data based markeing and mining to select the best prosopects for your product or service
DESCRIPTION
There are two distinct worlds that we live in today. Most people understand the real world and the virtual world of databases, Facebook, LinkedIn and so forth. If you have tried cold calling on the phone or in person today, it is "real world" brutal and usually does not work very well. I doubt many sales people can stand the constant rejection which is why there is so much turnover in sales. This leads to the idea of “Smart Prospecting” where businesses create mathematical definitions of their best prospects among a variety of market segments and then use these definitions to create data mining algorithms. If the budget is tight, there are expert system rules and Boolean logic algorithms that do not cost much to build and they work very well. Even if you do not have the resources to afford sophisticated data mining tools we can find qualified prospects by using search engines in Google as well as Data.com, Netprospex.com, Manta.com, Inside view, OneSource iSell, and of course linkedIn.com. The goal is to get lists of qualified prospects so the sales people do not waste time prospecting to the wrong companies that do not need the product they are selling and would never buy from them. This cuts down on rejection and wasted time as well as gas for the car. Then the sales person can smart call with specific targeted benefits that will help them to get appointments with qualified prospects. Every business and consumer exists in multiple databases and smart prospecting saves a lot of wasted mileage, rejection and turnover. There are 7,705,000 businesses in the United States and it costs very little to have the computers select only the best prospects. Most of these 7 million businesses are not your prospects - so don't try to sell them. If only 1 percent are your prospects that still leaves 77,000 businesses. Here is my presentation on data mining where we select the best prospects before we have the sales department call on them. I used this when I was a guest speaker at the Kodog business school and the data mining was used by me for Blue Cross Blue Shield, US Sprint, Fannie Mae and many other large firms in the DC area.TRANSCRIPT
The Art and Science of
Database Marketing
Robert J Fulmer – Database(d) Marketing. 19701 Grayheaven Manor Rd. Olney, MD 20832
Cell 240-328-5846 [email protected]
Virtually Every “Active” Consumer and
Business is in Multiple Databases
• Consumer Compiled Databases-Polk,
Metromail, Donnelly, Infobase, DBA
• Consumer Credit-Equifax, TRW, Transunion
• Consumer Psychographic-Lifestyle Selector
• Geodemographic-Prizm, Microvision,
ACORN, CRMI, Updated Census Data
• Business Compiled-ABI and D&B
• Consumer and Business Lists
• Internal Customer Databases-Name &
Address, Transaction, Billing, Customer
History
US Sprint, AT&T, MBNA, ING, Travelers Group,
Solomon Brothers, T. Rowe Price, Sony, Blue Cross-Blue
Shield are making Millions/Billions of dollars with
Databased Marketing.
We offer your firm the exact same...
• Computer Hardware
• Computer Software
• External Data Sources
• Proven Marketing Applications
• Statisticians-Model Developers
• Competitive Pricing
We Have Fortune 500 Experience &
We Outsource With the Best Vendors
• Marketing Research Experience-Projects for USA
Today, The Wall Street Journal, NBC Television, Fox
Broadcasting, Hearst, VNU
• Media Planning Experience-Toyota, Procter and
Gamble-Saatchi and Saatchi
• Strategic Business Relationship with Experian,
Interlogic, ASC Database Marketing, Merkle Direct,
Info USA and Dun & Bradstreet
Since every active prospect exists in multiple
databases, the key to database marketing is
picking the prospects that will..
• Respond
• Respond and make a
purchase
• Respond, make a
purchase and then make
more purchases
Direct Marketing Evolved into Database Marketing
• Direct Marketing
• Focus on Lists of
Consumers or
Businesses
• Uses Descriptive
Research to describe
Past List Performance
• Limited Targeting
Capability
• Database Marketing
• Focus on Individual
Consumers or
Businesses
• Uses Predictive Research
to Score Individual
Prospects
• Unlimited Targeting
Capability
Traditional Direct Marketing
Mailing List #1
Mailing List #2
Mailing List #3
{
{
{
Make a decision based on response by list
Highly Responsive
Continue to Mail
Marginal
Test Again
Poor Response
Drop List
Database Direct Marketing Marketing to the Individual
Create a
Customized
Marketing
Database by
Combining all the
lists into one file.
Database Direct Marketing Enhance the File with Internal and External Data
Add back all
records and
append additional
data overlays
Database Direct Marketing Model, Group and Score
Find the prospects
most likely to
respond
Traditional Direct Marketing
Income
Res
ponse
Rat
e
0 +
+
Response Rate Correlation:
Response Rate and File Information
Traditional Direct Marketing
Income
Res
ponse
Rat
e
0 +
+
x
2
y
2
x
1
y
1
Response Rate Correlation:
Linear Analysis - Two Dimensional
Database Marketing R
esponse
Rat
e
0
+
x
1
y
1
Response Rate Correlation:
Linear Analysis - Three Dimensional
Database Marketing R
esponse
Rat
e
0
+
Response Rate Correlation:
Multiple Linear Analysis - Three Dimensional
Database Marketing R
esponse
Rat
e
0
+
Response Rate Correlation:
Multiple Linear Analysis - Three Dimensional
Database Marketing
• RFM (recency, frequency, monetary)
• CHAID (decision tree)
• Linear and Logistical Regression
• Neural Networks, Systems Engineering
Modeling Methods
Database Marketing
• Select historical response data
– (past direct mail campaigns)
• Analyze the respondents and the non-
respondents
– 2,000 - 5,000 each
• Score (i.e. 0-999)
• Segment (quintiles, deciles, etc.)
Process
The Intelligent Closed Loop is a Source for
Positive and Negative Response Data, Demographics and
Payment History that is used to Build Predictive Scoring Models
Customer-Database-
Purchasing Information
Target
Selection •Acquisition
•Up-Sell
•Cross-Sell
•Retention
Message
Selection
Implementation Printing -- Data Processing
Personalization -- Letter-Shop
Response
Analysis
Data Cleaning
Data Upgrades •Purchasing Data
•Demographics
•Psychographics
Customer Profile
Predictive Model
Yes
No
Modeling Methodology Used by
Telematch, US Sprint and Experian
Large Heterogeneous Mixture of All Customers on a File
Segment A
Homogeneous
Segment B
Homogeneous
Segment C
Homogeneous
Segment A Response Model
Segment B Response Model
Segment C Response Model
List Selection
Segment C
List Selection
Segment B
List Selection
Segment A
Responders &
Non - Responders
Demographic
Attributes
Each List Selection Goes to Telematch
Database Marketing Focuses on Individual
Positive and Negative Consumers to Build
Predictive Scoring Models
• Negative Consumers
• Non-Responders
• Non-Convert to Customer
• Amount of Order is
Unprofitable-Unpaid Bill
• Customer Lifetime Value
is Low
• Financial Performance-
Low
• Positive Consumers
• Responders
• Converted to Customer
• Amount of Order is
Profitable-Paid Bill
• Customer Lifetime Value
is High
• Financial Performance-
High
The Enhanced Modeling Database is used to build
Predictive Models with a Positive and Negative
Customer Profile. These Positive and Negative
Profiles can then be used to Score the Entire
Range of a Prospect File or List
Positive
Customer
Profile
Negative
Customer
Profile
Male
Age 45
Sales
HH Income $55K
Female
Age 36
Professional
HH Income $65K
Prospects that closely match or are exact
opposites of your positive customer profile are
the easiest to identify.
Predictive Power of Models
Strong
Negative Average Strong
Positive Correlation +1.0 -1.0 Correlation
Positive
Customer
Profile
Negative
Customer
Profile
. .
.
.
. .
.
.
.
. .
.
.
.
. . .
.
.
.
Multivariable Regression-Simplified
y
x
Positive Value
Dependent Variable 1.0
Purchasers/Donors
Negative Value
Dependent Variable -1.0
Non-Purchasers
Independent/Predictor Variables
V
Regression is a technique for finding and describing a functional (mathematical)
relationship between a dependent variable (Purchaser) and one or more independent or
predictor variables.
v 1
2
Database Marketing
Database Append
Sample Mailing
Model
Score
Segment Validate
Non Response
LIST LIST LIST
Responses = Past Mailing
Positive Response
Responses
Responses
Process
S egm ent S core Range Database Quantity A nticipated % Rate
S e g m e n t 3 9 9 9 to 5 7 4 2 5 5 ,7 5 2 0 .8 1 %
S e g m e n t 2 5 7 3 to 4 6 5 6 0 6 ,3 7 1 0 .7 2 %
S e g m e n t 1 4 6 4 to 0 9 7 5 ,4 6 0 0 .4 7 %
1 ,8 3 7 ,5 8 3 0 .6 1 %
Segment Analysis
Segment 3
14%
Segment 1
53%
Segment 2
33%
R e sp o n se R a te A n a ly sis
0 .0 0 %
0 .1 0 %
0 .2 0 %
0 .3 0 %
0 .4 0 %
0 .5 0 %
0 .6 0 %
0 .7 0 %
0 .8 0 %
0 .9 0 %
Se g m e n t 3 Se g m e n t 2 Se g m e n t 1
A vg . % R a te
Database Marketing Final Analysis: Example of a homogeneous audience
Database Marketing
In a customer file with purchasing history and external demographic data
it is possible to build models that have a lift of 150% over the average response
rate and have a lift of 10 to 1 from the best to the worst decile in the file.
0
0.5
1
1.5
2
2.5
1 2 3 4 5 6 7 8 9 10
Percent
Response
Rate
Deciles
150% Lift above Average
1,000% Lift Best to Worst Decile
1%Average Response
Two Dimensional Matrix Probability to Respond & Estimated Spending
Response High Low
Spending
High
Low
DECILES 1 2 3 4 5 6 7 8 9 10
1
2
3
4
5
6
7
8
9
10
The Customer Life Cycle
and Marketing Activities
Lose Retain Develop Acquire
Network
Reactivating
Retaining Cross
Selling
Upselling
Networking Converting
Relationship
Marketing Prospecting
Objectives and Benefits of Modeling
• Acquisition Models: Raise response rates, number of orders and
amount of orders; acquire customers with higher lifetime value, lower
return rates and higher payment rates.
• Conversion Models: convert customers with high lifetime value and
medium and long term profitability.
• Relationship Marketing Models: Increase survival time and lifetime
value of customers.
• Cross selling and Up-selling Models: Increase business base with
customers selling greater depth and width of products and services.
• Networking Models: Acquire new leads to be filtered through
acquisition models.
• Retention Models: Increase all positive indicators of customer
performance while preventing customer defection.
• Reactivation Models: Restoring performance of inactive customers
clients or donors.
Case Study: New Customer Acquisition
Blue Cross/Blue Shield Federal
• Blue Cross/Blue Shield budget only permits mailing 1.25
million names on their prospect list of 2.2 million names.
• Previous years-random deselection was used to reduce
mailing quantity.
• Created a new customer acquisition model, validate,
overlay prospect file with demographic data from R.L.
Polk, score the file and select best 1.25 million names.
• Saved over $9 Million in foregone revenue-for a cost of
$50,000- 180 to 1 Return on Investment.
• Blue Cross/Blue Shield Federal now owns a predictive
model that is being updated and used for future mailings.
Database Marketing
Strategic Applications
• Strategic Business Development: Use a new customer acquisition model to
build the customer base with profitable new customers- with good lifetime
value.
• Improve financial performance through customer development. Target
customers for cross-selling and up- selling. Increase the amount and
frequency of purchases from the customer base.
• Improve customer retention by identifying “at risk” profitable customers
before they fire your firm. Offer special products and services to save these
“at risk” customers.
• Practice Relationship Marketing by offering special products, services and
communication targeted to a segmented customer database.
• Apply Financial Models to all stages of the Customer Life Cycle. Improve
new customer acquisition, customer development and retention. This will work
like compound interest to dramatically improve growth and earnings.