Transcript Prospect Research: In-house Options to Advanced Outsourced

Expanding the Scope
of Prospect Research:
Data Mining and Data Modeling
Chad Mitchell
Blackbaud Analytics
July 18, 2015
ANALYTIC
S E R V I C E S
Game Plan
•
•
•
•
•
•
•
Definitions, Overview and Why?
Data Mining vs. Data Modeling
In-house Solutions
Outsourcing Options
Examples and Cast Studies
Benefits and Risks
Q and A
ANALYTIC
S E R V I C E S
2
Background – Chad Mitchell
• Iowa State University
– Annual Phone-A-Thon
– Alumni Association Ambassador
– Major Gifts and Special Event Ambassador
• Experian
– Data Modeling and Demographic Data
– Blackbaud – Develop Prospect Screening Service
• Blackbaud Analytics
– 250 Clients
ANALYTIC
S E R V I C E S
3
Definitions
• Data Mining: Investigating and discovering
trends within a constituent database using
computer or manual search methods
• Data Modeling (Advanced Statistical
Analysis) : Discovery of underlying
meaningful relationships and patterns from
historical and current information within a
database; using these findings to predict
individual behavior
ANALYTIC
S E R V I C E S
4
Specific Applications of Data Modeling
• Determine subsets of similar individuals from
a larger universe
• Segment by characteristics
– Interests, finances, location, etc.
• Target marketing
• Predicting future behavior
ANALYTIC
S E R V I C E S
5
Why Use It?
• Classify donors & prospects by factors other
than wealth (or major gift potential):
–
–
–
–
–
–
Lifestyle/life-stage
Affinity
Interests/behaviors
Cultural
Demographics
Psychographics
ANALYTIC
S E R V I C E S
6
Go Beyond Capacity
Annual
Giving
Minimal
Investment
Major
Giving
Wealth Screening
Cultivate
Results
CAPACITY
ANALYTIC
S E R V I C E S
7
Benefits of Data Modeling
•
•
•
•
•
•
•
Reduce solicitation costs
Increase Response Rates
Understand donor/non-donors characteristics
Create cost-effective appeals
Increase gift revenues
Staffing and resource allocation
Turn knowledge into results
ANALYTIC
S E R V I C E S
8
Why Me? … New Roles for
Researchers!
• Prospect research is more than prospect
identification
• Leadership role of research
– Introduce new analytical/evaluation tools
– Results oriented change
– Giving is more than major gifts
ANALYTIC
S E R V I C E S
9
What Are My Options?
• Do It Yourself
– Simple statistics – Data Mining
– In-house Data Modeling
• Outsourcing
– Advanced Data Modeling
– Regression Analysis
– Consulting
ANALYTIC
S E R V I C E S
10
Simple Statistics
• What is simple?
– Frequency distributions
– Trend analysis
– Segmentation analysis
• Tools
– Existing Donor Management Application
– Microsoft Excel or Access
ANALYTIC
S E R V I C E S
11
Simple Data Mining - Examples
• Time of year giving
– Application: anniversary date solicitation
• Giving by solicitation type
– Application: segmented solicitations
• Geographic Analysis
– Application: special event and trip planning
ANALYTIC
S E R V I C E S
12
Anniversary Date Solicitations
• Objective: reduce solicitations to loyal donors
• Methodology: identify loyal donors with time
consistent giving patterns
– Contact donors at appropriate renewal time
– Mail or call these donors less frequently
– Increase value of their gifts
ANALYTIC
S E R V I C E S
13
Segmented Solicitations
• Objective: Increase solicitation effectiveness by
using ‘asking’ method appropriate to donor
• Methodology: Factor analysis
– Identify common characteristics of those who
give by phone, by mail, etc.
– Target groups sharing those characteristics
– Eliminate ineffective solicitations
ANALYTIC
S E R V I C E S
14
Special Event Planning
ANALYTIC
S E R V I C E S
15
Analyze Every Area of Giving
• Annual Giving
– Frequency at lower levels, highest propensity
– Most important donor segment
• Major Giving
– Determine an appropriate ask amount
– Maximize potential of each donor
• Planned Giving
– Frequency of giving – 10+ years
– No Major Gift giving history
ANALYTIC
S E R V I C E S
16
Case Study – Higher Education
Two similar organizations with vastly different
profiles
100%
Past Giving
80
Gender
60
Upscale
Credit Card
40
Wealthy Zip Code
20
Graduation Year
Demographic
Indicator
Past Giving
Alumni Member
Campus Leader
0
University A
ANALYTIC
S E R V I C E S
University B
17
Data Modeling –
How Do You Do It?
•
•
•
•
•
Challenge yourself
Identify the behavior to be predicted
– for example, annual giving likelihood
Identify variables to be used
Create a file (random sample)
– validate fields to be used
Utilize statistical software package
–
–
ANALYTIC
SPSS
SAS
S E R V I C E S
18
Types of Data Modeling
•
•
•
•
•
Clustering
Decision Trees (CHAID)
Neural Networks
Logistical Regression
Probit Regression
ANALYTIC
S E R V I C E S
19
How To (continued)
• Split the file in half at random
– modeling sample
– holdout sample
• Build model
• Apply algorithm to holdout sample
• Test the model
• Score the database
• Implement the model
ANALYTIC
S E R V I C E S
20
Yes, There Are Risks
•
•
•
•
•
Bad or misleading data
Off the shelf modeling programs
Time intensive
Test, test, test
Applying Generic models
– PRIZM, P$CYLE and MOSAIC
ANALYTIC
S E R V I C E S
21
Acceptable Risk
• Potentially rich data in your file
• Understanding the big picture
• Bringing focus to your development efforts
ANALYTIC
S E R V I C E S
22
Levels of Information
•
•
•
•
•
Individual
Household
ZIP + 4
Block
ZIP
Tip: start at smallest level possible - individual
ANALYTIC
S E R V I C E S
23
Types of Data
• Types of Client Data
– Demographic
– Giving History
– Activities/Relationships
– Transactional
– Attitudinal
– Interests
ANALYTIC
S E R V I C E S
24
Types of Data
• Sources of External Data
– Demographic/Census
– Single source databases credit
– Consumer transactional
– Aggregated (avoid
aggregated age)
– Cluster
ANALYTIC
S E R V I C E S
• Vendors
–
–
–
–
–
Experian
Acxiom
InfoUSA
D&B
KnowledgeBase
Marketing
– List Brokers
25
Creating Variables
•
•
•
•
Additive
Dichotomous (yes/no)
Continuous/quadratic
Composite variables
– State/city
• Missing data
ANALYTIC
S E R V I C E S
26
Maximizing Your Data
ANALYTIC
S E R V I C E S
27
Blending Data into Models
Appended Data
Client Data
Determine best candidate variables
for modeling process; create new
Composite and dummy variables
Final Unique
Identify best
models and test
results
ANALYTIC
Identify attributes with the
greatest explanatory value;
select and weigh data in
unique algorithm
S E R V I C E S
Algorithm(s)
28
Case Study – Family / Human Services
• Challenge
– Decrease direct mail
expense while increasing
annual contributions
• Before BBA
– Pieces mailed = 1,200,000
– Total No. of Gifts = 3,000
– Contributions = $300,000
• After BBA
– Pieces mailed = 200,000
– Total No. of Gifts = 10,000
– Contributions = $1,200,000
• ROI
– Contributions = 398%
ANALYTIC
S E R V I C E S
29
Outsourcing – Why?
•
•
•
•
•
Models specific to your donors and prospects
Speed
Cost
Accuracy
Consulting
ANALYTIC
S E R V I C E S
30
Vendor Qualification
• Methodology and Philosophy
• Experience
–
–
–
–
Number of clients
Personnel – Ph.D. Level Statisticians
References
Case Studies
• Integration with Existing Software
• Broad Range
• Deliverables, Follow-up and Consulting
ANALYTIC
S E R V I C E S
31
Outsourcing Examples
Every donor…
Annual Giving Propensity
478
0
1000
Major Giving Propensity
849
0
1000
Planned Giving Propensity
250
0
1000
Cash Capacity for Org in 12-mo. Period
$5,000-10,000
ANALYTIC
S E R V I C E S
32
Annual Giving Model
ANALYTIC
S E R V I C E S
33
Visualize Your Database
ANALYTIC
S E R V I C E S
34
Chart Your Ask Amounts
ANALYTIC
S E R V I C E S
35
Summary
• Data Mining vs. Data Modeling
• In-house vs. Outsourced Solutions
• Risks and Benefits
ANALYTIC
S E R V I C E S
36
Contact Information
• Chad Mitchell
–
–
–
–
–
–
–
Account Executive
Blackbaud Analytics
(800) 468-8996 x.5854 Toll-free
(404) 888-9353 Direct
(843) 216-6100 Fax
[email protected]
www.blackbaud.com
ANALYTIC
S E R V I C E S
37