Key points
- A consulting project answers one question, once, often after months.
- A platform lets your own team build, explain and refresh models as donors and campaigns change.
- It is not either/or. Teams with data scientists can export model code and use the API.
Two ways to get a model
A nonprofit that wants predictions has two routes. It can commission a model as a project, or it can run a platform its own team uses.
What a project gives you
A project answers one question well. It also tends to take months, and it ends. The model reflects your donors as they were when the data was pulled. When the file changes, or a new question comes up, the answer is another project.
What a platform gives you
A platform puts the modeling in your team’s hands. The fundraiser who has the question can ask it, see the reasons behind each score, and ask again next month with fresh data.
Side by side
| A consulting project | A platform your team runs | |
|---|---|---|
| Time to first scores | Often months | Days. UPMC Health System’s implementation took one day |
| Who runs it | The consultant | Your own team, with no data science degree needed |
| The next question | A new statement of work | Another model, built by the same team |
| Explanation | Varies by project | The drivers behind every score, in plain language |
| Keeping it current | A refresh project | Retrain as new gifts and results arrive |
| Commitment | A project fee | From $5,000 a month, month to month |
| Where results go | Often a file of scores | Any system: CRM, email, print vendor, ad platforms or the warehouse |
If you already have data scientists
This is not either/or. Teams that build their own models can export model code from Squark, call the API, and spend their time on the questions that need custom work.
A first step
Start with one use case and a CSV file. Compare the time to a first scored list, and the explanation you get with it, against your last modeling project.
See pricing, how Squark works and the UPMC Health System case study.