Propensity and capacity scores: what they are, and what they're not

A propensity score estimates how likely someone is to give; a capacity score estimates how much they could give if they did. They answer different questions, sit on different axes, and the most common scoring mistake in fundraising is treating a high score on one as if it were the other. A wealthy donor who never gives and a loyal donor of modest means are opposite errors, and only telling them apart makes either score useful.

What is a propensity score in fundraising?

Propensity is likelihood: based on someone’s history and engagement, how probable is a gift (or a next gift) in a given window? It’s built from behavioral signals: past giving, recency, interactions, and responsiveness. High propensity means “likely to say yes,” regardless of size.

What is a capacity score?

Capacity is ability: how large a gift could this person make if they chose to? It’s estimated from wealth and giving indicators, often via external data, and says nothing about willingness. High capacity means “could give a lot,” even if they never have and maybe never will.

Why confusing them is so costly

Lead with capacity alone and you get the classic major-gifts trap: a list of wealthy names with no relationship to you, most of whom will never give, and gift officers burning months on cold, high-net-worth strangers. Lead with propensity alone and you get lots of likely “yes”es for small amounts. That’s efficient, but you’ll miss the donor who could do something transformational. You need both, on two axes.

Read them as a quadrant, not a single rank

The useful move is to plot the two against each other:

| | Low propensity | High propensity | |---|---|---| | High capacity | could, won’t (yet): cultivate the relationship | focus here: able and willing | | Low capacity | low priority | will, but small: steward efficiently |

The top-right is where major-gift effort belongs. The top-left (high capacity, low propensity) is a cultivation problem, not an ask-now problem. Mistaking it for the latter is exactly the misread the single-number approach produces.

What they’re not

Three honest limits worth stating plainly:

• Not certainty. A score is an estimate with error, not a verdict. It orders your attention; it doesn’t predict an individual outcome.
• Not a substitute for judgment. It’s a starting point a gift officer refines with relationship knowledge no model has, which is the same posture as from hunch to score.
• Not trustworthy on bad data. A score computed on duplicated or undefined records is confident noise. Capacity from a stale external match, or propensity from a fractured giving history, ranks the wrong people convincingly.

What you get

Two scores you can actually act on, one for who to ask now and one for how much to aim for, read together as a map rather than a leaderboard. You stop chasing wealthy strangers who’ll never give and stop overlooking the willing donor who could do far more, because you finally measured the two things separately.

For where predictive work sits on the ladder, see What Becomes Possible; to check the foundation underneath, take the self-assessment.

Examples use synthetic data. The standard is open and early; treat current releases as drafts.