The determination made before this study mailed, and the reasoning behind it.

01 · What was determined

Minimal-risk observational research, held to the stricter of the two available readings.

The formal frameworks for human-subjects research in the United States — 45 CFR 46, the Common Rule — do not reach this study. It is privately funded by The Sweet Note, Inc., carries no federal support, and the company holds no Federalwide Assurance. No regulator, institution, or review board has authority over it.

I have held it to that framework anyway, because it is the best-articulated standard available for a study of this shape. Read against it, the study has the shape of minimal-risk observational research — the category the Common Rule describes at expedited category 7, research on individual or group behaviour presenting no more than minimal risk.

I claim no exempt status. There is an exemption in the rules for observation of public behaviour where subjects cannot readily be identified. It does not fit. This is a private response to a private piece of mail, and I can identify who was mailed — I have to, in order to mail anything at all, to send the debrief afterwards, and to honour a request to be removed. Claiming that exemption would contradict my own retention schedule.

The study is also conducted in accordance with the ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics (2025 revision), which is the self-regulatory code that does govern research of this kind in this profession.

02 · Respect for persons

Nobody consented in advance, and there was no version of this study where they could.

Telling someone before mailing that their scan is the outcome of a gift experiment destroys the thing being measured. There is no design in which upfront consent and an unprompted behavioural measurement both exist. So the study uses deferred consent with a debrief, which is the standard structure for behavioural research of this kind — and the burden that structure carries is that the disclosure has to be real, not nominal.

The Code this study is conducted under provides for this. Article 4(a)(iii) of the ICC/ESOMAR International Code requires that participation be voluntary and based on clear and accurate information about the general purpose and nature of the research, given at the beginning — and then says that where that is not possible because of the methodology, data subjects must be informed at the earliest opportunity. That is the clause this design runs on, and it sets the test: not whether disclosure eventually happened, but whether it happened at the earliest point the method allowed. What follows is my answer to that.

Three things are offered instead of prior consent:

• A disclosure available from the first moment. The full methodology is published at About this outreach, linked from the reveal page, readable the day the envelope arrives. It carries the study design, what is measured, what is not, the retention schedule, and an opt-out address.
• A choice at the point of engagement. The reveal page's consent notice states plainly that the outreach is also a research study, and says what declining does and does not change. Declining stops the reaction and the button tap being recorded. Whether the envelope was opened is still counted, and the notice says so.
• An unconditional way out, by either route. Every recipient who did not already opt out receives a debrief letter about two months after the mailing, and any of them can reply and have their record removed from the published dataset and the published event log, no reason required. An opt-out counts as that request too — you do not have to ask twice, and you do not have to wait for the letter.

The framework's own test for doing it this way, and all five of its conditions

The Common Rule — which, as section 01 says, does not reach this study and which I am holding it to anyway — describes deferred consent with a debrief at 45 CFR 46.116(f)(3), and sets five conditions on it. The provision is addressed to a review board, and no board has looked at this study; what follows is therefore the standard I measured myself against, not a finding anyone made. It is quoted in full.

45 CFR 46.116(f)(3) · requirements for waiver and alteration

In order for an IRB to waive or alter consent as described in this subsection, the IRB must find and document that:

1. The research involves no more than minimal risk to the subjects;
2. The research could not practicably be carried out without the requested waiver or alteration;
3. If the research involves using identifiable private information or identifiable biospecimens, the research could not practicably be carried out without using such information or biospecimens in an identifiable format;
4. The waiver or alteration will not adversely affect the rights and welfare of the subjects; and
5. Whenever appropriate, the subjects or legally authorized representatives will be provided with additional pertinent information after participation.

Quoted verbatim from 45 CFR 46.116(f)(3) as in force on 1 August 2026. Source: Office of the Federal Register, eCFR.

Condition (iii) is the one that bites. This study holds identifiable information, and it holds it because mail cannot be delivered to an unidentified address — and because neither the debrief nor the removal right could be honoured without a list of who was mailed. Withholding the research purpose is not what makes the study identifiable; physical mail is, and no design that mails anything to anyone avoids it. What the study does instead is bound it: the address is an office — the advancement or development office — rather than a person, so no individual's name is ever collected, stored or mailed at any point; what is measured is keyed to the random code from the moment it is captured, never to a name; and the link between the two is cut on a published schedule.

The other four, briefly. (i) The intervention is a piece of mail and the observation is whether a printed code was scanned — argued in full in section 03. (ii) Telling a recipient beforehand that their scan is the outcome of a gift experiment destroys the outcome it measures; there is no version of this design in which upfront consent and an unprompted behavioural measurement both exist. (iv) A recipient of this mailing has more say than a recipient of an ordinary one: an opt-out address readable the day the envelope arrives, a choice at the point of engagement, and an unconditional removal right afterwards. (v) The debrief letter, about sixty days after mailing, to every recipient who did not already opt out — including everyone who never scanned.

03 · Beneficence

What the recipient bears is an envelope. What is measured is whether they opened it.

The intervention is a piece of mail. The observation is whether a printed code was scanned, and what happened on one web page if it was. There is no deception about who is writing or why — the outreach is a commercial approach from a real company with a real product, and it says so on the card. What was withheld until the debrief is that the response was also being measured as research.

Three risks are foreseeable, and none of them is more than minimal:

• Unwanted mail. A recipient may find the envelope an intrusion. It can be recycled at no cost, and the opt-out address stops anything further.
• Discomfort at discovering the research. A recipient may dislike learning they were studied. This is the risk the disclosure page, the consent notice and the debrief exist to address, and the removal right exists to answer.
• A sense of obligation created by the gift. This one is the measurement target rather than a side effect — the study exists to find out whether a small gift changes behaviour. The gift is a pocket planner — $0.92 as goods, and $1.20 once setup, freight and tax are counted. Nobody is put under meaningful pressure by a pocket planner, and the effect is measured in aggregate without identifying anyone.

Nothing is captured that a person could not reasonably expect from opening a link: no camera, no microphone, no facial or voice analysis, no eye tracking, no keystroke logging, no tracking across other sites. That is a property of how the platform is built, not a policy that could be changed for one study.

The identifying information that does exist is bounded by schedule rather than by discretion. The link between an envelope and what was done with it is severed automatically about 60 days after mailing. I hold the mailing list itself for longer, because it is the only way to send the debrief and the only way to honour a removal request, and it is deleted once the last exclusion window closes. Every window is published on the disclosure page. There is one more, and it is the only place your identity touches the experiment. If you ask for the report by email, that address is stored alongside a code identifying the envelope you came from, and that code carries which arm you were in. Nothing else about you is attached to it, it appears in no published file, and it is not in the research dataset. The count is taken once when the study settles and the code is then deleted from the list, leaving an ordinary mailing list.

04 · Justice

The people who bear the study are the people the findings are for.

The 1,950 organizations approached are drawn at random, under a documented seed, from a fully enumerated frame built from two public federal registers, in equal quotas of 650 from each of three sub-sectors — university advancement, cultural institutions, and national charities. Nobody was selected for being vulnerable, cheap to reach, or unlikely to object. The frame is published, so the selection can be checked rather than taken on trust.

One kind of organization is deliberately not in the sample. Hospital-system philanthropy belongs to the same class of work this study is about and stays in the population the question is asked of — but it is not enumerated into the frame, because registry-of-record addresses for hospital-system foundations frequently resolve to a parent-system administrative office, so a letter addressed to the advancement office cannot be shown to reach one. That is a property of the public register, not of the sector, and it is a limit on what the study can claim rather than a judgment about who deserved to be asked.

The population is advancement and development offices — the people whose working lives involve sending exactly this kind of mail and wondering whether it worked. If the study answers its question, the answer is directly useful to the sector that carried it. The burden and the benefit fall on the same group.

The findings are published either way. The study is designed so that a null result and a negative result are as reportable as a positive one, the verdict wording is fixed in advance so it cannot be composed after the number is known, and the de-identified dataset is published with the paper. I have a commercial interest in the gift working. The pre-registration is what stops that interest from choosing the answer.

There is no compensation. Nothing is asked of a recipient beyond what arrives in the envelope, and no one is worse off for ignoring it.

05 · Where this is weakest

Three things I would press on if I were reading this.

The research and the sales approach are the same envelope. The ICC/ESOMAR Code requires a researcher to maintain a clear distinction between research and non-research activity (Article 1(d)), and requires a data subject to be told about any non-research activity — promotional, commercial, follow-up — before data collection begins, with separate consent obtained for it (Article 1(e)). This study deliberately fuses the two: the mailing is a genuine commercial approach and a research instrument at once.

The strongest thing I can say for it is about order, not fairness. Article 1(e) is written for an approach that runs one way — research first, then a commercial use of the people it collected from. This one runs the other way. The commercial approach came first and is the reason the envelope exists; it is on the card, at first contact, before anything was collected from anyone. And there is no per-recipient finding at the far end that could be turned back on you: the frame was drawn from public federal registers rather than from anything you gave me, the mail is addressed to an office rather than to a person, and what the study produces is aggregate. So the sequence the Article guards against is not something I chose to forgo. It is not available to this study.

What that argument does not do. It does not make the fusion go away. A reader who thinks one envelope should never be both things at once is not answered by the order they arrived in. Both purposes are at least stated plainly rather than one hiding behind the other — the commercial intent on the card, the research on the disclosure page, in the consent notice, and in the debrief. A reasonable person could weigh it differently.

Nobody independent checked the ethics. The design commits to an independent methodologist — paid a fixed fee that does not depend on what the study finds — to review the analysis plan before it locks and to re-run the analysis after the window closes. That is a check on the statistics. It is not a check on this document. The ethics of the study were assessed by the person with the most to gain from them being fine.

Consent, in this design, is thinner than the word usually implies. Deferred consent is standard, defensible, and the only structure that lets the measurement exist — but a recipient who never scans the code never sees the consent notice, and learns about the study for the first time when the debrief letter arrives two months later. For those recipients the disclosure is entirely retrospective. That is the cost of the design, and it is borne by them rather than by me.

06 · If you disagree

Two routes, and one of them is not me.

Write to me at lauren@thesweetnote.com — about this document, about the study, or about how a request of yours was handled. To stop everything, email the same address with the subject line "Gift Lift Study opt-out."

You are not obliged to come to me. This study is conducted under the ICC/ESOMAR International Code, and you can take a concern to ESOMAR — the body behind that Code — whose contact details are published at esomar.org. That route does not run through me.

Addendum, 23 September 2026

A short-lived record for honouring removal requests.

This addendum is published after mail day, as the freeze rule above requires: it adds to the determination and changes nothing written above the signature.

What it adds. The debrief lets any organization that was mailed ask, within 30 days, to have its record taken out of the published data. Those requests name the organization. Some may arrive before the platform's scheduled purge (around 1 November 2026), which removes the platform's own link between an organization's name and its envelope, and the rest arrive after it, when the platform can no longer match a name to a row. The mailing list I hold has names and addresses but no codes. So before that purge I capture one more thing from the platform: a map from each organization's name to the random code its row carries in the published dataset. Every request, whenever it arrives, is matched through this map in the same way. It holds no outcome of any kind. I hold it with the mailing list, use it only to honour a removal request, and delete it with the list once the last exclusion window closes. (Wording of this paragraph updated 11 October 2026, when the postcard moved earlier and requests could first arrive before the purge; the record it describes is unchanged.)

What it qualifies. Section 03 says that the report-signup list "is the only place your identity touches the experiment". That was not complete. While it exists, this map is a second place: it links a named organization to its published row, and that row carries the group it was in. There is also a third, which was already disclosed on /about-this-outreach: an email sent to me from the report page carries the same mailing reference in its body unless the sender deleted it, so the email address reaches me alongside a code that carries the group. That reference is deleted at settlement, with the signup list's.