---
title: "How to Know Which Donors Are At Risk of Lapsing (Before They're Gone)"
description: "Only 3% of lapsed donors ever come back. The warning signals that show up months early, a scoring method you can run today, and the software that predicts it."
canonical: https://gratefully.io/blog/donors-at-risk-of-lapsing
category: "Retention"
date_published: 2026-07-14
date_modified: 2026-07-14
read_time: "12 min read"
author: "Muddsar Jamil"
keywords: "donors at risk of lapsing, software to predict donor lapse, lapsed donor definition, LYBUNT, donor attrition signals, donor retention rate"
source: Gratefully — Donor Intelligence for Nonprofits
---

# How to Know Which Donors Are At Risk of Lapsing (Before They're Gone)

## Quick answer

To know which donors are at risk of lapsing, compare each donor against their own giving rhythm, not a fixed calendar. Risk shows up as a gap stretching past their normal interval, a sharply smaller gift, fading email engagement, or a failed recurring payment. Only 3 percent of lapsed donors come back, so the work is catching them early.

## Intro

Margaret gave every November for six straight years. Last November she did not, and nobody noticed, because nobody was looking for a gift that did not arrive. The CRM will notice next fall, when its lapsed report finally moves her over the line. By then she will be almost two years past her last gift, and the letter she gets will read like it was written to a stranger.

Here is the uncomfortable part: everything needed to catch Margaret in December was already in the database. The six-year November pattern. The email opens that stopped in spring. The gift officer who left in August and took the relationship's memory with her.

Knowing which donors are at risk of lapsing is not a data problem. Most organizations have the data. It is an attention problem: the signals live in five different places, and nobody has time to check them donor by donor. This guide covers what "at risk" actually means (and why every source defines it differently), the numbers that make early detection non-negotiable, the full set of warning signals, a scoring method you can run this afternoon with a worked example, and an honest look at the software that watches for you.

## What "at risk of lapsing" actually means

The most widely used definitions come from [Blackbaud's donor lifecycle framework](https://webfiles-sc1.blackbaud.com/files/support/helpfiles/rex/content/bb-donor-lifecycle.html), and they form a ladder:

- **At-risk:** no gift in the last 12 to 15 months
- **Lapsing:** no gift in 15 to 24 months
- **Lapsed:** no gift in two to five years
- **Lost:** no gift in more than five years

Useful vocabulary, and worth adopting for reporting. But before you build your alerts on it, notice that the sector does not actually agree. Funraise draws the at-risk line at just 6 to 12 months. The Donor Relations Group flags annual donors at 9 to 10 months of silence, and mid-level and major donors within the fiscal year. Fundraising Report Card counts anyone who gave one year and not the next.

The disagreement is not sloppiness. It is a clue that fixed thresholds are the wrong tool. A monthly donor whose gifts stop is at risk after six weeks, not twelve months, and Fundraising Report Card makes exactly this point: a monthly donor has lapsed after one missed month. Meanwhile a donor who gives every December, and only every December, is not at risk in June no matter what the dashboard says.

So the working definition this guide uses: **a donor is at risk of lapsing when their behavior deviates from their own established pattern.** The calendar thresholds are averages. Your donors are not average, and the deviation is measurable, which is what the scoring section below is for.

## The math that makes early detection non-negotiable

Three verified numbers define the problem. All are full-year figures from the [Fundraising Effectiveness Project's Q4 2025 report](https://publications.fepreports.org/), the sector's standard retention benchmark.

- **Overall donor retention is 43.3 percent.** Nonprofits lose more than half of their donors every year.
- **First-time donor retention is 18.9 percent.** Fewer than one in five new donors ever gives a second gift.
- **The recapture rate is 3.0 percent, and falling.** Of all previously lapsed donors, only about 3 in 100 resume giving in a given year.

Sit with that last one. Win-back campaigns operate in a 3 percent world. Retention operates in a 43 percent world. Every donor who crosses from at-risk to lapsed moves from the second world into the first, which is why identifying risk early is not a nice-to-have report, it is the highest-leverage work in fundraising.

Donor retention looks different measured year to date. Through the first quarter of 2026, under a methodology the Fundraising Effectiveness Project revised in the same report, overall retention runs at 18.0 percent, new donor retention at 7.1 percent and repeat donor retention at 25.8 percent. Year-to-date and full-year numbers are not comparable, so check which basis you are holding your own figure against.

The data also says where that crossing is most likely to happen. Retention by gift count in Q1 2026 runs 7.4 percent for donors who gave once the previous year, 19.4 percent for those who gave twice, 44.6 percent for three to six gifts, and 88.1 percent for seven or more.

The steepest cliff in the whole dataset is between the first gift and the second. A first-time donor is not at risk because of anything they did. They are at risk by default, and they are the largest group in most files. If you only have capacity to watch one segment, watch that one.

The cost side says the same thing. The classic benchmarks from [James Greenfield's fundraising cost research](https://www.abhe.org/wp-content/uploads/2023/02/Cost-to-Raise-a-Dollar-Perkins.pdf) put new-donor acquisition at $1.00 to $1.25 spent per dollar raised, versus roughly $0.20 per dollar to renew an existing donor. Keeping a donor is about five times cheaper than replacing one, which is the honest version of the "5x rule" you see quoted, and the reason [the 7x rule of donor retention](/blog/7x-rule-donor-retention) compounds the way it does.

And the reasons donors leave are mostly fixable. In Dr. Adrian Sargeant's landmark donor defection research, lapsed donors were asked directly why they stopped: 13 percent said they were never thanked, 9 percent had no memory of supporting the organization at all, 8 percent were never told how their money was used, and 5 percent thought the organization did not need them. (54 percent said they could no longer afford to give, a reminder that some lapse is genuinely not yours to prevent.) The study is from 2001 and remains the most-cited donor-side evidence in the field: most preventable lapse is a communication failure, not a change of heart.

**Author note:** In twenty years around development teams, the lapsed-donor report is the most quietly misleading document I have seen. It feels like intelligence, but it is an obituary. By the time a name appears on it, the odds of getting that donor back are about 3 in 100. Every one of those names was catchable earlier, and the signals were sitting in the CRM the whole time. The teams that retain well are not the ones with better reports. They are the ones who look earlier.

One honest caveat: nobody has published a reliable figure for exactly how far in advance the warning signals appear, and any article that gives you one is guessing. What the sector's own data does support is the qualitative version: donors rarely stop abruptly. Disengagement shows up in email behavior, gift timing, and payment failures first, and giving follows.

## The warning signals, all of them

Across the six most-cited resources on donor attrition, 26 distinct warning signals appear. They cluster into four families. No single signal means a donor is gone; each one means the relationship needs eyes on it, and two or more together mean it needs eyes on it this week.

### Giving behavior (the strongest signals)

- **A stretch past their own giving rhythm.** Every donor has a natural cadence: monthly, annual, every spring appeal. When the gap since their last gift grows noticeably past their usual interval, that is the earliest reliable warning. Watch the deviation, not the calendar.
- **A downgraded gift.** The Donor Relations Group flags a reduction of 25 percent or more versus the donor's previous giving as an at-risk trigger, and it is a good line. A donor who gave $1,000 last year and $250 this year still shows up as "retained" in your annual report. Treat the downgrade as a half-step toward the exit.
- **A broken pattern.** The donor who answered your year-end appeal five years running and skipped it has told you something no threshold report will flag. Skipped-their-usual-appeal is among the strongest single signals you have.
- **A first missed recurring gift.** Recurring donors are your most loyal cohort: Neon One's Recurring Donor Report measures their retention at 78 to 80 percent, versus the low 30s for one-time donors. Which is exactly why a missed or cancelled monthly gift is urgent. Their normal interval is measured in weeks, so six weeks of silence from a monthly donor is the equivalent of a year of silence from an annual one.
- **First-year donor status.** Not a behavior but a cohort fact: with 18.9 percent first-time retention, a new donor is at risk by default until their second gift. If your risk list does not include every first-time donor approaching month 9 to 10 without a second touch, it is missing the largest at-risk group you have.

### Communication engagement (the leading indicators)

- **Email engagement fade.** Opens and clicks that stop usually cool before giving does. The Donor Relations Group uses three consecutive unopened emails as its behavioral trigger. Calibrate expectations with the sector baseline: [M+R's 2026 Benchmarks](https://www.mrbenchmarks.com/) put the average fundraising email click-through at 0.59 percent, so the signal is not "did not click this appeal," it is "used to open and click, and stopped."
- **Unsubscribes and bounces.** An unsubscribe is a donor telling you the relationship is over-communicated or under-valued. A hard bounce means you have lost the channel entirely and may not know it.
- **Event absence and channel silence.** The donor who came to everything and stopped coming, the website visits that ended, the social interactions that dried up. Individually weak, meaningful together.

### Relationship signals (the ones that live in notes and inboxes)

- **Relationship silence.** No call, meeting, or personal note logged in six months or more. Gifts follow attention. If nobody has touched the relationship in half a year, the next gift is running on momentum alone.
- **Their person left.** When a gift officer or development director departs, every relationship they held becomes at-risk by default. The Donor Relations Group flags "no gifts since their gift officer departed" as its own signal, and they are right: donors lapse in the handover gap, not by decision. This is [institutional memory loss](/institutional-memory-loss-nonprofits), and it is the most predictable donor-loss event there is.
- **Friction in the record.** An unresolved complaint. A note about a life change. Dissatisfaction expressed at an event or in a survey reply. Human friction hides in free-text notes where no report can see it, and Sargeant's data says it drives more lapse than most teams believe.

### Operational signals (the ones your org causes)

- **Payment failure.** Expired and reissued cards quietly end more monthly-donor relationships than any decision to stop giving. The subscription industry, which measures this obsessively, attributes 20 to 40 percent of all churn to involuntary causes like failed payments, and there is no reason to think monthly giving is immune. This is the most fixable signal on the list, and the one nobody owns.
- **Acknowledgment latency.** A gift thanked late, or never, is a lapse generator you control end to end. Remember the Sargeant number: 13 percent of lapsed donors said they were never thanked.
- **Solicitation imbalance.** Ask-only messaging with no impact reporting, or total silence between appeals, both show up repeatedly in the research on why donors quit. If every touch a donor gets is an ask, the lapse is being manufactured in-house.

Signals 1 through 8 live in structured data your CRM already has. Signals 9 through 11 live in notes, inboxes, and people's heads, which is exactly why they get missed. Signals 12 through 14 are about your operation, not the donor, and they are the cheapest to fix.

## Score it yourself: a risk model you can run this afternoon

You do not need to buy anything to start.

Before you score anything, split the file by gift count. Donors on their first gift, their second, three to six, and seven or more behave so differently that scoring them together hides the pattern. In the Q1 2026 sector data those four groups retain at 7.4, 19.4, 44.6 and 88.1 percent respectively. A one-gift donor sitting at moderate risk needs a different response from a seven-gift donor at the same score, and a single ranked list will not tell you that.

Every CRM already knows how many gifts each donor has made, so this split costs nothing and takes minutes.

The deepest scoring treatment currently published (ReportingXpress's attrition risk framework) weights recency against the donor's own average interval at 35 percent, frequency deviation at 25 percent, giving trend at 20 percent, tenure at 15 percent, and thank-you latency at 5 percent. Directionally sound, but you do not need weighted percentages to get moving. Here is a simpler point model built on the same logic.

**Step 1. Export** gift history (donor ID, date, amount), last-contact date, and per-contact email engagement if your email tool reports it.

**Step 2. Find each donor's rhythm.** For every donor with three or more gifts, compute the typical gap between gifts (the median interval). This is their personal baseline.

**Step 3. Score the stretch.** Divide days-since-last-gift by that typical interval:

- Under 1.0: 0 points (on rhythm)
- 1.0 to 1.5: 1 point (drifting)
- 1.5 to 2.0: 3 points (stretched)
- Over 2.0: 5 points (nearly gone)

**Step 4. Add points for the other signals.**

- Most recent gift down 25 percent or more from their norm: +2
- Skipped an appeal or campaign they historically answer: +3
- Recurring gift failed or cancelled: +4
- Email engagement stopped (or unsubscribed): +2 (+3 if unsubscribed)
- No personal contact logged in 6 months: +1 (12 months: +2)
- Relationship owner departed in the past year: +2
- Unresolved complaint or friction note: +2
- First-year donor without a second gift: +2

**Step 5. Triage by total.** 0 to 3: healthy, keep stewarding. 4 to 7: outreach this month. 8 or more: outreach this week.

**Worked example.** Margaret, from the intro. Median interval between her gifts: 12 months (six consecutive Novembers). In mid-December she is at roughly 13 months, a stretch ratio of about 1.1: 1 point. Her email opens stopped in spring: +2. Her gift officer left in August: +2. No personal contact logged since: +1. Total: 6. Outreach this month. Note what just happened: no single signal was dramatic, and the calendar-based report will not flag her for another ten months. The combination flags her in December, while a warm call still lands.

If you want the more formal analytical backbone, [RFM analysis](/blog/rfm-analysis-ai) (recency, frequency, monetary value) is the classic framework and pairs naturally with this. RFM is also where the segment names you may have seen in donor tools come from: Champions, At Risk, About to Sleep, Hibernating. Those buckets are this same deviation logic, packaged into standing segments instead of a one-off score. And be honest about the spreadsheet's limits: it is a snapshot that starts aging immediately, it cannot see signals 9 to 11 (the departed gift officer, the complaint in the notes), and someone has to remember to rebuild it every month. Those three limits are precisely the line where software starts earning its cost.

## From spreadsheet scores to machine prediction

The step up from a monthly spreadsheet is machine-learning prediction, and it is worth understanding what actually changes, because "AI" gets used loosely here.

A static score, like the one above, checks a fixed list of factors at a moment in time. An ML model trained on your own giving history learns which combinations of factors preceded lapse for your donors specifically, checks far more factors than a person would (Dataro, the most explicit vendor about its method, says its models weigh hundreds), and re-scores continuously as new data arrives. The practical differences are three: earlier flags, fewer false alarms, and nobody has to remember to rebuild the spreadsheet.

What ML does not do is read minds. Models built only on transactions and email data still cannot see the complaint in a call note or the departed gift officer, unless the system also ingests that unstructured layer. And a prediction without an explanation is a black box asking for trust.

**Author note:** One thing I would push any team to demand from any predictive tool, ours included: the reason. If a tool says a donor is slipping, it should show you the evidence, the broken pattern, the silence, the note from the officer who left, so a human can judge whether the machine is right. If it cannot, keep your spreadsheet. A score you cannot interrogate is not intelligence, it is a rumor.

## Software that predicts which donors will lapse

We verified what follows on each vendor's own site (July 2026). The single most useful question to ask any tool: does it predict lapse ahead of time, or report it after a threshold? Most tools marketed for retention do the second.

- **Your CRM's built-in reports (Raiser's Edge NXT, DonorPerfect).** Reports LYBUNT/SYBUNT lists after the threshold. Not applicable for explained scores. Included; DonorPerfect from $99/mo.
- **Bloomerang.** Engagement Meter and Generosity Score gauge current warmth and capacity, no lapse prediction claimed. No reason codes. From $125/mo.
- **Fundraising Report Card.** Reports: retention, attrition, and LTV dashboards for diagnosis. Not applicable. Free.
- **GivingDNA.** Partly predicts: lapse-risk segments, refreshed monthly. Explainability not stated. Demo-gated pricing.
- **DonorSearch Ai.** Partly: year-over-year retention predictions, built for prospecting. Explainability not stated. Demo-gated pricing.
- **Dataro.** Predicts per-donor churn propensity from your CRM history. Yes, "transparent signals." From $15,000/yr.
- **Gratefully.** Predicts: nightly Donor Pulse score, At Risk and About to Sleep segments, drift alerts (LYBUNT/SYBUNT cohorts included). Yes, narrative reasons with sources. $400/mo billed annually.

A few honest notes on that list:

**If you only need a list of donors past 12 months, do not buy anything.** LYBUNT and SYBUNT reports (gave Last Year/Some Year But Unfortunately Not This) are built into Raiser's Edge, DonorPerfect, and most donor databases, and Fundraising Report Card's dashboards are free. That covers threshold reporting completely.

**True per-donor prediction is rarer than the marketing suggests.** Bloomerang's Engagement Meter measures current engagement (email clicks, volunteer hours, event attendance), which is genuinely useful, but the company claims readiness-to-give, not lapse forecasting. Blackbaud's AI investments center on major-gift prospecting. Among tools that do predict lapse, Dataro is the most established pure-play, its models are transparent about their signals, and its published accuracy claims are strong, though they are vendor claims, not independent benchmarks. It is also priced for large files, starting at $15,000 per year.

**Where Gratefully fits.** Grace, the AI inside [Gratefully](/), works through the whole portfolio every night and gives each donor a Donor Pulse score built from engagement, giving trends, time since last contact, and the qualitative signals in notes and documents, the unstructured layer where signals like the departed gift officer and the complaint live. When a donor starts to slip, a drift alert fires with the reason attached: not "Margaret is at risk," but "Margaret is at risk, her six-year November pattern broke, her opens stopped in April, and here is the handover note from the officer who left."

The portfolio is also segmented continuously rather than on report day. Every donor sits in a living [RFM-based segment](/blog/rfm-analysis-ai), Champions, Loyal, Promising, New, Need Attention, At Risk, About to Sleep, Hibernating, and Can't Lose, and moves between them as behavior changes, so "who is drifting" is a filter, not a project. Alongside the segments run working cohorts, including LYBUNT, SYBUNT, lapsed donors, and recurring churn, plus action flags like Re-engagement Target. The LYBUNT list your CRM gives you when someone remembers to run it exists here as a view that is already up to date. Every flagged donor comes with a drafted, personalized outreach a human approves before anything sends. Pricing is public: $400 per month billed annually, with a 4-week free trial.

For the full vendor comparison, including what each tool costs and the seven questions worth asking before buying, see [Software to Predict Which Donors Will Lapse](/donor-lapse-prediction-software). For the current retention benchmarks behind the signals above, split by donor type, gift size and gift frequency, see [Fundraising Metrics and KPIs](/blog/fundraising-metrics-benchmarks).

## What to do once you know

A risk list is only worth what you do with it, and the wrong move is the common one: dumping every at-risk name into a generic "we miss you" blast. Go back to Sargeant's numbers. The donor who was never thanked does not need a miss-you email, she needs the thank-you. The donor who was never told what her money did needs the impact report.

**Triage by stakes and by story.** Sort by giving history and by what the signals say. High-value, long-tenured donors with a broken pattern get a personal call this week. Mid-level donors with fading engagement get a personalized touch that references their actual history.

**Lead with gratitude and impact, not an ask.** A donor drifting away is telling you the relationship feels one-directional. The strongest re-engagement touch is evidence their past giving mattered, which is [the whole case for stewardship over acquisition](/blog/stewardship-vs-acquisition). Save the ask until the relationship answers.

**Fix the operational leaks first.** Failed card? Help them update it, this recovers revenue at a rate no appeal matches. Late acknowledgment? Send it, late and honest. Complaint in the notes? Address the complaint. Boring, specific repairs retain more donors than any campaign.

For the already-lapsed, the playbook changes from prevention to win-back, a 3-percent game that needs its own discipline: [donor re-engagement](/solutions/donor-reengagement). And when a donor does slip through, the work changes from prevention to repair. We wrote a full playbook on [how to re-engage lapsed donors](/blog/re-engage-lapsed-donors), starting with diagnosing why each one left.

## The bottom line

Which donors are at risk of lapsing is a knowable thing, months before the lapsed report says so. The signals are not exotic: a stretched rhythm, a smaller gift, a skipped appeal, a failed card, fading opens, a silent relationship, a departed colleague, friction in the notes. With 43 percent retention and a 3 percent recapture rate, the entire economic case sits on one side: catch them while they are slipping, because almost nobody comes back after they are gone.

Start with the point model and the monthly top-25 habit. And if you want the watching to happen every night, across every donor and every signal including the ones buried in notes, with the reason attached to every name, that is the job Grace was built for. A [donor health audit](/donor-health-audit) will show you, on your own data, who is slipping right now.

## Frequently asked questions

### What is considered a lapsed donor?

The most common framework, from Blackbaud's donor lifecycle definitions, treats a donor with no gift in 12 to 15 months as at-risk, 15 to 24 months as lapsing, 2 to 5 years as lapsed, and over 5 years as lost. Definitions vary by organization: some count anyone who gave one year and not the next, and a monthly donor is effectively lapsed after a single missed month.

### What is the difference between an at-risk donor and a lapsed donor?

An at-risk donor still counts as active but is showing warning signs: a stretch past their normal giving rhythm, declining engagement, or a failed recurring payment. A lapsed donor has already stopped giving. The difference is economic: retained donors renew at far higher rates, while the Fundraising Effectiveness Project measures the recapture rate for lapsed donors at only about 3 percent per year.

### How do you identify donors at risk of lapsing?

Compare each donor to their own giving pattern rather than a fixed calendar. Flag donors whose gap since their last gift exceeds roughly 1.5 times their usual interval, whose gift size dropped 25 percent or more, who skipped an appeal they always answer, whose recurring payment failed, whose email engagement faded, or who have had no personal contact logged in six months. Two or more signals together warrant outreach within the week.

### What is a good donor retention rate?

For spotting lapse risk, the aggregate matters less than the split. The Fundraising Effectiveness Project's most recent full-year figures are 43.3 percent overall, 18.9 percent for first-time donors and 59.3 percent for repeat donors, so anything above roughly 50 percent overall puts an organization ahead of the sector. The more useful cut for risk work is by gift count, where retention runs from 7.4 percent at one gift to 88.1 percent at seven or more. Full-year and year-to-date figures are measured differently, so compare like with like. The full benchmark set is in our fundraising metrics guide.

### What is a LYBUNT report?

LYBUNT stands for gave Last Year But Unfortunately Not This year, and SYBUNT for gave Some Year But Unfortunately Not This year. They are standard reports in donor databases like Raiser's Edge and DonorPerfect that list donors who have crossed a lapse threshold. They are useful for reporting, but they identify lapse after it has happened rather than predicting it. Some donor intelligence platforms, including Gratefully, maintain LYBUNT and SYBUNT as continuously updated cohorts rather than a report someone has to remember to run.

### Can AI predict which donors will stop giving?

Yes, within limits. Machine-learning tools train on an organization's own giving history to score each donor's lapse likelihood, checking far more factors than a manual score and re-scoring continuously. Vendors report high accuracy, though those figures are vendor claims rather than independent benchmarks. The feature to insist on is explainability: a risk score should come with the evidence behind it so a human can verify the machine's reasoning.

### How often should you review donor lapse risk?

Monthly at minimum for a manual process: rebuild the risk score, review the top 25 names, and assign follow-ups. Recurring donors need faster attention, since a single missed monthly gift is already a lapse signal. Tools that re-score nightly remove the need to remember, which matters because the most common failure of manual review is that it quietly stops happening.

### What should you say to a donor who is about to lapse?

Match the message to the cause. Donor defection research by Dr. Adrian Sargeant found 13 percent of lapsed donors were never thanked and 8 percent were never told how their money was used, so lead with gratitude and a concrete impact story, not an ask. If a specific friction caused the drift, a failed card or a missed acknowledgment, fix that directly and say so. Save the next ask until the relationship responds.


---

Canonical HTML version: https://gratefully.io/blog/donors-at-risk-of-lapsing
More about Gratefully: https://gratefully.io/ai-instructions