While look-alike models can be built quickly from within the WealthEngine (WE) platform, that is not the case with custom models. For this type of modeling, you will be working closely with your Client Engagement Manager (CEM) as a team of WE data scientists builds you a custom formula based upon both your specific goals and your constituency. Because of this, the process will take significantly longer and be somewhat more involved. In the end, however, you will see a massive return for that additional time and effort. This article will walk you through what you can expect.

  • Note: This process applies to our standard custom and Enterprise models, which may not already be a part of your subscription. To learn more about the different types, please see our other modeling resources.


Checklist of steps to follow:

  • Step 1: Consider your needs
  • Step 2: Kickoff Meeting
  • Step 3: Submit your file
  • Step 4: WE data scientists build the model
  • Step 5: Your CEM analyzes the results
  • Step 6: Your results are delivered
  • Step 7: Work with your CEM on turning these insights into action.




Step 1: Consider your needs – and whether you have the required information

As a general rule, the custom models are all intended to both deepen your understanding and predict behavior. However, by their very nature, the details will very based on your needs. This means that from the very beginning you should be thinking about:

  • Note: There are different types of custom models, some of which are more robust than others. To learn more, make sure to check out our resources.

Substep 1:  What specific behavior do you want to predict? 

The model will be designed to answer this question. To some, it may be: 

  • Moving from membership to donor
  • Buying a luxury suite, 
  • Donating $500,000+
    • But wait – do you mean in one check? Within a period of 5 years?

Substep 2: Do you have the information you need to model that?

To be successful, you must already have a sample of at least 200 – 300 people who have exhibited the behavior you want to predict. This is your “target” – and if that target is too small, what you get back from the model may not be statistically significant. It’s like in any scientific query: if your sample size isn’t large enough, your results may be skewed. 

  • Note: If you don’t yet have this information, that’s okay! Speak with your CEM about how you might go about gathering it so that you will be able to build a model in the future. 



Step 2: Kickoff Meeting

If you decide you are interested in building a custom model, you should contact your Client Engagement Manager to schedule the kickoff meeting. In that meeting, you and your CEM will explore and solidify what your goals are, as well as what variables will be added to customize the modeling formula for that purpose. 

A true custom model will be specifically designed with your organization’s goals and qualities in mind, so this step is critical to your future success. 

The meeting should take place before you screen your list. 

  • Note: If you aren’t sure who your CEM is (or whether that is included in your subscription) you should reach out to your Sales Representative instead. 



Step 3: Submit your file 

Having spoken with your CEM and gotten a list of next steps, you will submit your file.

Substep 1: Using the Batch Upload, send WE the list that you want screened against the model. 

Substep 2: Identify your Target group. 

Again, this will be a subset of your file made up of the people who behave the way that you want to predict. For example, if you are building a Major Gift Model, this sample will be made up of your top major gift givers. 

The file will then be screened in a process that typically takes 10 – 15 business days. 



Step 4: WE data scientists build the model

Once the screening is completed, our data scientists will analyze that data and use machine learning to create the custom formula that will make up your model. This will be specific to your organization and is based on both the data itself and your goals, as discussed during the kickoff meeting.

The first model will likely take 2 – 3 weeks to build. For every subsequent model, you can expect another week to be added to that timeline.


Once the model has actually been built, our data scientists will score every person in your file. against it.



Step 5: Your CEM analyzes the results

Your Client Engagement Manager will then take about 1 – 2 weeks to analyze your model results. This is a critical step, as they will be better able to guide you in successfully implementing those insights. 



Step 6: Your results are delivered

At this stage, you will finally be provided with an Executive Summary that takes you through the modeling process and explains exactly what you have gotten in return. 

You will be able to review the modeled scores themselves in My Profiles via newly added columns in your view of that list, or in a flat file that is sent to you outside of the WealthEngine platform. This will largely depend on the size of your file. 


Each person in the full screened list will have two new scores:

Model Decile: 

Individuals are ranked and divided into 10 deciles. Decile 1 is the best. 

Model Score: 

This is a raw number that is statistically adjusted to go from 100 – 1000. The higher the number, the more predictive.


We strongly encourage you to focus on the decile. The raw score is usually only utilized by clients who have such an enormous list that Decile 1 might include both people with a score of 1000 and a score of 700. This will be very rare, however. In most cases the decile alone will be sufficient and focusing on the minute differences between individual raw scores will not be a valuable use of your time and resources. 



Step 7: Work with your CEM to turn these insights into action.

By this point, your Client Engagement Manager will have thoroughly explored your needs during the kickoff meeting and analyzed the results of your modeled screening. This situates them perfectly to guide you through how to read the model scores, interpret the its independent variables, and finally, your actual next steps.