Methodology disclosure
About this outreach.
The package you received is the first application of the methodology we're inviting you to apply to your own brand. This page documents what's actually happening behind the scenes — the study design, your role, our commitments, and where the data goes. It is kept to its own page, and linked rather than placed in front of you, for one reason worth stating plainly: reading it tells you that you are in a study, and that knowledge can change what you do next. Scanning the code is the one measurement that cannot be affected — it already happened. Everything after it can be, and the pre-registration records it as such rather than claiming otherwise.
If you came here to leave the study: the opt-out is one email, no questions, and it is honoured unconditionally.
01 · What this is
The Gift Lift Study — a pre-registered gift-presence field experiment.
We're running SweetNote's inaugural methodology study on our own outreach, with SweetNote as the brand. The study tests whether including a physical gift in a cold-ship envelope causally lifts recipient engagement with the digital reveal page — relative to an otherwise-identical envelope with no gift.
The methodology is Behavioral Brand Moment Research™ (BBMR™). The Gift Lift Study is the first published application of this template. Future Gift Lift studies under the template will be run by paying brands on their own physical mailings — not by SweetNote on its own outreach.
It is both things at once, and neither is a cover for the other. The envelope was a genuine commercial approach: I want SweetNote to be useful to your organization, and if you reply I will treat that as a real conversation. It is also, at the same time, a research instrument — your engagement is measured and contributes anonymously to a study I intend to publish. The ICC/ESOMAR Code requires research to be kept clearly distinct from non-research activity, and requires you to be told when a single approach is doing both. This one is doing both: the commercial part is the invitation to talk, the research part is the measurement described on this page. The two are handled separately — ignoring the invitation does not take your envelope out of the study (the opt-out below is the route for that), and asking to opt out is honored on its own terms, with no effect on anything else between you and SweetNote.
The Code provides for withholding the research purpose, and sets the standard that withholding has to meet. It requires that participation be voluntary and based on clear, accurate information about the purpose and nature of the research, given at the beginning — and then adds that where that is not possible because of the methodology, people must be told “at the earliest opportunity” (Article 4(a)(iii)). Telling you in advance that your scan was the outcome being measured would have destroyed the measurement, so the earliest opportunity is the day the envelope arrives — which is this page, reachable from the reveal, published before the first envelope shipped and not waiting on anything. The standard is not that you were eventually told. It is that you were told as early as the method allowed, and you are entitled to judge whether that is what happened.
02 · Design at a glance
The protocol, in one table.
| Brand | SweetNote (self-administered) — conflict-of-interest disclosed in the pre-registration |
| Population | 1,950 US donor-stewardship organizations — one population, allocated by equal quota across three sub-sectors: university advancement (650), cultural institutions (650), national charities and fundraising (650). Mail is role-addressed to the advancement or development office, not to a named individual — the frame carries no personal names at all. A seeded random draw from a fully enumerated, criteria-defined frame of US donor-stewardship organizations (built from two public federal registers), documented in the pre-registration. Results generalize to that frame, not to US institutions generally |
| Allocation | 50/50, randomized server-side by the SweetNote platform at mint, blind to recipient characteristics |
| Gift-present arm | Card + a QR code unique to that envelope + a SweetNote-branded pocket planner |
| Gift-absent arm | Card + a QR code unique to that envelope, and nothing else. The two envelopes are the same size and stock but are not weight-matched, so the gift arm's envelope is heavier and thicker — a recipient could notice a difference before opening. That limitation is disclosed in the pre-registration rather than engineered away |
| Held identical | Envelope, postage class, ship window, card copy, reveal page, video, all reveal copy, action CTA, landing page. The two arms differ in exactly one respect: whether the planner is in the envelope |
| Primary outcome | Reach — proportion of randomized recipients who scanned the QR (95% confidence intervals reported) |
| Decision rule | Registered in advance as a cost break-even test, not a test against zero. The gift costs more to send than the no-gift envelope, so the study asks whether it raises Reach by enough to cover that extra cost, and reports one of exactly three verdicts: below the threshold, above it, or an interval that straddles it. The threshold is a ratio computed from the settled print and postage costs and frozen before any envelope is scanned — never after. The registered prediction is that the effect falls below it: a gift at this price is not expected to pay for itself |
| Tier | Causal-standard threshold (n=975 per arm, 1,950 total — pre-specified and fixed; no optional stopping). Unlocks causal-verb verdict language in the working paper. |
| Pre-registration | Filed on the Open Science Framework before any envelope was randomized, and public immediately — there is no embargo. It is one registration, public from the moment it is submitted; the assignment commitment digest — the fingerprint of who was assigned to which group, fixed before a single envelope moved — is stated inside the registration itself, so it cannot be edited or backdated afterwards. The one record that cannot exist yet is the cost threshold, which is recomputed from the settled print and postage bills once the first envelope is accepted at the counter; it publishes afterwards as a dated update to the registration, and here on this page |
| Pre-registration URL | osf.io/yq9te — the registration is public from the moment it is submitted, so this link resolves for you without an account and without asking me |
| Research ethics | Conducted under the ICC/ESOMAR International Code. A written human-subjects determination was made and dated before the first envelope mailed. See independent review below for how to raise a concern with someone who is not me |
03 · Your role + your data
What we measure, what we don't, and what stays yours.
- What gets observed. Whether the QR was scanned, whether video playback was started (pressed play), whether a 3-tap reaction was tapped, whether the action CTA was clicked — each a single binary signal per recipient — plus when: the day the first scan happened (reported in coarse day brackets, never exact times) and whether a scan happened again on a later day inside the 30-day window (the rescan measure, which requires per-day scan records). Deeper watch-depth measurement is captured only for recipients who opt in via the consent notice, and is reported as exploratory alongside the opt-in rate.
- What does NOT get observed. No biometric capture (camera, microphone, expression analysis, voice tone, eye tracking — none). No keystroke logging. No recipient identifier attached to engagement signals in any artifact SweetNote produces. No tracking of your behavior on the broader web.
- The two places your identity touches the experiment. First, if you ask for the report by email, that address is stored next to a code identifying the envelope you came from — and that code carries which arm you were in. The code is there only if you arrived by scanning your card; arrive any other way, or send Global Privacy Control, and it is simply absent — the request still works. Nothing else about you is attached to it, it appears in no published file, and it is not in the research dataset. When the study settles the count is taken once and the code is deleted from the list, which leaves an ordinary mailing list. Second, if you email me from that page — to ask a question, to take up the offer, or to opt out — the reply is prepared with the same mailing reference in its body, so your email address reaches me alongside the code that carries your arm. The line is visible in the draft before you send, and deleting it costs you nothing: the message still reaches me, and the opt-out still works. Same code, same deletion at settlement, and it is in no published file either.
- What stays yours. Your individual response is never published, shared, or attached to your name in any output. The architecture enforces aggregate-only output at the query layer — it is not a policy promise we made on a slide.
- How the paper reports it. A rate per arm, each with a 95% confidence interval, and the difference between them with its own interval — the form is "the gift arm scanned at x%, the no-gift arm at y%, a difference of d percentage points, 95% CI a to b." We are not showing example numbers here, because no result exists yet and an invented one would read like a prediction. No recipient-level data appears anywhere in the published paper.
How long it's kept — and what for
Your address: about 60 days. The mailing address on your envelope is the only thing that could tie an engagement record back to you. It is held to send the envelope, to send you the debrief postcard afterwards, and to honor an exclusion request — and the platform deletes it 30 days after the 30-day engagement window closes, which is roughly 60 days after your envelope shipped. That deletion is what severs the link; once it runs, the records below cannot be reconnected to you. I hold the mailing list itself outside the platform for the same three reasons, and delete it once the exclusion window has closed for the last piece in the debrief wave.
Engagement records: 24 months, under a random code. Whether the code was scanned, play was pressed, a reaction was tapped, the button was clicked — each recorded against the random code printed on your card, never a name. Nothing is recorded after the 90th day from mailing. What was recorded is kept for 24 months so this study can be compared against the ones that follow it, and then expires.
The token map: 120 days, no exceptions. Because this study measures whether the button led to a next step, it keeps a map of the random code to the study cell it was assigned to (no names, no addresses). It is purged unconditionally 120 days after the mailing date, roughly ten weeks after your debrief postcard arrives.
Opt-out — and exactly what it does
Opt out and nothing further is sent to you — no debrief postcard, no follow-up of any kind — and your envelope is counted, once, as an opt-out. To do it, email lauren@thesweetnote.com with the subject line "Gift Lift Study opt-out." No questions, and no effect on anything else.
The one thing an opt-out cannot do is un-count what was already measured. The number of envelopes the study set out to send is fixed before mailing and is never re-based, and any engagement recorded before your request arrived stays inside that aggregate count. That fixed number is the only thing standing between a researcher and the option of dropping whichever rows spoil the result and publishing the remainder. The pre-registration closes that option, and it closes it to me.
The ICC/ESOMAR Code requires that you be allowed to withdraw at any time and that your personal data be deleted when you do — and attaches one condition: provided doing so “does not compromise the integrity of previously collected or analysed research results” (Article 4(c)). That is a condition I have to satisfy, not a place to hide, so here is why it is satisfied. The number this study divides by was fixed before a single envelope shipped. Re-basing it afterwards is what would compromise the result; refusing to re-base it is what keeps the result checkable by you. Everything the Code does require on withdrawal — deleting the personal data, taking your record out of what is published — is done, and done without asking you why.
An opt-out is also a request to be removed from the published data, and it is treated as one. You do not have to ask twice, and you do not have to wait. Your record comes out of the published dataset and out of the published event log.
If you do not opt out, the same right arrives again with the debrief postcard about six to seven weeks after your envelope shipped: it points to the debrief page, and the email route on that page takes your record out. That route is registered in the pre-registration, and the number of people who take it is published in the paper.
04 · Why we're doing this on ourselves
Honesty up front — we have a stake in the outcome.
SweetNote's inaugural methodology study runs on SweetNote's own outreach. The choice was deliberate: practicing the methodology before charging anyone for it is the most honest demonstration available. It also creates an obvious commercial interest in the gift effect being positive — a positive Gift Lift Study result strengthens future Gift Lift sales.
The structural safeguards against this conflict are public pre-registration with fixed n (no optional stopping — the sample is fixed when the QR codes are minted, the engagement window is a hard 30 days, and the platform computes verdicts only in the settled report, so no interim look can alter the stopping rule, the sample, or the analysis) and a two-sided test (the study must be able to honestly report no effect or a negative effect, not only a positive one). The primary outcome is pre-specified — Reach was locked before mailing, so we cannot post-hoc choose the metric that looks best — and so is the analysis procedure: the primary decision rule is a Miettinen–Nurminen score interval on the relative lift, read against the pre-registered cost break-even threshold, with a Newcombe difference confidence interval on Δ and Holm–Bonferroni adjustment across the four secondary outcomes in support. A de-identified open dataset is archived after publication. A paid independent methods audit is built into the design, and the engagement is signed: Katherine Davis, an independent statistician engaged under a fixed fee that does not depend on the result, to review the pre-registered analysis plan, independently re-run the final analysis, and sign a replication statement published with the working paper. That engagement was signed on 17 August 2026.
Null and negative results — "the gift didn't help" or "the gift reduced engagement" — will be reported in the working paper alongside any positive result. The integrity of the methodology depends on this being a real commitment, not a hedge.
05 · Known limitations
What the Gift Lift Study can and cannot tell you.
- Three of the 1,950 pieces never mailed — a real-world fulfillment problem, and how we chose to handle it. (Dated addendum, added 1 September 2026 — the day before the batch entered the mail on 2 September, so every recipient reads the same page.) On mail day the print house delivered three pieces defective (two without an envelope, one without a card), and they could not go out with the batch. All three had been assigned to the gift arm at mint. We did three things, in this order. They stay in the study: the pre-registration fixes the denominator at the count randomized at mint — 975 per arm — and forbids re-basing it for any post-randomization event, and a print defect is post-randomization by construction, so the primary analysis runs on 975 / 975 exactly as registered. They are recorded, not deleted: the platform carries a fulfillment-exception mark that is distinct from cancellation — cancelling would have dropped the rows from the denominator and quietly turned the registered intention-to-treat estimate into a per-protocol one, so it was not used. They are reported: the paper also gives an as-delivered reading on 1,947 pieces (gift arm 972) as a labelled sensitivity, never as a competing estimate — the same form the registration already specifies for envelopes returned undeliverable. A piece that never mailed can only score zero on Reach, so the exception can only lower the gift arm's rate, by at most 3 / 975 = 0.31 percentage points — against the study's own hypothesis. It is logged in the published deviation log and as a dated update to the registration. We say it here because a brand running the same study will meet the same problem, and the honest test of the instrument is whether it has a place to put it that keeps the randomized denominator intact.
- You know you are in a study — and knowing can change what you do. A recipient of an ordinary Gift Lift study never learns they were measured. This one tells you outright, in more than one place: the privacy notice on the reveal page says the outreach is also a research study, the video's closing invitation describes the platform as research infrastructure, and this page sets out the protocol in full. Only the scan itself sits outside all of that — it necessarily happened before any of them could be read or heard. Everything after the scan — pressing play, tapping a reaction, clicking the button, coming back later — is treated in the analysis as partially affected by knowing, and the paper says so wherever it reports one of them. What this does not touch is the comparison the study exists to make: every recipient meets the same notice, the same video and the same page, whichever envelope they were sent, so the awareness is identical in both arms. It bounds what the absolute numbers mean, not the gift effect.
- A probability sample of a defined frame — not of US institutions generally. The 1,950 recipients were drawn at random, under a documented seed, from a defined and fully enumerated sampling frame built from public federal registers. What we learn generalizes to that frame — US donor-stewardship organizations meeting the published criteria — not to institutions or B2B audiences broadly. Every address was validated against USPS records before the random draw — only validated addresses entered the frame — so the effect we measure generalizes to the deliverable portion of that frame. It does not generalize to the other kinds of physical moment the methodology will eventually serve.
- Branded-gift confound. The Gift Lift Study uses a SweetNote-branded pocket planner. We chose a branded gift because brand-customer applications of this methodology will overwhelmingly use brand-presence-carrying gifts; testing with an unbranded gift would produce results that don't transfer to commercial application. This creates a confound: the effect we estimate combines tangible gift value with brand-presence reinforcement, and the Gift Lift Study cannot separate the two mechanisms from one study. Future Gift Lift studies should test the methodology with both branded and unbranded gifts to allow meta-analytic decomposition.
- Causal-standard tier — meaningful but bounded intervals. At n=975/arm we sit in the causal-standard band — comfortably above the ≥600/arm floor where causal-verb verdict language unlocks. The 95% confidence intervals are roughly ±4.4 percentage points at a 50/50 baseline — the worst case; they are narrower at realistic (lower) baselines. The verdict is read from the interval's position relative to the cost threshold in section 02, and it takes one of exactly three forms: the interval lies wholly below the threshold, wholly above it, or it spans it. A spanning result means the study could not settle which side of the threshold the effect falls on — it is not a finding that the gift did nothing, and the pre-registration forbids reporting it as one. The sample was not sized against an assumed effect: it was sized so the study can resolve that threshold across most of the plausible range of scan rates, which is a different and harder thing to ask of it. The peer-reviewed journal paper will use pooled multi-brand causal-robust data at the 24–36 month horizon.
- Inaugural single-brand study. The Gift Lift Study is one brand (SweetNote) on one specific outreach campaign with one specific branded gift. The pooled-brand analysis for the journal paper will introduce between-brand heterogeneity that the Gift Lift Study by definition cannot.
06 · Where the documents live
What's published from day one, and what follows.
The written determination. The human-subjects determination described in section 07 below — made and dated before the first envelope mailed — is published at thesweetnote.com/human-subjects-determination. It is not embargoed and does not wait on anything: it is readable from the day your envelope arrives.
Everything else. The pre-registration is public the moment it is filed, and its link is in the table in section 02 — it is not embargoed and you do not need an account to read it. The working paper draft, the methodology paper outline, and the experiment-readiness audit follow after the engagement window closes, and are linked here when they publish.
If you'd like to receive a copy of the working paper the moment it posts to SSRN, email lauren@thesweetnote.com with the subject "Working paper notify."
07 · Independent review
Someone other than me has to be able to hear about it.
The standard this is held to. This study is conducted in accordance with the ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics — the global self-regulatory code for research of this kind. Before the first envelope was mailed, I made and dated a written determination about the study's human-subjects status and the ethical basis for measuring it the way I have; it is published alongside the study materials, and you are entitled to read it and to disagree with it.
Raising a concern with someone who is not me. If you have a concern about how this research was conducted — including a concern about this page, or about how a request of yours was handled — you can take it to ESOMAR, the body behind the Code this study follows, without going through me. Their contact details are published at esomar.org. You can also write to me directly at lauren@thesweetnote.com, but you are not obliged to.
The methods auditor. The study's design carries a paid independent methods audit, and the engagement is signed. Katherine Davis, an independent statistician engaged under a fixed fee that does not depend on what the study finds, is engaged to review the pre-registered analysis plan, to independently re-run the final analysis on the de-identified dataset, and to sign a replication statement published alongside the working paper. They are named in the paper as the auditor and are not an author. The engagement was signed on 17 August 2026. If the replication is not delivered, the paper reports the analysis as unreplicated rather than quietly dropping the commitment.
The registration. The pre-registration is filed on the Open Science Framework before any envelope is randomized, and it is public from the moment it is filed — the link is in the table in section 02, and it resolves without an account. It fixes the primary outcome, the sample size, the analysis procedure and the reading rules in advance — which is what makes it possible for a reader to check, afterwards, that the answer was not chosen to suit us.