Introduction: Laurent Lo de Janvry



Similar documents
Predictive Modeling for Organizational Effectiveness

Today we ll talk about preparing and mining data to inform annual giving strategy

Identify & Engage: Analyzing your Donor Data to Discover and Prioritize Major Gift. Prospects. Oregon Nonprofit Leaders Conference, April 2014

Education. Lawrence Henze Elizabeth Crabtree. March 1, 2011

Modeling Lifetime Value in the Insurance Industry

Beyond Prospects: How Data Mining can Uncover Insights and Guide Program Decisions

Data Mining: An Overview of Methods and Technologies for Increasing Profits in Direct Marketing. C. Olivia Rud, VP, Fleet Bank

Creating a MultiChannel Marketing Plan & Timeline for Annual Fundraising

Young Alumni Giving Program Overview May 2007

Using Statistical Modeling to Increase

Using Your Fundraising Software to Effectively Manage Your Prospects

University Advancement Annual Giving. Program Review

The Pursuant Approach to Partnership

Lifetime Value A 360 Degree View of Donor Value

How Your Nonprofit Can Maximize Your Donor Database

How Organisations Are Using Data Mining Techniques To Gain a Competitive Advantage John Spooner SAS UK

Communications/publications Volunteer opportunities Events Awards PREAMBLE

Plan 2016 The Strategic Plan of University Advancement NC State University

Target Analytics Nonprofit Cooperative Database

If You Think Investing is Gambling, You re Doing it Wrong!

Experian TAPS SM Total Annual Plastic Spend

Reactivating Lapsed Donors: How to Use Loyalty and Philanthropic Segmentation to Optimize Donor Reactivation

A Basic Guide to Modeling Techniques for All Direct Marketing Challenges

By Peter Schoewe, Director of Analytics Mal Warwick Donordigital

Upending the Pyramid: Moving Donors to Mid Level Giving Kristin McCurry. If you want to change the world, change your MIND.

How to Get the Most Out of Your Fundraising Database. Robert Weiner

Building an Advancement Data Warehouse. created every year according to a study. Data Facilitates all Advancement Activities

Advanced Analytics for Call Center Operations

Free Trial - BIRT Analytics - IAAs

Data Mining in CRM & Direct Marketing. Jun Du The University of Western Ontario jdu43@uwo.ca

Reports and KPIs Guide

Embracing Technology for Moves Management Success

Relationship Management Policies and Procedures

An Overview of Predictive Analytics for Practitioners. Dean Abbott, Abbott Analytics

Data Mining Techniques Chapter 4: Data Mining Applications in Marketing and Customer Relationship Management

Nonprofit Fundraising Change in the Number of Companies

Data Mining Applications in Fund Raising

Steven R. DiSalvo, Ph.D.

Development Fundamentals: Cultivating Donors for Life. Kelly Pullin, Analytics Supervisor Kalina Cavallero, Senior Account Executive

Database Review Performed for Fine Arts Museum of Anytown Client ID: D10000 Performed by: Lori Wehnau Date of Review: 06/30/2009

Does Your Client s Current Life Portfolio Make the Cut?

Vital Signs Analysis 101

Simple Predictive Analytics Curtis Seare

Nonprofit Fundraising Study

Your presenter» With Steven Shattuck. What Every Fundraiser Can Do To Stop Falling Retention Rates

Introduction to Data Visualization

Query 4. Lesson Objectives 4. Review 5. Smart Query 5. Create a Smart Query 6. Create a Smart Query Definition from an Ad-hoc Query 9

Human Resources POSITION DESCRIPTION (HR 120)

Should I Stay or Should I Go? Determining if Convio s Common Ground is Right for Your Nonprofit

Predictive Modeling Techniques in Insurance

Online Survey of Employees Without Workplace Retirement Plans

What is Predictive Analytics?

University of California, Berkeley. John Wilton Vice Chancellor Administration & Finance

DIRECT MAIL THE POWER OF. Developing your direct mail program... For more information, contact:

Is There Future for Direct Mail Fundraising? Geoffrey Peters, CEO Moore DM Group

Leadership Appointments at the University of Georgia (UGA)

Neil Hayward Customer Intelligence Solutions Program Manager SAS EMEA Copyright 2003, SAS Institute Inc. All rights reserved.

Data Mining Applications in Higher Education

SWOT. SWOT for Fundraising. Internal. External. Strengths Weaknesses

Deloitte/Berkeley-Haas Partnership Update. October 26, 2015

Benchmarking Alumni Relations in Community Colleges

Product recommendations and promotions (couponing and discounts) Cross-sell and Upsell strategies

The Ariel Mutual Funds/Charles Schwab & Co., Inc. Black Investor Survey. Saving and Investing Among High Income African-American and White Americans

EXECUTIVE DIRECTOR OF PRINCIPAL GIFTS. UNIVERSITY OF MARYLAND College Park, MD

MISSING DATA TECHNIQUES WITH SAS. IDRE Statistical Consulting Group

DIRECT MAIL: MEASURING VOLATILITY WITH CRYSTAL BALL

Applied Data Mining Analysis: A Step-by-Step Introduction Using Real-World Data Sets

Role of Customer Response Models in Customer Solicitation Center s Direct Marketing Campaign

Maximize Revenues on your Customer Loyalty Program using Predictive Analytics

Direct Marketing Profit Model. Bruce Lund, Marketing Associates, Detroit, Michigan and Wilmington, Delaware

Using Analytics to Grow Your Fundraising Program

CONVIO LUMINATE Q&A. Summary: Luminate is comprised of two suites: What is new:

Insurance Analytics - analýza dat a prediktivní modelování v pojišťovnictví. Pavel Kříž. Seminář z aktuárských věd MFF 4.

Transcription:

-- Mining Your Data To Maximize Your Fundraising Potential Laurent (Lo) de Janvry UC Berkeley Haas School of Business CASE VII Tarak Shah UC Berkeley University Relations Introduction: Laurent Lo de Janvry BA in Economics, UC Berkeley MBA in Marketing/Strategy, USC Marshall School of Business Private Sector: 5+ Years Consulting, Marketing & Brand Strategy Arthur Andersen, Mars/KalKan Foods, DelMonte Foods, Prophet Brand Strategy Higher Education: 12+ Years 6-1/2 Years at UC Berkeley Director of Annual Giving - Strategic & Direct Marketing Services 4-1/2 Years at Berkeley-Haas School of Business Director of Development Annual, Reunion & Leadership Support Independent Consultant, ASquaredConsulting Development & Alumni Relations,, Market Research, Strategic Planning & Group Facilitation 1

Introduction: Tarak BS, UC Berkeley, Math/Philosophy 02 UC Berkeley University Relations ( 06-Present) Prospect Analyst, University Relations Report writing Campaign analytics uponinspection.wordpress.com (where I write about data mining and modeling) Strategic Marketing -- Going Beyond Your Institutional Intuitions & Assumptions Marketing Needs Marketing Tools Targeting Segmentation (Re)Positioning Branding Program Development Evaluation & Planning Data Mining & Modeling Market Research Benchmarking & Secondary Research Market Tests Measuring & Reporting 2

Strategic Marketing -- Going Beyond Your Institutional Intuitions & Assumptions Marketing Needs Marketing Tools Targeting Segmentation (Re)Positioning Branding Program Development Evaluation & Planning Data Mining & Modeling Market Research Benchmarking & Secondary Research Market Tests Measuring & Reporting -- Definitions What is data mining? 3

-- Definitions What is data mining? Data mining: process by which you sift through data to identify correlations between characteristic and desired behavior -- Definitions What is data mining? Data mining: process by which you sift through data to identify correlations between characteristic and desired behavior Desired Behavior Characteristic Correlation Annual Giving Association Membership +/- Major Giving 5+ Annual Gifts +/- App. Acceptance Catholic/Private HS +/- 4

-- Definitions What are predictive models? -- Definitions What are predictive models? Predictive models: mechanism to score constituents according to the likelihood of a desired behavior (i.e. giving, giving amount, membership, attendance, offer acceptance, etc.) 5

-- Business Objectives Why do you (we) want to develop donor models? -- Business Objectives Why do you (we) want to develop donor models? Targeting & Prioritizing Resources Determining Best Strategies Identifying New Target Markets 6

-- Development Objectives Targeting & Prioritizing Resources Marketing Strategies (Acquisition) Data Acquisition (Screening Services) Prospect Identification (Special Asks) Determining Best Strategy Hi Level Asks (Non-) Upgrade Strategies (Reunions) Identifying New Target Markets New Prospects (Acquisition/Reunions) -- Process & Application Who here has purchased or developed predictive models for their institution, and how did you do so? 7

-- Process & Application Who here has purchased or developed predictive models for their institution, and how did you do so? Hire Consultant vs. Do-It-Yourself? Acquire Third-Party Data vs. Your Data? Budget, time, tools, training, etc.? -- Process & Application How have you utilized the model(s) to modify your programs or marketing strategies? What was the outcome/result? 8

-- ABCs Cross Tabs to identify potential relationships and predictive vars (1/0) Donor Status x Membership Status Non- Totals Members 70% 30% 100% (40%) Non-Members 10% 90% 100% (60%) Totals 35% 65% 100% -- ABCs Cross Tabs to identify potential relationships and predictive vars (1/0) Donor Status x Membership Status Non- Totals Members 70% 30% 100% (40%) Non-Members 10% 90% 100% (60%) Totals 35% 65% 100% 9

-- ABCs Correlation Matrices or Linear Regressions to identify correlations and statistical significance Giving (1/0) Giving (1/0) Giving (1/0) Membership (1/0) Gender=Male (1/0) Proximity <50 miles (1/0) -- ABCs Correlation Matrices or Linear Regressions to identify correlations and statistical significance Desired Behavior Characteristic Correlation Annual Giving Association Membership +/- Gender: M +/- Proximity <50 Miles +/- T-Stat >1.67 >1.67 >1.67 Correlation Coefficient (or Beta): Weight of relationship or score T-Stat: Statistical significance of variable to help predict 10

-- ABCs Sometimes, transforming a variable will allow a relationship to present itself Giving Giving Age Age -- ABCs Sometimes, transforming a variable will allow a relationship to present itself Giving Giving Age (Negative) Square of Age 11

-- ABCs The characteristics should be independent of the desired behavior Desired Behavior Characteristic Correlation Annual Giving Donor s club member + + +! T-Stat >1.67 (Everyone who gave a gift was made a donor s club member) -- ABCs Linear Multi-Variable Regression Giving (1/0) = Score Constant Sum /Combo of Predictive Vars 12

-- ABCs Logarithmic Regressions or Binary Logit Regressions Giving (1/0) = Score Giving (1/0) = Score 1 Sum /Combo of Predictive Vars 0 Sum /Combo of Predictive Vars -- ABCs Model: Scoring Mechanism +5 Association Membership +3 Student Activities +3 Event Attendance +2 Family Relations +2 Zip Codes +2 Marital Status +1 E-Mail Address Presence +1 Degrees (Min 0; Max 19) 13

-- ABCs Constituents Model: Scoring Mechanism +5 Association Membership +3 Student Activities +3 Event Attendance +2 Family Relations +2 Zip Codes +2 Marital Status +1 E-Mail Address Presence +1 Degrees (Min 0; Max 19) -- ABCs Constituents Model: Scoring Mechanism +5 Association Membership +3 Student Activities +3 Event Attendance +2 Family Relations +2 Zip Codes +2 Marital Status +1 E-Mail Address Presence +1 Degrees (Min 0; Max 19) Scores 0 0 1 8 8 10 15 19 14

-- ABCs Constituents Model: Scoring Mechanism +5 Association Membership +3 Student Activities +3 Event Attendance +2 Family Relations +2 Zip Codes +2 Marital Status +1 E-Mail Address Presence +1 Degrees (Min 0; Max 19) Scores 0 0 1 8 8 10 15 19 -- ABCs Beware of over-fitting the data Finding People with the middle initial P are twice as likely to make a major gift Disposition Suspicious Set aside a clean set of data that is NOT used to create the model for testing 15

-- ABCs You may end up multiple different versions of a model, and need to select the best one to move forward with. The question you re trying to answer will help define what we mean by best -- ABCs For instance, sometimes it s more important to correctly predict yesses than it is to predict noes (or vice versa) Model 1 87% correct overall Model 2 74% correct overall MG Non MG Predicted MG 70 50 Predicted Non-MG 30 450 MG Non MG Predicted MG 95 150 Predicted Non-MG 5 350 16

-- Examples University of Melbourne Annual Giving Model Special Giving Model University of California, Berkeley Annual Giving Model Targeting Acquisition Efforts Major (Special) Giving Model Prioritizing Data Screening Resources Identifying Prospects (Beyond WOM) University of Melbourne -- Annual Giving Model +5 * 'Memb Affil 1/0 +3 * 'Al Act 1/0 +3 * 'Al Int 1/0 +3 * 'Res College 1/0 +3 * 'Affil 1/0 +2 * 'Male 1/0 +2 * 'Deg Stat B 1/0 +2 * 'Mar Status F-M-P-W 1/0 +2 * 'B E-Mail 1/0 +2 * 'H E-Mail 1/0 +2 * 'Sports 1/0 +1 * 'Family Rel 1/0 +1* 'Atnd Al Dnr 1/0 +1 * 'Atnd Tg 1/0 +1* 'Other Event 1/0 +3 * 'U Deg Med 1/0 +2 * 'U Deg Vet 1/0 +1* 'U Deg Eng 1/0 +1* 'G Deg Eng 1/0 +1* 'G Deg Med 1/0 +1* 'G Deg Sci 1/0 +1* 'G Deg Vet 1/0 [Min 0 / Max 43 (Act 0-33)] 17

University of Melbourne -- Annual Giving Model Score Count %Tot % %Nons Mean Median 0-3 4,724 23% 5% 42% $20 $0 4-6 5,292 26% 16% 37% $85 $0 7-13 5,436 27% 37% 17% $252 $25 13+ 4,753 24% 42% 4% $1,795 $98 Total 20,205 100% 100% 100% -- Business Objectives 50% 42% 37% % of Total Non- 0% 17% 4% 0-3 4-6 7-13 13+ Score 18

-- Business Objectives $2,000 $1,795 Avg. Gift Among $0 $20 $85 $252 0-3 4-6 7-13 13+ Score University of Melbourne -- Special Giving Model ($500+) +5 * $50+ First Gift +4 * Affiliations +3 * Atnd Tg +2 * 3+ Num Sum Gifts +2 * 50+ Years Old +2 * Alumni Activity +2 * Family Relations +2 * Business E-Mail +2 * Sports +2 * Resident College +2 * Marital Status +2 * Male +2 * Atnd Al Donor Event +2 * Other Events +2 * U-Faculty Law/Arch +1 * G-Faculty MBS, Law & Arch +2 * Country MY +1 * Country HK, SG & TH +1 * State NSW & MSL +2 * Zip 3000 & 3141-3147 +1 * Zips (8) 19

University of Melbourne -- Special Giving Model ($500+) Score Count %Tot %$500+ %<$500 Mean Median 0-4 2,304 23% 1% 24% $137 $20 5-7 2,088 22% 8% 21% $186 $51 8-10 2,279 20% 15% 23% $296 $100 11-13 1,903 19% 26% 18% $637 $200 14+ 1,631 16% 50% 14% $4,815 $375 Total 10,205 100% 6% 94% -- Process & Application Annual Gift ($) High Level Low Level Non- Low Score High 20

-- Process & Application Annual Giving ($) High Level Cash Cows (Special Prospect) Low Level Non- Low Score High -- Process & Application Annual Giving ($) High Level Cash Cows (Special Prospect) Low Level Retain & Upgrade (Giving Society) Non- Low Score High 21

-- Process & Application Annual Giving ($) High Level Cash Cows (Special Prospect) Low Level Retain & Upgrade (Giving Society) Non- Targeted Strategy (Higher Ask) Low Score High -- Process & Application Annual Giving ($) High Level Cash Cows (Underserved?) Cash Cows (Special Prospect) Low Level Retain & Upgrade (Giving Society) Non- Targeted Strategy (Higher Ask) Low Score High 22

-- Process & Application Annual Giving ($) High Level Low Level Non- Cash Cows (Underserved?) Cut Down and/or Retention Program (Sustainer/EFT) Cash Cows (Special Prospect) Retain & Upgrade (Giving Society) Targeted Strategy (Higher Ask) Low Score High -- Process & Application Annual Giving ($) High Level Low Level Non- Cash Cows (Underserved?) Cut Down and/or Retention Program (Sustainer/EFT) Eliminate and/or Targeted Strategy (Acquisition) Cash Cows (Special Prospect) Retain & Upgrade (Giving Society) Targeted Strategy (Higher Ask) Low Score High 23

UC Berkeley -- Examples Annual Giving Model: The Cal Fund Prioritizing acquisition efforts Major Gift Model Prioritizing screening investments Prioritizing personal-volunteer solicitation efforts UC Berkeley -- Annual Giving Model Business need: Target limited resources to increase ROI of mass marketing programs Analyzed a random sample of 10,000 donors to Cal Fund (our annual fund) in previous fiscal year Determined variables that predict giving to Cal Fund & developed scoring model 24

UC Berkeley -- Annual Giving Model +3 for alumni who are lifetime members of the CAA +2 for alumni with both undergraduate & graduate degrees from Cal +1 for alumni with undergrad degree only +1 for alumni with bus. phone in database +1 for alumni with an e-mail in database +1 for alumni with Mrs. stated as a prefix +2 for alumni with Cal activities listed +1 for alumni with Dr. stated as a prefix +2 for alumni with Cal children +1 for alumni with an interest listed in db +2 for alumni with Cal spouse +1 for alumni in San Mateo & Santa Clara +2 for alumni with current annual +1 for alumni from Col. of Letters & membership of CAA Science +2 for alumni who have given to +1 for alumni with marital status married campus, excluding Cal Fund +1 for alumni with marital status divorced +1 for alumni with lapsed membership of the CAA +1 for alumni with marital status widowed +1 for alumni with Cal parents UC Berkeley -- Annual Giving Model Reach: Maximum Mailing Population 250,000 231,000 200,000 193,582 183,366 150,000 136,176 100,000 92,391 50,000 - FY01 FY02 FY03 FY04 FY05 25

UC Berkeley -- Annual Giving Model Response Rate 4.5% 4.0% 3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% 4.2% 3.1% 2.6% 2.2% 2.1% FY01 FY02 FY03 FY04 FY05 UC Berkeley -- Annual Giving Model Revenue FYTD $1,200,000 $1,000,000 $800,000 $1,084,769 $747,669 $838,108 $870,536 $947,251 $600,000 $400,000 $200,000 $- FY01 FY02 FY03 FY04 FY05 26

UC Berkeley -- Annual Giving Model Average Gift $250 $200 $150 $181 $172 $213 $226 $228 $100 $50 $- FY01 FY02 FY03 FY04 FY05 UC Berkeley -- Annual Giving Model Net Return on Investment $7.00 $6.00 $6.07 $5.00 $4.00 $4.07 $4.22 $3.00 $2.00 $1.00 $2.54 $2.16 $- FY01 FY02 FY03 FY04 FY05 27

UC Berkeley -- Annual Giving Model 7,000 6,000 5,000 4,000 3,000 2,000 1,000-5,995 4,344 3,932 3,852 4,156 FY01 FY02 FY03 FY04 FY05 UC Berkeley -- Special-Major Giving Model Business need: Prioritize special-major gift efforts personal-volunteer prospect pool during reunion years -- to maximize staff time & ROI Also: rank unqualified prospects Prospecting Cold call lists Database screenings 28

UC Berkeley -- Special-Major Giving Model 2662 $50K+ donors 10,000 random sample donors < $50K (5,140) and non-donors (4,860) 242 data points downloaded from donor database (for prospect and spouse) Split into Test and Control samples Used statistical package (Data Desk) to analyze variables that predict major giving UC Berkeley -- Special-Major Giving Model Prospect has given 10 or more gifts Prospect or spouse has at least one contact recorded in database Prospect or spouse has attended at least one event Prospect or spouse has been rated Prospect or spouse's first gift amount was greater than $100 Prospect's employer is listed in database Prospect's business zip code is listed in database Prospect's business telephone is listed in database Prospect or spouse has at least one affiliation listed in database Prospect is aged 50 years or more Spouse's birth date is listed in database Prospect made their first gift to Cal 25 years or more ago, or their spouse made their first gift to Cal five or more years ago Prospect is/was a member of a campaign committee or volunteer 29

UC Berkeley -- Special-Major Giving Model Percentage of Class Reunion alumni accepting kick-off event invitation by Major Gift Score 8% 7% 6% Percentage 5% 4% 3% 2% 1% 0% 7.14% 5.39% 2.78% 1.68% 1.69% 0.00% 0-17 18-35 36-53 54-71 72-89 90-107 Major Gift Score -- Innovations External Data Sources Wealth indicator/screening Assets Investments Length of time as current address Philanthropic giving 30

-- Innovations Event Attendance Online Alumni Community Membership & Usage Electronic Newsletter Readership Facebook/LinkedIn??? -- Innovations Use of visualization To uncover relationships To communicate results Other types of models Relationships, text-mining, social media, and beyond... 31

-- Warnings!!! Understand Your Business Objectives (And Their Inherit Tradeoffs) Targeting resources (ROI) vs. investing in growth (donor base) Know (How To Apply) Your Model Predicting the future using the past New Markets Market Saturation Don t eliminate your younger alumni even though they have low scores! -- Mining Your Data To Maximize Your Fundraising Potential Laurent (Lo) de Janvry UC Berkeley Haas School of Business Laurent_deJanvry@haas.berkeley.edu W 510.642.1224 M 510.589.5944 Tarak Shah UC Berkeley University Relations Tarak_shah@berkeley.edu 32