We recently had this question asked of us: How do you build a Prospect Model based on one prospect? Our President and Founder, Doug Cogswell, took a moment to answer this and give some ideas on how to build this kind of modeling with Advizor.
We build a lot of models for fundraisers. And we train teams on how to build effective models. We have a reference guide for this which you can grab at http://internal.advizorsolutions.com/Help/Analyst/ADVIZOR%20Modeling%20Workbook.pdf
The basic concept is:
- Pick a Target Group. This is the group that has the behavior you want examine
- Compare it to a Base Population. This is the larger group which has the potential to have the behavior you want. Some are in the Target Group, most are probably not
- Develop / synthesize / create a set of Explanatory Factors. These are things that you think might explain / show why the target is different than everybody else in the Base Population.
Then use the modeling tools to build a set of models. The modeling algorithms will determine how the different explanatory factors describe / influence the target, and can be used to predict which others in the current or future base population would be mostly likely to exhibit the behavior of the target.
So your question is basically what’s the smallest number that that can be in the target. And from a statistical perspective most would say 25. But ideally many more. That is because for the model to work there has to be enough members in the target to distribute across and populate the explanatory factors. So a target of 1 is not going to work in a predictive model. For example, say the one person is a large donor. And he/she lives in Greenwich, CT, has a Harvard MBA, works in investment banking, and has been to 15 reunions. But you have other large donors who live in other locations, have different degrees, and have done other things post-graduation that show engagement. If you look at just the one person all of these other attributes are going to be minimized and your model will not be very effective.
Now that being said, understanding the characteristics of one top donor is not a bad thing, it’s just going to build a very robust or useful model.