7 Key Steps to a Successful Predictive Model
We often see teams jumping at analytics tools before they've properly framed what they are trying to do. We see a ton of focus on the actual math in the models, vs. the process to set up for the model.
In our work we recommend a process that has 7 steps:
- Clearly identify the problem you are trying to solve
- Create hypotheses about what might influence that problem. What data do you have? What additional data can you get your hands on?
- Blend and synthesize your data into explanatory factors that will work in a model.(most of your data does not come out of the database in this form)
- Visually explore the data and adjust your hypotheses (step #2)
- Build predictive models
- Examine the output and adjust the models and re-run them
- Bin and name the outputs so that the team can easily understand them.
In our opinion running processes like this is far more important than which tool you use. And there are data analytics tools (e.g. ours...) that do all of this in a "no coding required" approach.
A couple of tutorials:
- Awesome Analytics in Fundraising (30 mins)
- Million Dollar Donor Workshop (30 mins)
- Attachment Scores- How to Create them and Where to Use Them (30 mins)
- Overview on what ADVIZOR Does (3 mins)