What is data mining?

 

"The science of extracting useful information from large data sets or databases"

 David Hand, Heikki Mannila & Padhraic Smyth, Principles of Data Mining

"Non-trivial extraction of implicit, previously unknown and potentially useful information from data"

Vipin Kumar, Michael Steinbach, Pang-Ning Tan, Introduction to Data Mining

There are many definitions of data mining, but basically it is the process of extracting or uncovering hidden patterns in data. As far as fundraising is concerned, data mining works by uncovering and combining hidden markers which give clues or hints to a person's ability to make a gift.

 

So what are these hidden markers?

 

Here is one example. The data mining expert Peter Wylie conducted some research into lifetime giving based upon the length of the donor's first name. Donors on a fundraising database were grouped into pairs by formal name and informal name.

1. William versus Bill
2. Robert versus Bob
3. Richard versus Dick
4. Edward versus Ed
5. Kenneth versus Ken
6. Michael versus Mike
7. Ronald versus Ron
8. Thomas versus Tom
9. Donald versus Don
10. John versus Jack
11. James versus Jim
12. Raymond versus Ray

As can be seen below, in 11 out of the 12 pairs (Raymond versus Ray was the only exception), the formal names gave considerably more than the informal names. The probability of this having occurred by chance (if there were no true relationship between first name length and giving for this population of donors) is less than one in four thousand.

 

 

What are the other hidden markers?

 

Research by Wylie and others has suggested the following fields:

 

Age, home phone number, business phone number, marital status, Email, donation history.

 

How do we use these fields to extract useful data?

 

Each field must be assigned a number depending upon the information it contains.

 

First name                                Formal, score 1; informal, score 0

Age                                          Above 50, score 1; Below 50, score 0

Home phone number                 Listed, score 1, not listed, score 0

Business phone number            Listed, score 1, not listed, score 0

Marital status                           Not single, score 1; single or not listed, score 0

Email                                       Listed, score 1, not listed, score 0

Spouse/family details                Listed, score 1, not listed, score 0

 

The higher the score, the more likely the person is give a regular donation and/or a large donation.

 

This can be made more specific to major donors by adding the following:

 

Post code prefix*               EH4, AB15, NW3, etc. score 1, other postcodes, score 0

Largest donation               Greater than £100, score 1; less than £100, score 0.

 

Some of these markers are better indicators of wealth than others but by combining many markers together, one can get a very good indication of a donor's likely ability to make a major donation.

 

How does it work? It's not exactly clear, but the good thing is: you don't need to know.

 

Supporting Advancement list a comprehensive selection of data mining resources here.

 

Prospect-dmm is the discussion group for development professionals involved or interested in data mining and modeling, particularly as such concepts may be applied to major gifts.

  

You may also be interested in Data Mining for Fund Raisers by Peter Wylie and Fundraising Analytics: Using Data to Guide Strategy by Joshua Birkholz.

 

 

* Why did I choose these postcodes? They are the first postcodes on Eurodirect's Millionaires List.

 

 

Page Updated: 24/04/08