customer decision making through well-aligned
programs.
The best analytic products draw information from systems throughout an
enterprise, including front-office systems such as sales force automation,
marketing automation, the customer interaction center, financials and
others. Customer data, purchasing information, sales force contact
histories, call records and all other customer-related information
companywide should be included, as well as third-party demographics.
Placed in a central data repository (data warehouse), the result is a very
broad, dense range of consistent data.
Descriptive analytics use these data to take a historical look at
organizational performance. From a customer relationship viewpoint,
descriptive analytics reveal how effective the company's marketing
efforts have been, how the sales force is performing and if support and
services have been meeting client expectations.
Predictive analytics use customer data to provide a forward-looking
perspective about an organization's consumers. Predictive analytics help
identify customer segments so organizations can better understand customer
behavior, gaining the ability to leverage opportunities and proactively
control risks.
Three Steps To Predicting Customer Behavior
Customer behavior analytics use the predictive analytics process to
identify customer segments, predict customer behavior and ultimately help
make customer interactions more effective and profitable.
Predicting customer behavior involves three steps:
Profiling. Companies first build a profile of information
about customers who have previously exhibited a targeted behavior.
Profiling requires rich customer data, including enterprisewide
transactional and behavioral data such as call center and purchase
information. Other data sources include key performance indicators and
third-party demographics. An example of profiling might be building a
profile for customers who bought new homeowners' insurance policies in
the past two years. The goal is to determine characteristics to look for
in future buyers.
Modeling. By using data mining on the profile
information, analysts can uncover the most relevant characteristics of the
customer segment being analyzed. For example, the most significant
attributes of customers who bought homeowners' insurance are gleaned
from the profile via the data mining application. Such characteristics
comprise the model of customers most likely to purchase homeowners'
insurance in the future.
Scoring. Analysts use predictive analytics to score
existing customers by comparing them to the model. Those most closely
matching the characteristics included in the model are most likely to
exhibit the targeted behavior. In the insurance example, a company can
rate its customers numerically to indicate how closely they match the
model of the person most likely to buy homeowners' insurance.
Once customers are scored and the analysis pinpoints customers most
strongly correlating with the model, a company can address those
customers, especially the top prospects. Customers scoring a nine or above
might receive a special promotion for homeowners' insurance, while a
separate, incentive-based offer might entice those scoring seven and
above.
Targeting high-scoring customers increases the likelihood of correctly
identifying needs, generating positive responses and achieving campaign
goals. ROI is enhanced by eliminating efforts directed toward customers
who are not part of the targeted segment defined by the predictive model.
Analytic Feedback Loop
It is best practice to test a model by scoring a sampling of existing
customers. Once scoring results are analyzed, the model can be fine-tuned
and tweaked to increase its accuracy.
When using predictive models for live campaigns and initiatives, it is
critical to analyze the results with descriptive analytics for an
understanding of the model's effectiveness and success with respect to
campaign objectives. Descriptive and predictive analytics reinforce each
other. An iterative loop results, helping companies identify accurate
models in which they have confidence ' these are the models relied upon
for reuse over time. Companies that fail to use descriptive analytics to
measure results lose out on opportunities to clearly see how effective
their efforts are.
Web Enablement For Time- And Cost-Savings
Ideally, a company will leverage a Web-enabled CRM analytics
environment to automate and streamline all analyses. Speed, efficiency and
convenience are maximized with integrated Internet applications providing
secured access to front-office data, call center and support records, the
data warehouse, customer profiles and models, scoring mechanisms, the
campaign results repository and every other aspect of CRM analytics.
Close integration of customer analytic solutions with numerous data
points makes predicting customer behavior an increasingly accurate,
productive undertaking. Web enablement increases speed and reduces
deployment efforts and costs.
A Foothold In The Future
According to META Group, 'Our research reveals that in the next 12
to 18 months, analytical CRM solutions will be among the fastest growing
CRM technology segments. Indeed, executives believe that optimizing
customer-related business processes, maximizing customer value and
personalizing customer experiences is more than an art. While analytical
CRM solutions come in many forms (e.g., general query and reporting,
scorecards, data mining, packaged analytic applications), those that are
most tightly integrated with multiple data sources and into operational
and collaborative CRM business processes will yield the greatest returns.'
Analytics are becoming increasingly mainstream. It will be critical to
establish platforms for consistent delivery and management of information
across all analytic applications for collaboration within the extended
enterprise. Analytics are supporting more users and more roles. Fragmented
approaches can generate scalability and information consistency issues.
By modeling and predicting customer behavior, companies are beginning
to use the wealth of information gathered by their front-office CRM
systems to tap the potential of CRM analytics. The early bird gets the
worm ' and the organization that leverages customer information by
skillfully applying descriptive and predictive analytics reaps ongoing ROI
from well-crafted campaigns and initiatives.
Don Murphy manages
CRM Analytics product strategy for PeopleSoft
Inc. He has extensive direct-response and database marketing
experience, as well as experience with multichannel targeted marketing
solutions.
[ Return
To The January 2002 Table Of Contents ]
|