Lead-Scoring for Charities, Without the Hype

Every CRM vendor sells a lead-scoring feature. Most charities try to use it once, get confused by what the score means, and quietly stop looking at it. Then a year later someone proposes building a "proper" predictive model - usually with a six-figure budget and a 12-month timeline - and the whole conversation starts again.

There is a much smaller, much more useful version of lead-scoring that almost no charity I have worked with had in place when I arrived: three signals, three tiers, refreshed weekly. It is not glamorous. It does not need a data scientist. It works.

What lead-scoring is actually for

A lead-score answers one question: "Out of all the supporters in our database, who should the major-gift fundraiser call this week?" That is it. If the score does not change that decision, the score is decoration.

Predictive models try to answer a more elaborate version of the question - "what is the probability of this person making a gift of £X in the next Y months." That is a harder question, and most of the time, charities do not need the harder question. They need the simpler one, answered reliably.

The three-signal model

Pick three signals you genuinely capture and trust. The shortlist for most charities looks like this:

• Recency: how recently did they last engage in a meaningful way (gift, attendance, opening multiple emails in a 30-day window)?
• Capacity: a capacity proxy - could be largest historical gift, postcode-based wealth screening, employer information, or a simple "have they ever given more than £500" flag.
• Affinity: a behavioural proxy for closeness - number of touchpoints in the last 12 months, programme attendance, volunteer history, response to surveys.

Each signal scores 0–3. Sum the three. The result is a 0–9 score. Three tiers map to action:

1. 0–3: routine cultivation. Stay in newsletters, no individual outreach.
1. 4–6: warm relationship. Personal email from a fundraiser within 4 weeks.
1. 7–9: priority. A named relationship manager, with a 7-day SLA on first contact.

That is the entire model. It fits in a single SQL view. A fundraiser can read it in 30 seconds. You can deploy it in two weeks.

Why the simple model holds up

The temptation is always to add more signals. Resist. There are three reasons the three-signal model outperforms a 30-signal model in most charity contexts.

1. The data is rarely clean enough for sophistication

A predictive model trained on noisy data produces noisy predictions. Most charity CRMs have completeness in the 50–80% range across key fields. Three signals you trust beats thirty signals you half-trust.

2. The number of "actionable" supporters is small

For a charity with 10,000 active records, the top tier (score 7–9) is usually 200–400 people. A fundraiser can humanly contact 30–50 of them in a quarter. The score does not need to discriminate within the top 200; it needs to find the 200. A simple model does that.

3. Maintenance cost compounds

A 30-signal model needs 30 inputs to keep clean. When one breaks, the score breaks. A 3-signal model has three things to maintain. When the data warehouse hiccups, you can fix the score in an afternoon.

How to deploy it - in two weeks

Week 1: Define and validate

Sit with the major-gift team. Ask: "Show me ten supporters you wish you had called sooner this year." For each, record their score under your draft signals. The model should rank those ten in the top 10–20% of your supporters. If it doesn't, your signals are wrong.

Iterate the signals until the model puts the "wish I'd called" supporters in the top tier. That is calibration. It is not a science; it is a sanity check.

Week 2: Operationalise

Build the SQL view (or CRM filter). Schedule it to refresh nightly. Build a single dashboard: top tier, count, week-on-week change. Give the major-gift team access. Agree the SLA: anyone newly entering the top tier gets a personal email within 7 days.

Two weeks. One view. One dashboard. One SLA. That is a working lead-scoring system.

When to add complexity

There is a real moment to graduate to a more sophisticated model. The signals to look for:

• Top-tier population is consistently above 1,000 - too many for personal contact, so finer ranking matters.
• You have 5+ years of clean gift history with attribution to campaigns and channels.
• You have either an in-house data team or a budget for a 6-month build with a partner.
• Your fundraising team trusts the simple score and is ready to push it harder.

If three of these four are true, predictive modelling will pay back. If two or fewer, stick with the simple model and put the budget into supporter experience instead.

Two traps to avoid

Trap 1: Hiding the score behind a black box

A predictive model that fundraisers cannot interpret will not be trusted, and an untrusted score is unused. Always be able to answer: "Why is this person tier 1?" If the answer is "because the model said so," your fundraisers will quietly ignore it.

Trap 2: Confusing score with strategy

A high score does not mean "ask for money now." It means "this is a person worth a human conversation." The conversation might lead to a gift, a major-gift cultivation plan, a legacy conversation, or simply a deeper relationship. Treat the score as a triage signal, not a sales lead.

A short closing

Lead-scoring is one of those areas where charities have been sold a complex solution to a simple problem for a long time. The simple version - three signals, three tiers, weekly refresh, owned by one person - outperforms the complex version for the vast majority of charities, costs almost nothing to maintain, and survives staff changes. Build that first. Add complexity only when it earns its place.

Most fundraising teams do not need better predictions. They need better triage. The simple score is triage at scale.

Further reading

Choosing a Charity CRM in 2026 | Donor Segmentation That Actually Moves Money | The Five-Minute CRM Health Check