The Development Shelf.

Twelve thank-yous are drafted and waiting. The executive director, who usually signs them, is traveling until Friday. Holding the queue feels careful, but every donor now waits, including those whose messages never needed an executive’s eye.

An executive who reviews every thank-you is usually protecting several things at once: the right name and amount, wording the organization stands behind, relationships they hold personally, promises they made and exceptions only they can settle. Someone else can own the accuracy check. Standard wording can be approved once for an agreed set of routine messages. Relationships and promises need the executive personally, but the note or call can follow the thank-you. Under that agreed rule, exceptions get a fresh decision before sending.

If you are not sure what the review protects, ask what it has caught in the past. The answer tells you which check to build.

The third row takes the executive’s personal touch out of the approval path. The book’s major-donor sequence already works this way, with several thank-yous in different forms from more than one person. At any scale, the same idea lets an accurate thank-you go out promptly from the person coordinating support while the executive’s note follows where the relationship calls for it, and neither waits for the other. The coordinator, who sees every record, is also well placed to tell the executive what that note could mention.

Sign each message with the name of the person who wrote it. A standard letter can carry the executive’s signature when the executive has approved both the wording and that use of their signature. A message that reads as personally written implies the signer wrote or read it. If someone else prepares it, make sure the signer sees the message and knows where replies will go.

The book also argues for more than one thank-you on practical grounds: a single message can be misdelivered or filtered, and nobody finds out. A sent email is not yet a delivered thank-you. Check bounce reports and returned mail, and for the gifts where a lost thank-you would matter most, send a second message by another route.

When the executive knows most of your donors

In a small organization the executive may know most donors personally, so the third row covers most of the file. The change is smaller but still worth making: approve the wording once, send the prompt thank-you under the fundraiser’s or coordinator’s name, and spend the executive’s review time on personal notes instead of proofreading.

If you are the only fundraiser, you hold both the coordinating and the relationship work. Give them separate blocks of time so the checks don’t wait for the notes. Any gift acceptance rule that requires executive sign-off stays where it is, in the last row.

Twelve gifts, three things for the executive

Riverbend received twelve gifts on September 14. Maya coordinates donor support, Luis leads programs and Leah, the development director, is authorized to approve routine acknowledgement copy. Anika, the executive director, is away until September 18.

Leah sorts the queue by what each record needs rather than by who usually signs.

Anika is needed for three things, one note and two calls, and none of them has to hold an acknowledgement. Ten thank-yous go out on September 16. Only the two unmatched gifts wait, and they wait for finance.

Routing all twelve through Anika would have held ten messages until her return. Marking every record complete after an email would have hidden the promised calls; sending before finance finished matching would have risked thanking the wrong people. Official receipts follow the team’s existing process; delegating these personal messages changes neither the rules for financial documents nor who may approve an exception.

Change one fact and the route changes. If Anika had told Ruth she would call, Ruth’s gift would join the promised calls. If finance matches an unknown gift to a donor Anika knows, that gift takes Ruth’s route: a prompt thank-you from Maya, then Anika’s note.

Check the gift, designation and any request against the record before sending. Add a colleague’s personal note only after they have written or reviewed it.

Send a report like this only if your authority to approve routine messages was agreed beforehand. If it wasn’t, ask before the queue goes out. Silence isn’t delegation.

Ten messages can’t give you a reliable rate, but they show what kind of review Anika is doing. Every change here has somewhere else to happen. The purpose error belongs in Maya’s record check, and the claim in the approved wording. The two personal touches can follow a prompt thank-you as Anika’s own note. The one message she held was a real exception, and the proposed rule would still send it to her first.

Your review may point somewhere else. If most of the executive’s edits add facts nobody else could know, or correct relationships the record gets wrong, fix the information that reaches the draft, or agree a fixed review time. Delegation is only one possible answer.

Record what Anika chose and when. If she doesn’t reply, the current process continues; raise it again at your next one-to-one. Silence isn’t acceptance of the rule or of the fallback.

The next check

At the 4 p.m. review, match each of the twelve records to a sent message or a named open task, and check for bounces. Keep the two calls and Ruth’s note visible until Anika has done them. If she agrees to a trial, bring the approval log to the review date and let her decide what happens next from what it shows.

A first gift arrives with more information than most teams read. The record shows what the donor chose to fund, how the gift reached you, whether it was one-time, monthly or in someone’s honor, and any question or preference they left. What it rarely shows is why they gave.

Hallett notes that some of what you want to know, such as how a gift was structured, can be looked up rather than asked. Most of the record works that way. Read it first, then ask the donor only what the record can’t tell you. The question becomes shorter and more specific, and nobody asks the donor to repeat what they already told you.

Most teams can’t follow up personally with every new donor, so someone has to choose. Put obligations first. A donor who asked a question is owed an answer whatever the gift size. That answer takes someone’s time, so count it before anything else; the hours left over decide how many personal conversations you can offer. Next come donors who told you what interests them. Hallett’s move from project to program gives that conversation a subject without an ask.

Among the rest, gift size, screening and a colleague’s judgment can reasonably decide who gets your time; Hallett uses all three. They are weak evidence of what the donor wants next. Let potential decide where your time goes, and let the donor’s own words, or your question, decide what the contact is about.

Everyone else still gets a useful welcome, and it can carry one easy question. If you ask at scale, someone has to read the answers. Before sending, write down what you hope to learn and how each likely answer would change the plan. If every answer leads to the same scheduled message, drop the question.

The order holds at any size. A sole fundraiser might answer what’s owed, hold one or two conversations a week and send the general welcome to everyone else. A larger team can read source, structure and tribute fields in bulk and save individual reading for the gifts that stand out.

Six new donors, two conversations

Riverbend’s recent first-time donors include the six below. Each has received an accurate acknowledgement. Leah, the development director, has time for two personal conversations this week, so she reads each record before choosing.

Leon and Marcus get the two conversations, for different reasons. Leon has said what he wants to understand, so the subject is set. Marcus has said nothing, so the conversation’s job is to ask. Priya’s answer comes out of Maya’s week and was counted first; if Maya had no time left, the answer would still come before either call. The Second Gift Starts Here works through a full week of this, including a question that arrives midweek.

Different facts would change the plan. If Marcus says his granddaughter’s school uses the program, Leah can offer an overview like Leon’s. If he would rather not be called, she records that and stops. If Noor asks to visit, respond to the request and let what she says next shape anything further.

Offer the overview only once Luis has agreed to prepare it on workable timing. Otherwise, acknowledge the question now and give a date you can keep for checking back. When Leon replies, record what he asked for, how he wants it and when it will be sent. His reply narrows the work; it doesn’t establish interest in a meeting or another gift.

Leon may answer, “I only wanted the receipt. Please don’t turn this into a fundraising conversation.” Respond to that without defending the plan:

Apply the preference to every list and sequence before marking the work complete. A polite response alone does not fix an unchanged mailing list.

Check the work separately from the next gift

At the next review, check that Priya got her answer, Leon got the explanation he chose and Marcus’s call was made or closed. Keep unanswered questions and changed preferences visible. Answering Priya shows that you kept a commitment; it can’t tell you whether she will give again.

Later, look at who gave again over an observation period you chose in advance. Record what you knew at the first gift and what you did, so that patterns in your own donors can emerge, as Hallett suggests. Donors chosen for conversations may have given again anyway. Compare donors with similar starting circumstances, including those you couldn’t reach because time ran out. That can reveal a pattern to investigate, but it still won’t isolate the effect of the conversation. Count monthly donors separately; a scheduled payment says little about a decision to give again.

A board member who invites a friend is vouching for the experience. If the hour is dull, or turns into a surprise appeal, the friend will remember who invited them. Reluctance to “bring names” is often a reasonable reading of that risk.

Axelrod’s own sister, an early guest, said plainly that she would not give. She came anyway, and afterwards brought her own circle of friends, because the hour had been good enough that she trusted they wouldn’t resent being invited.

That confidence is what a board member needs before inviting anyone, so their own first visit does two jobs. It gives them the firsthand understanding Axelrod asks for before anyone invites others, and it tests whether the experience is ready for guests. Axelrod asked that same sister for an honest verdict on the tour. Ask your board member whether they would be comfortable bringing a friend and, if not, what would need to change. Fix that before anyone sends an invitation.

Before anyone invites a guest, settle three things the guest will be told: what they will see, whether any request for money or time will be made, and who will contact them afterwards. Axelrod’s hosts told guests in advance that a staff member would call. That openness is what lets a board member invite without worrying about an ambush.

Use something you already run, such as an orientation, open class or program showcase. If your program can’t safely host visitors because of confidential services or young participants, offer a staff-led briefing with aggregate information instead. The same three things still need settling.

Keep this apart from commitments already made. Axelrod advises against reworking an imminent event on the strength of a new idea, and the full model takes most of a year. Try one visit alongside existing plans, and don’t rewrite anyone’s board agreement through an invitation.

The orientation gets tested first

Oak Street Reading runs a monthly orientation for prospective adult volunteer tutors. It uses program materials and aggregate information; guests don’t observe children or see participant records.

Leah, the development director, talks to four board members about introducing people to the program. Two agree to attend the September orientation first. One would rather thank recent donors by phone. One has no time this quarter, and Leah leaves it there.

Elise, one of the two, hasn’t attended since the program changed. Afterwards she says the explanation of tutor preparation finally made the work clear to her. She also says the session ended abruptly, and she wasn’t sure what a guest would be expected to do next. Leah asks the volunteer coordinator to close future sessions by explaining how to apply and who will be in touch.

Elise thinks of Sam, who asked her about volunteering last spring. Leah confirms that the October session has space and the coordinator can receive a guest. Only then does Elise send the invitation.

If Elise had said she wouldn’t bring a friend, Leah’s next question would have been why, and the session would have been fixed before anyone else was asked. If Sam declines, the invitation ends there. Sam doesn’t become a prospect because Elise suggested the name.

If Sam attends and asks about evening roles, Leah answers that question instead of sending a general appeal:

A question about volunteering hasn’t established interest in giving. Record the question and the reply date, and carry both through the handoff.

The card gives the board member what a short call needs and nothing more. Staff choose the donors, check preferences first and collect the cards afterwards.

What completion looks like

The guest got the experience described, any question they asked has an owner and a reply date, and the board member knows where their part ended. Count those before counting names gathered or invitations sent. A guest who doesn’t reply has left the invitation unanswered; that doesn’t authorize adding them to a campaign.

A right to withdraw works only if withdrawing is affordable. Someone who knows their film is the centerpiece of the dinner program may feel they can’t say no without leaving a hole, and agree to what they would rather refuse. A replacement that is approved and ready removes that pressure. It also makes a yes easier to trust.

Watanabe’s experience shows that audience means particular people as well as numbers. A guest list can help someone judge an event’s audience; public distribution is much harder to bound. Explain any recording or livestream too. Ask whether there is anyone they wouldn’t want to see or hear the story, without requiring them to name anyone. The production team needs the decision, and any personal reason can stay with the staff contact.

Agreeing to an interview isn’t agreeing to every sentence in it. Show the person the actual cut, with its captions, thumbnail and the appeal around it, and take out anything they ask to remove without asking why.

A broad release form may cover uses the person never pictured when they signed it. Ask about the specific use anyway.

A backup is ready when it is approved for this audience, the file is in the show folder, it fits the same slot and the emcee has words to introduce it. Build it around what the segment was meant to tell people rather than around a substitute speaker. Watanabe advises choosing two things you want guests to know before they leave, and the backup should still deliver one of them. It might be a staff member describing the program, an approved anonymous story, or a clearly labeled composite built from material you have permission to use. A small team without a video budget can use a short paragraph read by the executive director, with a photograph of the space and no participants in it.

Name who can remove the original from every scheduled use: the playback queue, the website, the social calendar and any email. If something has already gone out, tell the person what can be stopped or taken down and what can’t be recalled. Don’t promise that every copy will disappear.

The request that changes the job

Nia, an adult participant in Harbor Skills’ training program, has approved a two-minute film for the October 22 dinner. Her approval names that event and the final cut, which she reviewed on October 17. On October 18, Theo in communications asks to add the film to the public campaign page and make three social clips. “Her release covers all media,” he adds.

Lena, the development director, checks the request against the record. The release is broad, but Nia agreed to one dinner. The campaign page would reach people nobody can predict, the clips would cut her words down, and public copies can circulate beyond Harbor Skills’ control. Lena keeps the dinner plan and pauses the online uses.

She shows Nia the draft page and the three clips, and asks whether there is anyone Nia wouldn’t want to see them. Nia is comfortable with the invited dinner audience and its agreed no-recording arrangements, but she hasn’t told her current employer about her time in the program. She declines online use. Lena records the decision for the production team and keeps the reason to herself.

The dinner already has a replacement. The program director describes the training, using verified activities and an approved photograph of the workshop with no participants in it, and the segment carries the evening’s first message, that the program trains adults for skilled work. Nia could withdraw from the dinner film too without leaving an empty slot.

Tell Nia it’s done only after Devon confirms the film is out of the playback queue, then give the emcee the words for the replacement segment.

Retention starts from last year’s donors and asks which of those same people gave again. A common shortcut starts from this year’s donors and subtracts the new ones. Subtraction works if you remove everyone who didn’t give last year. Under FEP’s definition, a new donor has no earlier recorded gift, so subtracting only new donors also leaves people who gave in an earlier year, skipped last year and came back. They are real donors, but they were never in last year’s group.

The shortcut overstates retention by exactly the number of returning donors divided by last year’s total. The error grows with every lapsed donor you win back, so a good reactivation year can make retention look healthier than it is.

Report returning donors as a result of their own. Inside the shortcut they inflate one number and get no credit.

When two reports disagree, ask their owners for the donor lists behind each figure, the period boundaries and how much history they checked. They may be answering different questions. Keep both counts if both questions matter; don’t settle the disagreement by picking the larger percentage or relabeling a database measure as annual retention.

With a small cohort, report the count beside the rate. In a fixed cohort of 80, each additional prior-year donor who gives again moves retention by 1.25 percentage points. A five-point change is four people.

Calendar boundaries create absences of their own. A donor who gave on December 20, 2024 and again on January 5, 2026 gave about a year apart but counts as absent from 2025. Check for gifts received just after a period closed before calling anyone lapsed.

Two reports, two different questions

Harbor Skills had 96 donors in 2025, up from 80 in 2024. Its finance pack reports 80% retention; the development dashboard reports 50%. Before recommending another campaign, operations lead Jules asks both owners for the donor IDs and counting rules behind each figure.

The finance pack took the 96 current donors, subtracted 32 first-ever donors and divided the remaining 64 by 80. Its “new” meant first-ever, which is right for counting new donors and wrong for this subtraction. The 64 included 24 people who gave before 2024, skipped that year and returned in 2025.

The dashboard matched the previous year’s people: 40 of the 80 gave again. After confirming that both exports used the same identity, date and gift rules, Jules reports retention of 40 ÷ 80 = 50% and donor growth of 16 ÷ 80 = 20%. The gap of 30 percentage points between the reports is the 24 returning donors divided by 80.

The export had several gift rows per person. Jules removed duplicates only after settling the donor identity rule. Deduplicating by name would have been unsafe: two people can share a name, and one person can hold more than one account.

The 40 prior-year donors without a 2025 gift make a review list. Jules’s first pass, using the export from January 9, 2026:

The calculation shows which records need attention. It can’t say why anyone is absent, or make another contact appropriate.

Harbor Skills counts individuals. FEP’s definitions also allow households or organizations; whichever you choose, apply it to both periods and say so in the report.

If identities remain unresolved, withhold the rate and name who is checking the mapping. If the prior cohort is empty, report “No eligible prior cohort”; a 0% rate would mean something different. For an unfinished year, name the year-to-date periods and use the same cutoff in both; don’t label the result completed annual retention.

Questions about money often arrive inside a thank-you exchange, so they tend to get thank-you answers: prompt, warm and vague. Treat each one as a request for a specific record. The reply can only be as precise as the record behind it.

One person owns the reply even when others supply the facts. When finance provides the figures, name the colleague who will check the attachment and write to the donor. Answer the confirmed parts, and keep an unfinished part open with a date.

For the first question, how specific you can be depends on what your records show:

“Please keep my name private” needs a scope before anyone can honor it. The donor may mean off published donor lists, not named at events, hidden on online giving pages, not shared with other organizations, or all of these. Ask if it’s unclear. Record each choice separately from contact preferences; someone can choose public anonymity and still want receipts and program updates.

Then find every place the name appears or is due to appear: printed and online donor lists, event programs and signage, donor walls, social posts and the donor feed on an online giving or peer-to-peer page. Some may already be live. Confirm each change with the person who owns that list, and describe only what was actually done.

Three requests inside one email

Rosa gave $500 to Northbank Reading on September 8 through the online giving page, after the September appeal. On September 9 she asks three things: what her gift pays for, whether she can see the latest financial statements and whether her name can be left off the public thank-you list.

Hana, the donor support lead, separates the work and keeps the reply. Eli in finance checks the gift’s designation and which statements are the latest completed set. Communications checks where Rosa’s name appears. Hana will answer Rosa by September 12.

The gift form and receipt record the reading program. Eli checks the September appeal too: it used $50 for ten books as an example cost and said that gifts support the whole program, including tutors. The budget includes tutor time and reading materials, and the records don’t trace any single gift to a particular purchase. Rosa could still picture her $500 as a hundred books. Hana explains what the example cost meant and what the records show.

Communications finds Rosa’s first name and initial in the giving page’s public donor feed, and her full name on the draft list for the annual report, which hasn’t been printed. Both are removed. Rosa’s existing request for program updates stays as it is.

Check what the appeal promised

Use the pooled-program explanation only if it matches the appeal and the gift record. If the appeal promised a narrower use, don’t describe that promise as an example cost. Ask finance and the responsible fundraising lead to reconcile what was promised, recorded and spent before answering.

Do this while the appeal is still a draft. Every example cost you keep needs a current figure and a line saying what gifts actually support, unless your accounting really does link gifts to purchases. Had the September appeal been checked this way, Hana’s reply to Rosa would have confirmed what the appeal said instead of correcting it.

If Rosa had asked for identifying details about the students, Hana would offer approved program-level information and explain the boundary. Answering one person’s question shouldn’t disclose someone else’s story.

If one person handles both the finance records and the reply, the separation still helps: check the record before you write the sentence it supports.

Every published result has a setting: one organization, one group of people, one change and one response counted over a set period. Headline figures tend to travel without it. In the five field experiments below, the average sometimes hid how differently groups of donors responded, and in one the response counted wasn’t money.

Before you borrow a message, find out who received it, what changed and what was counted, for how long. Then decide what you’d test, and what you’d keep doing whatever the answer.

Describe a loss only when it’s real

Tanner (2026) compared two campaigns to about 4,000 previous donors of the Charles Darwin Foundation. One reported an extinction, the other the rediscovery of a species. Roughly 2.1% gave to the extinction appeal and roughly 1.2% to the rediscovery appeal. The campaigns described different real events, so this compares two true stories rather than one story reworded as a loss, and at this list size each rate rests on few gifts.

If a service will really end on a date, say so and keep the evidence with the draft: who confirmed it, and when. If nothing will be lost, describe what a gift makes possible. Don’t make donors responsible for a loss nobody has confirmed.

Say what a priority choice does

Kessler, Milkman and Zhang (2019) mailed about 32,000 previous donors to a university annual fund. Reply cards for a randomly chosen group invited alumni to mark which of four priorities mattered most; every gift stayed unrestricted whatever they marked. Among alumni in the highest-income census tracts, those who gave gave about 2.4 times as much as similar donors with the standard card, with no statistically significant increase in the likelihood of giving. That’s one subgroup, identified by neighborhood income, at one university.

If you invite a choice, say what it does: restricts the gift, informs planning or shapes what you report back. For example: “Tell us which part of our work matters most to you. Your gift supports all of it, and your answer decides which program updates we send you.” Use that only if finance and the program team have agreed it’s true.

Treat a survey answer as an intention

Godin and colleagues (2014) assigned 7,000 lapsed blood donors to six questionnaire versions or a control group, then counted blood-drive registrations over 6 and 15 months. Some questionnaire versions increased registration frequency. About 23% of recipients with valid addresses returned one. The researchers compared everyone originally assigned to each group, including nonrespondents. Returning the questionnaire was a choice, so a comparison of respondents alone would mix the effect of the mailing with who chose to reply.

This is evidence about giving blood; it can’t forecast cash gifts. When a donor replies “I plan to give again this year,” record a stated intention, keep it apart from pledges and leave it out of revenue forecasts. If you test a survey, compare giving from everyone assigned to receive it with everyone who wasn’t, never respondents alone.

Test an ask inside a thank-you by giving history

Sudhir, Fong and Roy (2019) varied HelpAge India’s annual thank-you letter to roughly 200,000 previous donors. Across everyone, adding an explicit request for another gift had no statistically significant effect. Split by history, recent, frequent and higher-value donors gave less when asked, while lapsed, infrequent and lower-value donors gave more. Their estimate that targeting the request could raise donations by 6–11% is a back-of-envelope calculation, not a measured result.

Leave receipts out of any test like this; they go promptly and unchanged. Don’t turn the lapsed-donor result into a rule to ask every lapsed donor. Define your groups before sending, and read results within each group as well as overall.

Make a contact promise you can keep

Damgaard and Gravert (2018) emailed about 43,500 donors on DanChurchAid’s list. One version added a line promising a single email in the next three months, where the usual was monthly. Within three days, unsubscribes were 0.30% in that group against 0.49% for the usual message: 0.19 percentage points lower, or about 39% in relative terms. Giving didn’t differ significantly. Monthly direct-debit donors weren’t included.

That’s evidence about three days of unsubscribes. It doesn’t make quarterly contact right for everyone or show that fewer emails raise more. First check every sending queue: newsletters, events, renewals and automated reminders. Make the promise precise: “We’ll send you one fundraising email between [start date] and [end date]. Receipts and replies to your questions will come as usual.” Use dates your team can keep.

Turn a finding into a local test

Choose one change and donor groups defined before you send. Count one primary measure for everyone you assigned, including people who never opened or replied. Fix the window and decide in advance what you’ll do if the result is unclear. On a small list, a difference of a few gifts is a reason to look closer; don’t call it a winner.

Riverbend’s thank-you test

Riverbend Literacy sends donors an annual thank-you email in November. This year Leila, the development director, wants to find out whether adding an invitation to give helps or hurts, and whether the answer differs between donors who gave recently and donors who haven’t given for a while.

She writes the plan before anyone is assigned. Recent donors gave in the 12 months before October 26; older donors last gave more than 12 months and up to 36 months before that date. Sam checks contact preferences and duplicate identities. This test uses individual donors; joint households, monthly donors and relationships managed personally receive their usual separate letters. Anyone who asked not to be emailed or solicited is excluded. Their preferences still govern any stewardship.

That leaves 1,000 recent and 1,000 older donors. Sam randomly splits each group in half, so each version goes to 500 recent and 500 older donors: 1,000 people. Version A is the usual thank-you. Version B is identical except for one added paragraph inviting a gift toward a new Saturday reading session. Both go out at the same time from the same sender. Receipts for any new gift go out as they always do, and Omar in donor care answers every reply within two working days, whichever version the donor received.

The first count

The observation window closes on November 23. Sam matches every gift received in the window, from any channel, to the version each donor was originally assigned. For the giving rate below, he counts each donor once, even if they made several gifts.

Overall, version B looks 10% better: a giving rate of 5.5% against 5.0%, half a percentage point, which is five more donors out of a thousand. Among recent donors, 25 gave after B and 40 after A. Among older donors, 30 gave after B and 10 after A. The overall result combines two groups moving in opposite directions.

These are counts from a single send. Five extra donors overall doesn’t establish a winner, and 15 fewer recent donors giving doesn’t prove the invitation put them off; a gap can come from chance. The split resembles what Sudhir, Fong and Roy (2019) reported, which makes it worth checking. It doesn’t confirm a rule for Riverbend.

Leila’s decision

Version B brought in $375 more net revenue overall. One donor in its older group gave $750 net, twice that margin. Keeping that gift in the primary result, Sam also shows the comparison without it: $5,375 − $750 = $4,625, or $375 less than A’s $5,000. The reversal shows how much one gift influences this result. It is not a reason to discard the gift, and it does not measure the uncertainty of the test.

All 13 unsubscribes arrived in the first three days. One recent donor assigned B included a complaint with their request; no other complaints or separate no-contact requests were recorded by November 23. Omar honored every request and confirmed that no further messages went to those donors. The contact figures merit attention, but they do not establish a lasting effect on donor relationships.

Leila records the test as inconclusive and keeps the usual thank-you. Before considering an older-donor test next year, she reviews the opt-outs and complaint with Omar. She does not turn either group’s result into a permanent rule.

Reading the result

Report every result against the version each donor was originally assigned, including people who never opened the email, unsubscribed later or gave by post. A comparison of respondents alone describes who chose to respond.

If someone asks not to be contacted after assignment, remove them from every remaining send straight away and keep them in the report under their original version. Dropping them from the count would hide exactly the reaction a guardrail is there to catch.

If a send normally produces only a few gifts, treat the test as exploratory. Set the window and review date before sending, report the counts beside every rate, and don’t call an early count a winner.

The finding in its setting

Tanner (2026) compared two Charles Darwin Foundation campaigns sent to 3,983 previous donors. One reported that a species had gone extinct; the other that a species believed lost had been found. About 2.1% of donors gave to the extinction campaign and about 1.2% to the rediscovery campaign.

Both campaigns reported something that had happened, to people who already supported that conservation work. The study didn’t reword one program as a loss, and it gives no reason to describe a loss nobody has confirmed. Nor does it show that loss framing doubles giving for other organizations. It supports a narrower point, that a true account of a loss can draw a response from people who already care about the work.

Three things that get called a deadline

Appeal drafts often treat three different situations as one. A program consequence means a specific service will stop, shrink or not start on a known date unless money arrives, and someone with authority has said so. A finance shortfall means the budget is short. That’s real, but a gap in the general budget rarely means one particular service ends on one particular day; the organization may draw on reserves, cut elsewhere or finish the year in deficit. A calendar date is one the campaign chose: the end of the tax year, the last day of the appeal, the day the printer needs copy.

Each supports different words. Only a confirmed program consequence lets you say something will be lost. A shortfall lets you describe the gap and what gifts will help with. A calendar date lets you ask for a gift by that date and say why you chose it, without implying anything happens to the program afterwards.

Where the money goes

A second claim usually travels with the deadline: what a gift buys. “Gifts to this appeal support our after-school program” and “Your $500 gives a child a place” describe different arrangements. The second implies the gift is spent on that place, which is only true if your organization records and uses it that way. Ask finance how gifts made through this appeal will be recorded before you choose the wording.

If gifts go to a general program fund, you can still show what things cost. “A Saturday place for the spring term costs about $500” gives donors a sense of scale, provided the same paragraph says gifts support the whole program. Write the calculation into the brief and have finance check it, so whoever answers a donor’s question can show where the figure came from.

Plan for the facts to change

The facts behind an appeal keep moving after the copy is approved. A grant arrives, the board changes its plan, a building fails an inspection. Before sending, decide who will tell you if the consequence changes, what you’ll send instead and how you’ll correct anything already out. Write the replacement version at the same time as the first, while the facts are fresh, and get it approved in the same sign-off.

The sentence in the draft

Millbrook After School runs a weekday program for 60 children, Monday to Thursday, funded through June by a city contract and two foundation grants. In September it added a Saturday morning session for 24 children whose schools had referred them for extra help with reading and maths. A one-off grant pays for Saturdays until December 12.

On November 4, Ama, the development director, receives the year-end email draft. Its opening line reads: “This winter, 60 children could lose their after-school place unless you give by December 31.” The appeal is scheduled for November 17. Ama opens a claim brief and checks each part of the sentence with the person who would know.

The cost, and the condition

Jonah and Rosa cost the spring term together. The funder of the autumn grant has agreed in writing that $3,000 left unspent can be used in the spring.

The board minute sets the condition. If $12,000 is confirmed by December 8, all 20 Saturdays run. If less is confirmed, Millbrook will run as many full sessions as the money covers, at $600 each, as long as that’s at least ten. With less than $6,000 confirmed in total, which means less than $3,000 raised, the Saturday session ends on December 12 and families hear that week.

A spring place costs $12,000 ÷ 24 = $500. The appeal can use that figure to show scale, in the same paragraph that says gifts support the whole after-school program.

Ama’s decision

The appeal goes out on November 17 about the Saturday session, with December 8 as its date and the board’s condition stated plainly. December 31 comes out of the copy, along with the 60 children. Jonah and Rosa agree to tell Ama the same day if a grant decision or anything else changes the gap. The replacement version is approved in the same sign-off, so the team can switch without another round of review.

When the funding arrived

On November 30, the Pell Family Fund confirmed in writing a $4,000 grant for the spring Saturday session. With $5,350 already given to the appeal, confirmed spring funding reached $3,000 + $4,000 + $5,350 = $12,350, a week before the date in the appeal. Dee confirmed that the board’s condition had been met.

Ama followed the brief. The giving page changed on November 30 to say the spring term was funded. The December 6 reminder, which repeated the deadline, was cancelled. The update below went out on December 1 to appeal recipients who could still receive it under their contact preferences. The grants and $5,000 of the appeal gifts covered the Saturday cost. The remaining $350, and any appeal gifts received after November 30, went to the weekday program, under the terms the appeal had described. Donors still heard the board’s decision on December 11, as promised.

Suppose a donor gives $250 with “Family reading workshops” ticked on the reply card. Six months later, they ask how many workshops the gift paid for. If the form never said what the tick did, there’s no good answer. The gift went into the general fund, the tick sits in a field nobody reports on, and the donor had reason to assume otherwise.

A priority box can do three different things. Each needs its own sentence on the form and its own person behind it.

The finding in its setting

Kessler, Milkman and Zhang (2019) randomly assigned 32,174 previous donors, alumni of a university, to receive the standard reply card or one that also invited them to mark which of four priorities mattered most. Gifts stayed unrestricted whatever they marked. Among those in the highest-income census tracts who gave, gifts were about 2.4 times as large with the choice card; the study found no statistically significant increase in the likelihood of giving. A census tract describes a neighborhood, so it’s only a rough guide to any one person’s wealth. The authors note that some alumni may have believed their choice directed their gift.

The experiment doesn’t establish what every donor understood. If you offer a choice, say plainly what it does.

Kestrel’s new reply card

Kestrel Reading Project runs three things: one-to-one tutoring for children in grades 1 to 4, books children take home to keep, and evening reading workshops for families. Gifts go to the general literacy fund, which pays for all three. Kestrel has never offered a restricted option on its standard forms.

Marisol, the development director, wants to add a priority question to the spring reply card and the online form. Before drafting, Marisol meets Ben, the program director, and Yusuf in finance with three questions: what will the answer do, who will do it, and what must the form not imply.

They agree that the answer will decide which program stories go to donors who have asked for updates, and that Ben will receive the totals each June. It won’t change where any gift goes. Ben commits to one short update per program area every six months, so each option has something behind it. Yusuf asks to see any note requesting a restriction before anything is recorded, because Kestrel’s gift acceptance policy requires finance approval for a restricted gift.

Five replies from the first week

The spring appeal goes out on March 2. Marisol picks five of the first week’s replies that show the decisions the team now has to make the same way every time.

Joan’s gift

On the call, Marisol explains that Kestrel gifts normally support all of its reading work, and asks what Joan would prefer. Joan wants the $500 used only for books children keep at home. Yusuf had confirmed on March 6 that Kestrel can accept a gift restricted to home books under its policy, so Marisol agrees, and sends the handoff below the same day.

What Kestrel left out

Marisol’s first draft of the card said “Choose where your gift makes the biggest difference.” Read on its own, that tells the donor the gift will go where they choose, which wasn’t true for any of the options. It came out.

The team also kept the interest field away from the appeal schedule. A donor who ticks Tutoring hasn’t asked for the tutoring appeal and no other, and hasn’t agreed to more email.

When the answer changes or doesn’t fit

If a donor ticks something different next year, the new answer is current. Keep the old one with its date; the donor may mention the change, and it helps to know.

If a program area pauses, tell the donors who chose it and asked for updates, and ask whether they’d like a different focus. If a donor asks how “their” program used the gift, answer truthfully: the gift went to the whole program, and here is what that program did.

If the donor doesn’t answer a restriction query by the agreed date, finance decides the next step under the gift acceptance policy. Keep the original instruction, assign a person and a next review date, and record any contact the donor’s preferences allow. Silence doesn’t turn the gift into an unrestricted one. Recording the money and settling its use are separate tasks; follow finance’s policy for both.

The development director owns the form and the six-month check. The program director owns update content and receives the totals. Finance owns every restriction decision. Whoever processes gifts owns the fields and the review flags.

Check after six months

Count the notes that asked for a restriction and how quickly each was resolved, and read every donor question about where a gift went. If several donors thought they had chosen where their money went, rewrite the sentence on the form.

A higher average gift after adding the question doesn’t show the question caused it. To find out, send the two cards to randomly assigned halves of the same mailing and compare everyone assigned, including people who didn’t give.

The finding in its setting

Godin and colleagues (2014) randomly assigned 7,000 lapsed blood donors to one of six questionnaires or to a control group that received none, then counted registrations at blood drives over the next 6 and 15 months. Some questionnaire versions increased registration compared with the control group. About 23% of people with valid addresses returned a questionnaire.

The outcome was blood-drive registration, so the study cannot tell you how much money a cash-donation survey would produce. Returning a questionnaire was a choice, which is why the comparison that counts is everyone assigned against everyone in the control group. People who reply differ from people who don’t in ways no mailing caused.

How an answer turns into a number

Survey answers drift as they move through a team. “I plan to give again” becomes “likely donor” in the database, then “41 expected renewals” in the board pack, then a revenue line built on the average gift. Each step seems reasonable to whoever takes it, and none of them is something the donor said.

The donor answered one question, on one date. That’s what the record should hold, with the wording of the question. Anything staff infer goes in a separate note and doesn’t change the donor’s status. Permission to follow up is a separate answer too; a donor can plan to give and still not want a call.

Give every answer a recheck date

An answer describes one moment. Set a recheck date when you record it, such as the end of the period the question asked about. After that date it leaves the active list and stays in the donor’s history. Nobody should stay a “likely donor” because of a box they ticked two years ago.

To find out whether the survey changed giving, hold back a randomly chosen group that doesn’t receive it, and fix the dates before sending. Then compare the share of each whole group who gave, counting each donor once. The reply rate and the giving rate have different denominators. Keep them apart in every report.

Fernwood’s check-in survey

After Femi checks duplicate identities and excludes donors whose contact preferences do not permit this survey, Fernwood Community Pantry has 1,200 eligible individual donors whose last gift was 13 to 36 months before March 1. Rhea, the development manager, wants to ask them whether they’re thinking of giving again. Femi, the operations lead, wants to know whether asking makes any difference.

On March 2, Femi randomly splits the 1,200 into two groups of 600 and saves the file. One group receives the survey on March 3. The other doesn’t. Both groups get Fernwood’s usual communications. The comparison window runs from March 3 to August 31, which is also the recheck date for every answer.

96 of the 600 replied, a reply rate of 16% of everyone sent the survey. Six replies mentioned an amount in the comment box, such as “probably $50 again.” Rhea copies each comment into the donor’s record word for word and adds none of them up.

Rhea owns the follow-ups and works through the 37 by the channel each donor chose. Two donors who planned to give gave online before Rhea reached them; their follow-ups were cancelled and they received the usual receipt and thanks. Of the 35 Rhea contacted by March 31, four replied “no thanks,” and their records now say so. The 22 who said “not right now” get no personal ask; the one who asked about volunteering in their comment gets the information they requested.

The April report

Rhea’s first draft for the April 14 finance committee counted the 41 “plan to give” answers as expected renewals at the group’s average past gift of $85: 41 × $85 = $3,485 in the spring forecast. Femi took it out. The answers did not establish 41 gifts of $85. Some comments mentioned possible amounts, but none created a commitment through Fernwood’s pledge process. The $3,485 was Fernwood’s estimate, presented as donors’ commitments.

The report gives the survey counts in full, reconciled to all 600 recipients, and leaves revenue as gifts received. Two donors made written pledges after their follow-up; finance recorded them through its usual pledge process, and the report counts them there.

August 31: did the survey make a difference?

Femi counts, for each group of 600, the donors who made at least one gift between March 3 and August 31, from any channel, counting each donor once. Everyone stays in the group they were assigned to, including bounces and people who never replied.

This comparison covers the survey and the agreed personal follow-ups it triggered. It cannot isolate the effect of asking the question. Eight more donors gave in the survey group. With 600 in each group, a gap that size could easily come from chance, so Femi reports a small difference that doesn’t settle whether the survey helped. Repeating the same design next year would add evidence.

The 39 survey-group donors who gave include 22 who said they planned to, 7 who weren’t sure, 1 who said not right now and 9 who never replied. So 22 of the 41 who planned to give did give, which is 54%. Femi doesn’t set that beside the 5.2% in the no-survey group. Those 41 chose to reply and chose that answer, and many would probably have given anyway. The 54% describes them and says nothing about what the survey did.

On August 31 every answer reached its recheck date. The 19 donors who said they planned to give and haven’t are no longer flagged, and their March answers stay in their history.

Who does what

Whoever processes replies records each answer word for word and opens a follow-up task only where the donor ticked a box. One named person owns each follow-up and its due date. Finance owns pledges, through whatever process your organization already uses; the survey doesn’t create a new kind of commitment. Whoever reports to leadership owns the reconciliation, the comparison and the recheck date.

The finding in its setting

Sudhir, Fong and Roy (2019) varied HelpAge India’s annual thank-you letter to about 200,000 previous donors. Across everyone, adding an explicit request for another donation had no statistically significant effect on giving. Split by giving history, the response differed. Donors who had given recently, often or in larger amounts responded less well to the request, while lapsed, infrequent and smaller donors responded better. The authors estimate that sending the request only to the groups that responded well could raise donations by about 6–11%, a back-of-envelope calculation from their results rather than a separate test.

The letter was a once-a-year thank-you. The study shows that giving history can change how a request inside one is received. It doesn’t give you a permanent rule about whom to ask.

Four messages that all start with thank you

Most teams send several kinds of message that open with thanks. They have different triggers and owe the donor different things. This guide considers an invitation inside the annual thank-you; decisions about welcome messages belong with the welcome series.

Eligibility comes before assignment

Two questions get mixed together in a thank-you test. Eligibility asks whether this donor should receive an invitation at all, given everything else going on with them. Assignment asks which eligible donors get which version, and it should be random if you want to learn anything.

Decide eligibility from facts in the record: the donor’s preferences, how and when they give, whether a payment is failing or a pledge instalment is due, whether someone owes them a reply or a report, and what else is scheduled to reach them in the next few weeks. Don’t decide it from who you think will give.

Then define your comparison groups by date, with a reference date, before anyone is assigned. People who unsubscribe or ask for fewer emails after assignment leave every later send at once and stay in the analysis under the version they were assigned. Dropping them from the count would hide the reaction you most need to see.

Linden’s November letter

Linden Neighbors pairs volunteers with older people who live alone and runs a shopping minibus for them. Each November it emails donors an annual thank-you with a short account of the year. Its year-end appeal on November 24 and Giving Tuesday email on December 1 use the donor list frozen on October 31: donors who gave between November 1, 2024 and October 31, 2026.

This year Tomás, the development director, proposes adding a paragraph to the November 10 thank-you inviting a gift toward a second minibus. On November 2, before any talk of a test, Tomás asks Aisha in donor care for the contact timeline of every donor with open activity. The first one Aisha brings back belongs to Irene Kowalski.

Irene has three messages planned in three weeks, at least two of them requests for money, and an answer that is almost two weeks late. The proposed invitation is about the shopping service, the very thing Irene asked about. Sending it before the answer would tell Irene the organization had time to ask but not to reply.

Tomás pauses the invitation for Irene. Aisha gets the answer from finance and sends it with the annual report by November 6. Irene receives the plain thank-you on November 10. The November 24 appeal goes ahead, because Irene’s preferences allow appeal emails.

From one record to the whole list

Aisha applies the same checks to all 2,400 donors who gave between November 1, 2023 and October 31, 2026. Each donor goes into the first row that applies, so nobody is counted twice.

A decision for each group

Tomás defines the groups before anyone is assigned. Current donors made their last gift between November 1, 2024 and October 10, 2026; lapsed donors made theirs between November 1, 2023 and October 31, 2024. Of the 1,500 eligible, 1,100 are current and 400 are lapsed.

Aisha assigns the lapsed donors on November 3 and saves the file read-only. Gifts from November 10 to December 8 will be matched to the version each donor was assigned, and the team will review on December 15. Tomás uses the completed appeal test linked below to show colleagues how to read the result group by group, against everyone originally assigned.

Check before you press send

Ask a colleague who wasn’t in the planning to pick five eligible donors at random and read each one’s contact calendar for the next six weeks. If any would receive a request before an answer they’re owed, or two requests close together, fix the list before sending.

After the window, confirm that every receipt went out on its normal timing and every reply was answered within the agreed time, in both versions. Development owns the decision sheet, donor care owns open issues and replies, and finance owns receipts.

The finding in its setting

Damgaard and Gravert (2018) emailed about 43,500 people on DanChurchAid’s list. One version added a postscript promising a single email in the next three months, where the usual pattern was monthly. Over the first three days, 0.30% of that group unsubscribed against 0.49% for the usual email: 0.19 percentage points lower, or about 39% fewer in relative terms. Giving didn’t differ significantly. Donors who gave monthly by automatic payment weren’t included.

The tested promise covered email in general. A promise about fundraising email only, like the one below, is narrower and wasn’t what the study tested. The study doesn’t show that quarterly contact suits everyone or that fewer emails raise more money.

One promise, many senders

A development team can send one appeal and still break the promise, because the events team sends a ticket invitation, a relationship manager sends a personal ask and the platform sends an automated reminder. Each team sent one email. The donor received four.

Classify every message by its purpose, whatever the sending team calls it. Then check each donor’s delivery history across every system that can send. A combined contact history tells you whether the promise held; a count of templates per team doesn’t.

Writing the promise

Use exact dates, name the kind of email covered and list what will still arrive. If the donor’s own words were broader, such as “just one email please,” either honor that for all email or ask what they meant before applying a narrower promise.

Keep each preference with its source and date, apply the latest one and keep the history. A preference outranks anything staff have said they’ll send. If a relationship manager offered to send a proposal and the donor has since chosen fewer fundraising emails, the proposal becomes the one email or waits until the period ends.

Ruth’s preference, and what was already scheduled

Cedar Valley Hospice’s preference page lets donors choose “One fundraising email between November 1 and January 31.” Ruth Adair, who has given every December for six years, chooses it on October 20 and receives the confirmation below. Ruth’s record also shows a September conversation in which Dan, a relationship manager, offered to send details of the family room renovation in November, and Ruth said that would be fine.

On October 23, Keisha, the operations manager, pulls every message scheduled to reach Ruth between November 1 and January 31 from every system that can send email.

How Keisha resolved it

Dan’s proposal becomes the one fundraising email. Ruth had agreed to receive it, it’s the most specific of the six fundraising messages scheduled, and it comes from someone Ruth knows. Keisha asks Dan to keep it to one message. If Ruth replies with questions, Dan answers them, but there’s no follow-up nudge unless Ruth asks for one.

Lin’s platform can’t remove the appeal section for individual readers, so Lin sets up a news-only edition, with no appeal and no donate button, for donors with this preference. In December, when the appeal leads the newsletter, the edition is suppressed for them.

Keisha adds Ruth to the suppression list in every system that sends: the email platform, the events platform and the automation rules. Each queue owner confirms in writing that it’s applied. Dan’s calendar entry now reads “Ruth: one fundraising email agreed, November 18.”

The audit

Keisha reads delivery events every Monday during the period, and again on February 2. Drafts and scheduled sends don’t count. Keisha reads delivery logs from the email platform, the events platform and the CRM’s record of staff email, and asks relationship managers to confirm anything sent outside those systems.

The December 3 reminder came from a weekly contact import into the events platform, and the import didn’t carry suppressions. Keisha finds it on the December 7 check. Ruth received two fundraising emails instead of one. Keisha fixes the import, checks the other donors with the same preference and finds eleven more who received the reminder, and tells Paul and Dan that day.

Ruth’s preferences still allow email, so Keisha sends a short apology with no mention of the concert or the appeal and no request of any kind. For the one affected donor who had since asked for no email at all, Keisha fixes the cause and records the breach without writing.

Ruth’s December gift doesn’t restart anything. The receipt went out as usual and the promise ran to January 31. The February 2 audit found no further breaches.

Who owns it

One person owns the register and the audit, usually in operations. Each queue owner confirms suppression in their own system, and confirms again after any import or list rebuild. Relationship managers log personal emails in the CRM or tell the register owner, because personal email often bypasses the systems the audit reads.