WealthEngine (WE) models can help you unearth new donors; gain a greater understanding of donor motivations and behaviors; and help you drive personalized campaigns – but when it comes down to it, what is the system doing? What does that mean?

The term “modeling” gets thrown around all the time, but if you ask a given person to describe exactly what a model is, even professionals often draw a blank.

In this article, we’ll answer those questions!

*Note: for information on the specific modeling packages WE offers, see our other modeling resources.*

This article will cover:

- Modeling is math
- Where does the formula come from?
- Descriptive vs Predictive

**Modeling is math**

Let’s strip away the business lingo and the marketing speak. When it really comes down to it, modeling is an equation:

A + B x C – D = E

That mathematical process is happening entirely on the back-end of our system; you are simply reaping the benefits. WealthEngine will take quantitative elements, plug them into a formula, and provide you with a simple modeled score that measures how valuable a specific individual is to your fundraising efforts.

For example, let’s say that Jane Smith was someone that you screened, and now you are comparing her to a model.

- WealthEngine knows that Jane:
- Has visited the hospital 5 times in the last 3 years
- Is married
- Has 2 minor children in the household
- Has given $5000+ in political donations
- Has green eyes
- Has a degree in English Literature

Each of these qualities are deliberately ordered and weighted within the model’s formula (A + B x C – D = E) so as to prioritize the most statistically significant. Each attribute needs to be entered into that formula, which means that each one is given a numeric value. So, when the system saw that Jane was married and had a degree in English Literature, it used that model’s point system to calculate her score:

- Marriage status:
- unmarried = 1 point; married = 2 points.

- College Degree:
- no degree = 0 points; Mathematics = 1 point; English = 2 points; Economics = 3 points, etc.

The system then combines those categorical variables (attribute and statistical value) to rate Jane. Once Jane’s information is plugged into the entire, lengthy equation, WealthEngine concludes that her modeled score is 740, out of a maximum 1000. In other words, it has taken subjective qualities and turned them into an objective calculation! The result then provides a sophisticated measure for you to determine who you should be prioritizing and how they might fit into your current/future campaign.

So again - modeling is math! You won’t be seeing any of this, but it is the process that gets you the final score.

*Note: This is the core concept of modeling. There are however many different models you might use. To learn more, check out our other modeling resources!*

**Where does the formula come from?**

Different models will have different formulas.

Some types, such as the Look-Alike Models, use enormously complex, pre-built formulas that WealthEngine has leveraged our 20 years of pioneering engagement science to design. These formulas will be applied across the board. This means that if you were to build five different Look-Alike Models, they would all use that same formula.

On the other hand, you can also work with our data scientists to develop a formula that is specifically customized to your needs and constituents. Our team will conduct a statistical analysis of your database to determine which variables are significant in predicting who is most likely to, for example, contribute a major gift. By determining how these unique donor traits influence each other, WE is then able to create a brand-new formula that will more accurately indicate which individuals should be your focus.

*Note: If you are interested in working with WE to create a custom model, our Designing a Custom Model Guide will walk you through what that process will look like.*

**Descriptive vs Predictive:**

There are two key categories of models: descriptive and predictive. Both kinds of models will allow you to gain greater insights into what makes your donors unique and to identify promising new leads - but there are key differences:

- Descriptive modeling
- Based on a pre-built algorithm.
- Jane Smith looks like your best donors. She may or may not behave like them, however.

- Predictive modeling
- Based on a custom-built algorithm.
- Jane Smith behaves like your best donors – even if she doesn’t look exactly like them!

In other words, predictive modeling predicts behavior, while descriptive modeling merely describes an individual or group of individuals. That description in and of itself could be profoundly telling, but it does not convey the sophistication and confidence of a predictive model.

*Note: To learn more about the difference and to see a breakdown of which models fall into which category, check out our Descriptive vs Predictive Guide.*