Transcript Fundraising Intelligence - Institute of Fundraising

Fundraising Intelligence:
Data Mining & Analytics
RIF Scotland
3rd October, 2011
Marcelle Jansen, WealthEngine
Agenda
• Definitions
• Data Mining
• Bringing Analytics to your
Organization
• Harnessing the Power of Data
through Analytics
– A Case Study
2
Definitions
• Data Mining: “the extraction of meaningful patterns
of information from databases”
• Analytics: “how an entity arrives at an optimal or
realistic decision based on existing data”
• Predictive Modeling: “the process by which a
model is created or chosen to try to best predict the
probability of an outcome”
3
The Goal: Fundraising Intelligence
“Fundraising Intelligence can be described as the
process of gathering data, turning it into actionable
information through analysis, and making it
accessible to the right people, at the right time, to
support fact-based decision making.”
4
“In God we trust. All others
bring data.”
-- Barry Beracha
Sara Lee Bakery retired CEO
5
Agenda
• Definitions
• Data Mining
• Bringing Analytics to your
Organization
• Harnessing the Power of Data
through Analytics
– A Case Study
6
Data Mining
• What data is important?
• What type of data should we collect?
• Where are the sources for data?
7
Key Considerations with Data
• Rules to Follow:
– Clean
– Consistent
– Structured
– Codified
• GIGO: Garbage In Garbage Out
8
Data: Emirates Airlines Flight 407
• Context: Airbus A340-500
– Flight from Melbourne, Australia to Dubai on
March 20, 2009
– Flight included 257 passengers and crew of 18
• Events
–
–
–
–
Aircraft not accelerating normally
Traveled the length of the runway (more than 2 miles) still unable to lift-off
Plane’s tail struck the ground at least 5 times before lift-off
Clipped a strobe light and flattened a navigation antenna as it struggled to gain
altitude
• Reaction
– “This would have been the worst civil air disaster in Australia’s history by a very large
margin” Ben Sandilands, Aviation Expert
– “They were lucky that…a lot of people (didn’t lose) their lives” Dick Smith, former
head of the Civil Aviation Safety Authority
What Happen on Flight 407!?!
9
Data: Emirates Airlines Flight 407
• Flight 407 had four experienced pilots in the cockpit
– The captain and first officer completed a preflight checklist
including a four-part process cross check
• Data entry of plane’s calculated weight: 262 metric
tons
• Flight 407’s actual calculated weight: 362 metric tons
– This is the equivalent of not calculating the weight of 20
African elephants stored in the belly of the plane
• The Australia headline said it all:
10
“The Devil is in the Data”
‘The Devil is in the Data” The Australian, September 12, 2009
11
The Devil is in the Data: Types of Data
• Financial
• Biographical
• Philanthropic
• Behavioral
• Other………..
12
The Devil is in the Data: Sources of Data
• Internal
• Volunteer Information
• Research Information
• Electronic Screening
13
Data, Data and More Data
• Giving History
– First Gift Date / First Gift Amount
– Last Gift Date / Last Gift Amount
– Total Giving / Total # of gifts
– Largest Gift Amount / Largest
Gift Date
– Average gift (annual vs. major)
– Other factors?
14
Data, Data and More Data
• Relational
– Family Ties (legacy alums, multiple family members with
affiliations)
– Alumni Association or Member
– Volunteer Roles
– Connection to an organization insider
– Product Purchases
15
Data, Data and More Data
• Biographical
– Age
– Marital Status
– Gender
– Business Title
– Email address
– Business/home phone
– Others?
16
Data, Data and More Data
• Contact
– Last Staff Contact
– Event Attendance
– Last Solicitation
– Amount of Last Solicitation
– # of Contacts Overall
– # of Contacts in last 3 years, 5 years
– Others
17
Electronic Screening: Making It Work
• What should screening my data accomplish?
– Identify new prospects
– Qualify existing prospects
– Prioritize existing prospect pool
– Segment prospects into solicitation pools
18
Electronic Screening Data
• How do I select the right type of
screening for my organization?
– Determine your organizations needs
– Do you need to screen your entire database or
does it make more sense to screen a targeted
sample
– Do you want hard asset data?
– Or demographic data?
19
Electronic Screening Data
• Types of Data Returned
– Geo-demographic Data
– Hard Asset Data
– Wealth Indicators
20
Electronic Screening Data: Results
• Capacity Ratings
• Propensity to Give Ratings/Indicators
• Financial Information
– Income
– Real Estate
– Stock Holdings
• Gifts to Others
• Age/Children
• Household Interests
21
Agenda
• Definitions
• Data Mining
• Bringing Analytics to your
Organization
• Harnessing the Power of Data
through Analytics
– A Case Study
22
Data Analysis vs. Statistical Modeling
Data
Analysis
23
Statistical
Modeling
Definition
• Analysis of specific business
questions and the development
of foundational insights that feed
into statistical modeling
• Building statistical models to
predict desired behaviors
Techniques
• Hypothesis Based Approach
(vs. Boiling the Ocean)
• Univariate & Bivariate Analysis
• Multivariate Analysis
• Linear/Logistic Regression,
Cluster Analysis, etc
Tools
• MS Excel most commonly used
• SAS, SPSS are most popular
Modeling: What Do You Want To Accomplish
– Major Gift Model
– Annual Fund Modeling
– Planned Giving Model
24
Model Variables
Dependent Variables
• Overall likelihood of giving a gift
• Likelihood of giving a gift over $X
• Likelihood of being a Major Donor
• Likelihood of upgrading a gift
• Next Gift Amount
• Lifetime Giving Amount
• Next Ask Amount
Independent Variables
• Giving History
- RFM* & Trend Attributes
• Constituent Type
- Parents vs. Alumni
• Wealth Indicators
- Capacity Ratings & Wealth
Components
• Demographics
- Age, Marital Status, Education
• Contact Info
- Phone number, email address
*RFM corresponds to Recency, Frequency, Monetary Value
25
Agenda
• Definitions
• Data Mining
• Bringing Analytics to your
Organization
• Harnessing the Power of
Data through Analytics
– A Case Study
26
A Non-Profit Corporation Illustration
Context – A Public Radio Station
• Understand the profile of the organization’s client base
• Rank order the client base on desired behavior using statistical models
• Determine criteria to identify best prospects in the broader universe
Available Data
• Organization data– Biographical, Giving History and Relational
• Screening data – Wealth Attributes, Geo-demographic, and
Philanthropic
Objectives
• Profiling analysis identified predictors of the desired behavior - Major Gifts
greater than $250
• Rank order the client base
• Profiling insights were used to develop a custom prospect identification
strategy
27
Dependent Variable Illustration
Distribution by Largest Gift Amount
60%
•
Selected ‘Largest Gift of at least
$250’ as the dependent variable
•
6,149 donors met this threshold
(incidence of 15% in the sample of
41,759 records)
•
Identified predictive attributes by
analyzing across giving, wealth
and demographic variables
•
Metrics used for analysis
- Incidence = (# of
constituents with largest gift
of at least $250)/(total
number of constituents)
- Distribution = % of total
constituents in a segment
% of Total Donors
50%
40%
Dependent Variable
Largest Gift >=$250
30%
20%
10%
0%
0-100
100-250 250-500
500-2K
Largest Gift Amount
28
2K+
Illustration: Number of gifts and years since first gift are
strong predictors of giving at least $250
By Number of Gifts
30%
By Years Since First Gift
50%
Distribution
45%
24%
23%
Incidence
25%
25%
Incidence
40%
45%
24%
40%
Distribution
20%
35%
18%
35%
20%
30%
30%
15%
25%
12%
15%
15%
16%
25%
20%
20%
10%
8%
10%
8%
15%
10%
5%
10%
5%
5%
5%
0%
0%
0-3 Gifts
29
15%
4-6 Gifts
7+ Gifts
0%
0%
Last 1.5 1.5-3 Years 3-5 Years 5-7 Years 7-9 Years 10+ Years
Years
Illustration: Political giving and number of contact data
points also strongly differentiate givers over $250
By Political Contribution Level
35%
33%
Incidence
30%
80%
By Number of Contact Methods
30%
90%
25%
70%
Distribution
70%
60%
25%
20%
21%
50%
40%
60%
Incidence
20%
Distribution
50%
12%
40%
15%
15%
30%
11%
10%
10%
30%
20%
20%
5%
5%
10%
0%
0%
0 Contributions
30
80%
25%
$1 to $2,000
>=$2,000
10%
0%
0%
0 or 1
2,3,4
Illustration: Property Value and Giving Capacity scores
slope giving behavior (>=250)
Property Value
40%
Incidence
Giving Capacity
37%
80%
30%
Distribution
35%
70%
Incidence
Distribution
70%
26%
60%
25%
30%
60%
50%
Incidence
20%
25%
19%
50%
40%
20%
40%
18%
15%
14%
30%
15%
30%
13%
10%
10%
20%
5%
10%
10%
20%
5%
0%
0%
<=$500K
31
<=$2MM
>$2MM
10%
0%
0%
$0 - $11K
$11K - $226K
$226K - $336K
$336K+
Statistical Modeling: Making it Work
• How will all this help me determine the
philanthropic characteristics of my
organization’s donor base?
– Allows you to determine which assets have the most impact
on organizational giving
– Allows you to become more efficient in selecting the best
prospects for your programs and organization
– Allows you to segment, build and grow your prospect pool to
focus on your best prospects
32
Summary
• Fundraising intelligence allows you to
optimize your data for:
-
maximum return on investment,
-
effective strategy development
-
efficient fundraising management
• It all starts with good data!
33
Questions?
Contact:
020 3318 4835
[email protected]
www.wealthengine.com