As a WealthEngine client, you can purchase a number of different models. This article will provide a high-level overview of the different categories.
Note: For a deep understanding you should also check out our guide: What is Modeling? hyperlink
This article will cover:
- Why use a model?
- How it works
- Look-Alike Models
- Custom Models
- Enterprise Models
Why use a model?
The core of successful fundraising lies in targeting individuals you know will contribute to your organization. This begins with developing a keen understanding of your top donors. But, as you identify new people to engage, how do you know you’re reaching out to the right ones? And, how do you know that the people you’re reaching out to will give or spend just like your best?
Modeling will help you understand what makes your top donors unique – and then apply those insights to the rest of your constituents. In some cases, you can even work with WealthEngine (WE) data scientists to create a custom algorithm that is unique to you. Our most advanced models will not just help you find the people who resemble your top donors – they will predict, based on an analysis of your file and numerous customized attributes, who is most likely to behave in the way that you want.
In other words, WE models will utilize sophisticated computer learning to help take the guesswork out of fundraising for you.
Note: If you want a walk-through of how this might apply to your work specifically, please reach out to your Sales Representative.
How it works
First, you begin by asking a question:
- Who is likely to give me a major gift?
- What should my next ask amount be?
Then our data scientists, after analyzing a file that you have provided, will create a mathematical formula that is built to answer that question
Once you provide us a file, including a segment identifying your very best people, formulas to identify common traits among your donors. Different data points, or attributes, are then taken to create the perfect model. They have the capability to convert qualitative attributes into quantitative scores.
So, in order to predict this or deepen your understanding, our data scientists will conduct a statistical analysis to determine what qualities are statistically significant and (in some instances) predictive.
By determining how these unique donor or customer traits influence each other, they are then able to more accurately determine who would buy or give to your business or organization.
Note: To explore this further, see our guide: What is Modeling? hyperlink
This is the simplest and most commonly used type of modeling. It is the only kind that is fully integrated into the WealthEngine platform, meaning that you can easily build and analyze the model on your own, from within your account.
This feature may already be included in your subscription, right along with Searching, Screening, and Prospecting. To be sure, check to see whether “WE Analyze” is included in your contract.
The accessibility and speed of this model mean that it should be an integral part of any prospecting effort.
created to help you understand the unique qualities among your customers or donors (typically in comparison to the national average).
Look-alike modeling is descriptive, which means that it will tell you what individuals look like. You can identify individuals who closely resemble your top donors, but the system is not necessarily predicting that those people will actually give at that level. From there, the rest is up to you!
Look-Alike Modeling is descriptive, which means that it allows you to identify individuals who display similar characteristics to your best donors based on factors such as key wealth and giving scores as well as demographic and lifestyle data. This information can help you identify individuals who closely resemble your ideal existing donors, but the system is not necessarily predicting that those people will actually give at that level. From there, the rest is up to you!
Value/What you will learn from using this model: Within minutes, you can find new, promising prospects based on existing profiles in your database. This flexibility lets you build multiple models to look like any number of different segments of their donor or customer base.
Depending on the volume of people you are modeling, a look-alike model can be made in mere minutes. The greater the number, the more time it will take to create the model.
You will be able to rank people in a larger screening or prospect list based on how closely they resemble your best donors.
Note: any key sidenotes and/or related links to further dive into the above-stated details.
These are our standard (non-Look-Alike) models.
4-Pack Custom Modeling Suite:
- Major Gift Model
- Planned Giving Model
- Likelihood to Give Model
- Next Gift Amount Model
In this Modeling Suite, we’ll build models using up to 6 client attributes in addition to existing WealthEngine attributes (net worth, age, estimates giving capacity, total real estate value, etc.). Depending on what you are looking to predict, a model apart of this pack can be swapped out for either the Mid-Range Gift Model or the Sustainer Model.
Once we’ve received this information, we’ll be able to create these 4 models for you. You’ll gain insights on which prospects have the greatest potential to provide a major gift or planned gift; who is most likely to give in general; and how much an individual may give next.
Note: For more, see our Custom 4-Pack Model Guide. hyperlink
Clients provide us a file with donor and prospect Names, Addresses, past giving history, and additional client attributes they might have in their system. For example, let us say we have a University client then the client attributes could be:
- Alumni Flag
- Year of Graduation
- Club membership
- kids go to the same university
- College within the university
- Undergrad/Graduate degree
Features of Custom 4 pack models:
- So the models are built using WE attributes (total_networth, age, ECG, total_real_estate value, etc..) and up to 6 client attributes from the list of attributes provided by the client.
- Custom 4 pack is limited to the 4 models listed above.
In addition to the custom modeling suite, we also offer Enterprise Models, which are even more complex and robust than our standard custom models. Instead of selecting 6 unique attributes to customize the algorithm behind the model, you can send our data scientists as much data as you’d like. This will result in a fully-tailored analytic solution.
Essentially, you can hand our data scientists a problem statement, or specify what type of model you would like. This is to ensure that your experience is as personalized as possible. From there, our team will work closely with you to accomplish your targeted goal using the model of your choosing.
“We can build a model for anything if we have the necessary elements. Specifically, you can describe a specific behavior you want to predict and have a large enough sample “target” of people who have exhibited that behavior. That target needs to be large enough to predict that behavior
- For example, let’s say that you want to convert people who have a transactional relationship with you (e.g. they attend plays) into becoming donors. So long as you have 200-300 people who have already done that, we can build a customized model for that. WE will analyze those individuals, factoring in both the unique attributes you have selected and the traits that we find to be most statistically predictive to create that customized algorithm. That model will finally be compared against a larger screening and each person in that file will be given a new model score predicting how likely they are to take that same action.
In doing so, you will have an even more personalized experience and can work more closely with our team to accomplish your goal.
Under enterprise modeling, we have more relaxed modeling criteria. Mostly for-profit clients go for Enterprise modeling. The salient features of Enterprise modeling are as follows:
- The problem statement for the model is defined by the client.
Maserati came to us with a list of their current and past car customers. The modeling problem was to find the likelihood of a person buying one of the 4 Maserati models.
Gucci provided a list of customers in 2018 and wanted us to build a model which would identify people who look like customers who spent in the top 10% bracket in the year 2018.
- No limit on the number of clients attributes to be used as predictors in the model.
This is a key difference when compared to the modeling approach taken in the Customer 4 pack. We need to take into account all the client attributes provided to us and attempt to find the best subset which gives the most performant model. For the technically inclined this problem is an NP hard problem in optimization and so each model requires the Data Scientist to build multiple models by varying the predictor set. This adds to the model complexity and delivery time.
Note: any key sidenotes and/or related links to further dive into above-stated details.