Value × health: where to spend staff time
Action list
at-risk first, then by giving at risk (annual gift × chance of lapsing)
| Member | Type | Annual gift | Lifetime value | Health | Why the score | Recommended action |
|---|
The model weighs dozens of signals at once. These are the ones that move a member's health score most, grouped into themes. Click a theme to see what's inside it. On your real file, this page is measured from your data, not assumed.
The bar is how much of the model's decision comes from that theme. The arrows say which way a signal pushes:
Every "why the score" line in the action list comes from these same signals, for that one member.
Mostly current gift size, whether gifts are growing or shrinking, years of giving, and sustainer status. A small sustainer of twelve years can be worth more than a one-time larger gift.
Value says what a member is worth keeping. Health says how likely you are to lose them. Staff time goes where both are high: valuable members starting to drift. Automated touches go everywhere else, and healthy members mostly get left alone.
| Member | Type | Annual gift | Lifetime value | Health | Why the score | Recommended action |
|---|
An export from your CRM (Raiser's Edge, Salesforce, Bloomerang, and so on): gifts, recurring payments, and tenure. Add newsletter, podcast, and event engagement if you have it.
Rich history: a panel model that learns month-by-month patterns. Sparse history: a recency, frequency, and gift-size model. Both are tested on members who already lapsed.
Health is the chance a member is still giving in twelve months. Value is predicted lifetime giving. Every score comes with the reasons behind it.
Each segment gets a plan, and the list refreshes on a schedule so new risks show up while there's still time to act.