Planned giving

Which donors look like bequest prospects?

Can't tell yet. No public dataset we ship, or know of, records which donors actually left a bequest, so there is nothing honest to report here yet: no accuracy number, no chart. Anything we showed here would be scoring the model against a stand-in for bequest intent, not bequest intent itself, and would look like proof it isn't.

None of the real files we test other models on (KDD Cup 1998, DonorsChoose, PSID, Karlan and List) records a bequest or estate-intent signal either, so this gap is not specific to the sample data.

What this means for your file

There is no number yet for how many bequest prospects the model would find in a file of 10,000 donors, and no simple rule to compare it against.
PlannedGivingIntentScorer needs donors labelled with whether they made a planned gift (a bequest, a charitable trust, a beneficiary designation), not just whether they kept giving. That label usually only exists inside a gift officer's own CRM, built up over years.
If your shop tracks planned-giving intent, even informally as a flag or a moves-management stage, you can run the check below on your own file.

Try it on your own donors

import numpy as np
from philanthropy.models import PlannedGivingIntentScorer

# Replace with your own donor feature matrix and bequest-intent flags
# (1 = made a planned gift, 0 = did not); this uses random arrays only to
# show the API runs end to end.
rng = np.random.default_rng(0)
X_train, y_bequest_train = rng.random((200, 3)), rng.integers(0, 2, 200)
X_test, y_bequest_test = rng.random((50, 3)), rng.integers(0, 2, 50)

model = PlannedGivingIntentScorer(random_state=0).fit(X_train, y_bequest_train)
score = model.predict_intent_score(X_test)

Compare score's ranking against y_bequest_test using whatever top-N hit rate matters to your program, the same way the other Results pages do. If you're willing to share an anonymized version of that comparison, it would let this page report a real number instead of "untested."

How we tested

We have not: there is no labelled file to test on. There is also no established simple rule for planned-giving prospects to compare against, the way "rank by giving so far" exists for major gifts, so a real test needs both a labelled file and someone to define what "beating no model" looks like for this question.

Numbers for analysts

The drivers below come from a giving-response stand-in label on sample data, not real bequest intent; read them as "what the estimator weighs when fit on this kind of label," not as a finding about who actually leaves a bequest.

Sample data

On this file the model looks at 3 things about each donor.

What it knows	What that means
Giving history	how much and how often they have given in total
Recency	how recently they gave

This file has no engagement and mailing history records, so they are not used here. Wealth & demographics is in this file, but this model is not given it here.

These matter most:

What raises or lowers the score
lifetime giving	▲ raises the score

Only 1 feature had a measurable effect; the rest are in the analyst note below.

For analysts

Columns: total, n, recent

Column	Importance	Direction
`total`	0.190 (range 0.155 to 0.222)	+
`recent`	0.019 (range -0.000 to 0.047)	mixed
`n`	-0.003 (range -0.007 to 0.001)	-

Method: permutation importance (roc_auc), partial-dependence sign for direction ("mixed" if it changes sign). Split: walk-forward (train on fiscal years < T, test on T). Git SHA: d857e8aac5ff4cdbb8c91450cec7583480c1d7af.

Results on your own file will differ.