Transcript DonorCast - Minnesota Planned Giving Council

Predicting the Next Planned Gift
Josh Birkholz
Bentz Whaley Flessner
Bright Spots
1
Plan
 Setting
the stage
 Introducing
 How
Predictive Analytics.
is it accomplished?
 Just
One Statistical Principle:
Randomized Testing
 Bringing
2
Analytics In-House
Setting the stage
3
Fundraising Has Three Primary
Business Processes
Base Development
One-to-many strategies of engagement
Major/Planned Gift Development
One-to-one high ROI strategies
Prospect Development
Conversion from base to major
Prospect Development has Three Stages Feeding Major
and Planned Gift Cultivation
Market Research
Identification with screening and modeling
Prospect Research
Qualification with data
Field Research
Discovery / qualification
through interaction
Plan Strategy
Stewardship
Major Gift
Fundraising
Cycle
Solicitation
5
Cultivation
Effective Prospect Development for Planned Giving
 Identifies
-
prospects meeting the criteria planned gift donors.
Traditional characteristics
Characteristics unique to your organizations
 Works
with fundraisers to develop strategies for aligning the
prospects with the institution for a philanthropic partnership.
Characteristics
Assumptions
Observations
 Consistent
 Assumptions
 Old
donors
donors
 Donors
assets
with appreciated
generally
accurate for most institutions.
 Other
common characteristics
from our research:
 Legacy
families
 Multiple
property owners
 Employment
in education
and public service
 Donor
loyalty
 Positive
7
donor experience
How is Loyalty Achieved?
Needs
+
Needs met
consistently
=
Loyalty
Example: Positive Donor Experience
9
Introducing Predictive
Analytics
10
What Is Meant by “Analytics?”
Analytics describes the statistical
tools and strategies for:
 Analyzing
constituencies.
 Building
models to predict
constituent behaviors.
 Evaluating
program performance
using relevant metrics.
 Projecting
11
future program performance.
Analyzing Constituencies
 Identifying
 Defining
core constituent groups.
their characteristics.
 Understanding
their motivations.
Applications



12
Portfolio optimization.
Segmentation strategies.
Event programming.
Data Mining and Predictive Modeling:
What Is “Data Mining?”
Using statistics
to identify patterns
in data.
 Comparing
characteristics of
people or things doing a
behavior with people or things
not doing the behavior.
13
Data Mining and Predictive Modeling:
Predicting Behaviors from the Patterns
 Common
-
14
non-fundraising examples:
Credit ratings
Meteorology
Airport security
Modeling Can Predict Many Things
 Major,
planned, and annual giving
 Program
or department models.
(giving to fine arts, capital needs,
scholarships, patient care, etc.)
 Membership
likelihood
 Season ticket subscriptions
 Alumni affinity
 Channel preferences (mail,
phone, email)
 Next gift amounts
 Loyalty scoring with
precise weightings
15
Effective for Planned Giving:
Your constituents compared to
Your success stories using
Your data to identify
Your unique opportunity
16
How is it accomplished?
17
Method
18

Understand your goals before
you begin.

Gather your data. Included
demographics, giving, research,
and screening data.

Prepare the data for modeling.

Model.

Evaluate the results against
existing donors and prospects.

Score the file and implement
the results.
Common Score Format (Fractional ranking displayed)
All records have a ranking and a 0–1,000 score.
Planned Giving
Rank Label
Minimum
Maximum
4
500
1 Top 50%
500
750
2 Top 25%
750
900
3 Top 10%
900
950
4 Top 5%
950
975
5 Top 2.5%
975
990
6 Top 1%
990
995
7 Top 0.5%
995
997
8 Top 0.25%
998
999
9 Top 0.1%
999
1,000
0 Lower 50%
19
Planned Giving Score
20
PG Donor
Not PG Donor
9 Top 0.1%
8 Top 0.25%
7 Top 0.5%
6 Top 1%
5 Top 2.5%
4 Top 5%
3 Top 10%
2 Top 25%
1 Top 50%
0 Lower 50%
Percentage
Evaluate by Comparing Scores to Actual PG Donors
100
80
60
40
20
0
Categorize Variables From Output
Giving
Geography
Demographics
Management
Capacity
21
Sample of Possible Variables in Your Model
Category
Variable
Length of Giving Relationship
Giving
Frequency Index
Monthly Payment Preference
Capacity
Multiple Property Ownership
>100 miles from campus
Geography
Wisconsin (-)
55439 (+)
Event Attendance (+)
Management
Survey Response (+)
Alumni Volunteer(+)
Demographics
22
Education Job Title(+)
Single(+)
Opportunity:
Review Portfolio, Prioritize Direct Marketing Appeals
Planned Giving
Model Rank
23
Not Assigned a
Prospect Manager
Managed
0 Lower 50%
53,425
92
1 Top 50%
26,507
257
2 Top 25%
15,330
724
3 Top 10%
4,767
585
4 Top 5%
2,201
474
5 Top 2.5%
1,197
410
6 Top 1%
326
208
7 Top 0.5%
129
139
8 Top 0.25%
59
101
9 Top 0.1%
20
88
Examples: Successful Implementation
New planned giving director.
 Prepared new prospect list.
 Felt it was a “stacked deck.”

24

Program needed jump-start.

Purchased predictive models.

Aggressively marketed and
discovered new names.

Had best planned giving year
in history.
Just One Statistical Principle:
Random Testing
25
Drawing Planned Giving Donors Out of a Hat
 Imagine
a hat with 130 slips of paper.
 About
31% of the slips have the words
“planned giving donor” written on them.
 If
you draw a slip out of the hat,
approximately 1 in 3 will be a PG donor.
 For
most organizations, planned giving donors
represent a far lesser portion (<5%).
Can We Improve This Ratio?
We could survey our actual planned giving
donors asking:
 How
-
-
would you describe yourself?
A blue slip of paper
A green slip of paper
A yellow slip of paper
Survey Results
Now Which Slip of Paper Will You Select?
If You Choose Blue…
Will you draw a
planned giving
donor on average
1 out of 2 times?
The Answer: Unknown
 There
 You
30
is not enough information.
do not know the distribution of the random population.
Consider Your View
 Now, which slip will you select?
Total
Count
% of Total
PG Donors
% of PG
Donors
% of Color that
are PG Donors
Blue
60
46%
20
50%
33%
Green
60
46%
12
30%
20%
Yellow
10
8%
8
20%
80%
130
100%
40
100%
31%
Population
Total
33%
67%
1 in 3
1 in 5
4 in 5
Principle
 Common
characteristics may not be distinguishing
characteristics.
 How
populations are different (target vs. random) is more
interesting statistically and predictive than common characteristics
of a target group.
32
Bringing Data Mining
In-House
33
Bringing Data Mining In-House
 More
and more organizations
have in-house data mining
capacity, from large shops to
small shops.
 Large
shops generally have
dedicated staff.
 Small
shops have developed the
skill sets in research,
advancement services, or
annual giving.
34
Making the Case
 Gather
references of peers and aspirant peers.
 Build
a cross-functional project team.
 Start
with short-term projects—specific appeals.
-
Communicate goals before the project.
Communicate the success after the project.
 Educational
-
35
and research institutions:
Explore on-campus knowledge resources (economics, statistics,
business departments).
Explore on-campus software resources.
Statistics Software

SPSS
-
-

SAS
-

Very powerful for large data sets
Needed for regulatory testing
(not necessary in fundraising)
Good network of researchers using SAS
DataDesk
-
-
36
My personal preference
User friendly for expert and novice alike
Large network of other researchers using SPSS
Object-oriented format easy to understand
Excellent for exploratory analysis
Large network of other researchers using DataDesk
Training
 Software
training courses
 Conferences
and users groups
 Learning
through outsourcing (you
are buying methodology as well
as analysis)
 Onsite
consulting
 Campus
37
resources
Learn Through Outsourcing
Many organizations outsource their analytics;
benefits include:
 Expert
analysis.
 Opportunity
 High
38
to learn from their methodology.
level of service over the short term.
Developing In-House Capacities
 It
is not hard to learn.
 Analytics
is becoming part of the constituent relations and
admissions skill set.
 Nobody
 Ability
knows your data like you do.
to create multiple models and analysis—not to be
restricted by costs.
39
Final Thoughts
40
When You Leave Today, Remember:
 Start
with your bright spots.
 Build
a prospecting plan around
your characteristics.
 Consider
predictive analytics to
identify and prioritize your list.
 Comparing
PG donors to random
donors is more valuable than
summarizing common PG donor
characteristics.
 Whether
you outsource or build
analytics in-house, analytics is
within your reach.
41
Questions?
Joshua Birkholz
Principal, Bentz Whaley Flessner
Founder of DonorCast
89646:JMB:abl:050410.
7251 Ohms Lane Minneapolis, Minnesota 55439
ph: 952-921-0111 fax: 952-921-0109
[email protected] www.donorcast.com