What RFM scoring tells you that instinct does not

A fundraising manager four years into the job can usually name the top forty donors without opening the database. They know who gives the most, who has given every year since 2019, and which two will pick up the phone in November if the ask is about the shelter. When the first RFM report lands, most of it agrees with them. The big donors score high on value, the loyal ones on frequency. It reads as a tidy confirmation of what the team already believed, which is a strange thing to buy software for.

That reaction is fair. It is also where most RFM work stops being useful, because the three letters, recency, frequency and monetary value, do not carry equal weight against a fundraiser's own judgement. Recency is the one that argues back.

The gift that still feels recent

A team's memory of a donor updates on the size of the gift, not the date of it. A supporter who gave $2,000 to the tax appeal two years ago is filed under major donors and stays filed there. Nothing in the ordinary run of a week moves them out: gifts generate receipts, thank-yous and a line in the weekly figures, and twenty-six months of silence generates no paperwork at all.

Recency is the one letter measured against the calendar instead of against the team's impression of a donor, which is why it disagrees with them. The supporters most likely to slip out of view are the mid-value ones. They gave enough to be remembered warmly, and not often enough for the silence to register.

In practice that shows up as a call eighteen months later than it should have been, to someone who has since started giving somewhere else. That is what the second gift that was never asked for usually is: an ask that was owed on a date nobody had in front of them.

A 5 is a rank inside your own database

RFM scores are relative. Each donor is ranked against the rest of that organisation's donors and placed in a quintile, so a 5 for value means the top fifth of this donor base and nothing more. In a program where the median gift is $40, a 5 on value might be $250 a year. Put a major-gift pipeline behind it and the same score might be $25,000.

The relative scale is the right design. A fixed scale built around some sector-wide idea of a good donor would put almost every supporter of a volunteer-run committee in the bottom quintile, which tells that committee nothing about who to ring first. A ranked one works in a program of 300 donors and a program of 300,000 without being tuned.

It is also misread constantly. A 5/5/5 looks like a verdict on the donor when it is a verdict on the room. Two organisations can hold the same supporter, score them 5/5/5 and 2/3/2, and both be right. That is why the score belongs on a call list rather than in a board pack, where the only number that travels is the dollar figure.

The donor who gives twice a year on purpose

Frequency is almost always counted as a raw tally of gifts. A donor on a $10 monthly debit produces twelve gifts a year and lands in the top quintile. Someone who gives $500 in July and $500 again in December produces two, and lands in the bottom half. That is $120 of giving ranked above $1,000.

A count of twelve is a fact about how a payment was set up. Patterned giving is a different question: does this person give at the same couple of moments each year, at a similar size, without being chased.

Plenty of Australian donors think of themselves as committed to an organisation and give on a schedule of their own. They would describe themselves as regular supporters and mean it. Most fundraising teams only treat someone as a regular giver once there is something on a monthly debit, and the case for one recurring donor over several one-off gifts is often made in a way that leaves them out. The frequency letter takes that narrow definition and hard-codes it into the score.

A program that wants frequency to mean something can count distinct giving years instead of gifts, or gifts that arrived outside an appeal, or whether the interval between gifts holds steady. None of those is harder to compute than a tally.

From a score to a call list

A mid-sized Australian charity spent months on predictive modelling. The analysis was sound and the enthusiasm was real, across the fundraising team and up to the board. Nothing came out of it that changed a fundraiser's week: no ranked list, and nowhere in the existing workflow for a score to land. The work went unused.

That is a more common failure than a bad model. A segment is only worth having if it changes the next action a fundraiser takes, and a score is under the same obligation. What comes out the other side has to be a list of names in priority order, with an amount beside each and a reason attached to it.

Around 22,000 Australian charities, 43 per cent of the sector, are run entirely by volunteers, according to the ACNC's research on volunteer-run charities. In most of the rest, the person who would act on a donor score is the same person writing the newsletter and chasing the acquittal. A model that needs someone to sit down and interpret it loses to the newsletter every time.

Together computes RFM the same way everyone does (a quintile is a quintile), then treats the three letters as an input rather than an answer. Upgrade, lapse-risk and reactivation signals are calculated on top of them, each carrying the factors that produced it, and the result lands as a ranked worklist with a suggested amount beside each donor. For organisations running NationBuilder, the scores, suggested asks and segment tags are written back as custom fields, so they turn up where the fundraiser already works rather than in a second system somebody has to remember to open.

All of it rests on the donation history being complete and the donor records being clean. A supporter who appears three times in the database, once per channel, gets ranked three times and scores badly in each. Scoring a messy database produces a confident ranking of the wrong thing, which is harder to argue with than no ranking at all.

The version that takes an afternoon

The cheapest version of this needs no composite score at all. A team can sort the donor list by the date of the last gift, oldest first, and read the first hundred names against what the fundraisers believe about those donors. The names that surprise someone are the ones worth a call this month, and that list arrives in an afternoon rather than a quarter.