You bring the question. We bring the rigor.
Research-grade studies for the physical sends you already make, each answering a different question. Start with yours.
Patent pending.
What recipients see when they scan.
The page after the scan, where every signal in your dashboard is generated.
Brand-owned, behavioral, one clear next step. Every interaction on this page is a signal in your dashboard.
- Video
- Reach + Engagement
- Three-tap reaction
- Affirmation
- Brand CTA
- Action · + optional Outcome
Why every element is there
Universal layout
- 01
Greeting at top, recipient name personalized.
The recipient’s name appears in the first ~200ms, above the primary content. Sahni, Wheeler & Chintagunta (Marketing Science, 2018), minimal name personalization raises open rates in randomized field experiments.
- 02
Tap-to-play video. No autoplay.
Video starts on tap. Preserves user agency, matches mobile-native patterns, satisfies WCAG 2.2. Apple HIG / Google Material consent guidelines.
- 03
Vertical 9:16 video aspect ratio.
Primary video is 9:16. Maximizes mobile screen real estate; matches native portrait viewing. 9:16 dominates mobile video consumption time across TikTok, Reels, Shorts.
- 04
Three-tap warmth ladder.
Sweet · So sweet · The sweetest. Three taps rising in warmth, never five-point or open-ended, the cognitive sweet spot for fast emotional response. Reichheld (Bain, NPS) on radically simplifying response scales.
- 05
44×44 px minimum tap targets.
Every interactive element meets Apple’s 44pt touch-target minimum. Apple HIG (44pt); Google Material recommends 48dp; satisfies Fitts’s Law.
- 06
No consent wall. The content is never gated.
The reveal never blocks the video or message behind a dialog. It plays for everyone. Opening the reveal, pressing play, the reaction, and the CTA click are all counted anonymously (no name) for every recipient, on a legitimate-interest basis; only how far someone watches the video (the continuous watch-depth measurement) stays off until the recipient taps Count me in on a dismissible ribbon, and Global Privacy Control is honored automatically as an absolute opt-out. No forms, no surveys, no dark patterns. GDPR Art. 7 + Art. 4(11) (unambiguous, freely given consent); CCPA §1798.140(h).
- 07
Body text 16px minimum, line-height 1.5+.
Type at the mobile-readable floor. No pinch-zoom, no eye strain. Line-height 1.5 supports WCAG 2.2 SC 1.4.12 (Text Spacing); 16px is mobile-readability best practice, not a fixed WCAG minimum.
- 08
4.5:1 minimum on primary text. WCAG AA.
The auto-contrast engine holds primary text at 4.5:1 or better on any brand color (WCAG AA); the default light surface typically exceeds AAA. Readable for older recipients without adjustment.
- 09
CTA placement: post-content, pre-reactions.
Primary CTA appears at content completion, above the reactions. Captures intent at the moment of full engagement.
- 10
Brand mark in header. Platform attribution in footer.
Sender’s brand at top (~10% of page area). “Powered by SweetNote” small at the bottom. Brand presence at the moment that matters.
- 11
Single scrollable page. No pagination.
The digital reveal is one scrollable surface; no “next” button between sections. Preserves narrative arc.
- 12
Video runtime 15–90 seconds recommended.
Soft 15–90s methodology range. Hard ceiling 90s. Below 15s saturates watch% near 100% and reduces discriminating power.
Brand-authored fields
- 01
Inscription paragraph, the brand’s short message.
Freeform brand-written prose, ≤ 500 chars. Universal use: thank-you context, invitation event details, announcement narrative, acknowledgment framing. The 500-char cap bounds reading-time variance across tenants so scroll-to-CTA behavior stays comparable.
- 02
Action CTA, one required next-step button.
Single required URL + label, capped at 40 chars. One slot, one shape, every reveal, the click is captured as the cross-tenant-comparable Action signal. The CTA is also the handoff point for optional outcome attribution: brands that relay their processor's webhook events receive the downstream conversion (donation, purchase, registration) attributed back to the reveal cohort.
- 03
About-your-brand paragraph, an optional closing word.
A short brand paragraph, ≤ 300 chars, near the foot of the reveal under a small “About” label. The words are entirely the brand’s, SweetNote writes no headline over them. Set once as a workspace default and optionally overridden per campaign. Brand-constant identity copy, the same for every recipient, so it adds no per-recipient variance and results stay comparable. Omit it and the section simply doesn’t render.
Things missing, by design.
- 01
No open-ended text input from recipients.
The digital reveal does not accept free-text input. Reactions are the 3-tap structured choice; no comment box. Mixed-instrument research warns that free-text + reaction in the same surface dilutes both signals.
- 02
No RSVP form management.
For invitations, brands write event details into the inscription paragraph and put their existing RSVP URL in the action CTA. The brand’s system (TheKnot / Cvent / Eventbrite / Splash / Bizzabo / Classy / Givebutter) stays the source of truth.
- 03
No biometric capture, anywhere.
No camera access, no microphone access, no facial-expression inference, voice-tone analysis, or eye tracking. No BIPA-sensitive biometric data is captured (BIPA, 740 ILCS 14; biometric information per CCPA §1798.140(c)). Sentiment is the 3-tap reaction only.
- 04
No video autoplay.
Video starts on tap, never autoplays. Preserves user agency; matches mobile native patterns.
- 05
No reaction options beyond three.
Three taps, fixed labels. Not a 5-point scale, not a Likert, not an emoji palette. Comparable across studies and across brands.
- 06
No per-recipient drill-down for senders in SweetNote’s output.
Senders never see per-recipient engagement detail in SweetNote’s own dashboard, deliverable reports, or the cross-tenant benchmark, that’s the structural separation that makes the platform research-firm-credible. Integrations a brand opts into are a separate layer: per-recipient activity is documented to flow to the brand’s own webhook destination keyed by an anonymous code the brand joins to their own records on their side, the brand using their own data through us, same way Qualtrics or Sprig handle response forwarding.
Every design runs on the same architecture.
Same instrument every time; what differs is how recipients are grouped and whether assignment is random. Pick the question; the study follows. ClickTap any design for the full method.
How is the moment landing?
Snapshot- Q4 client appreciation gifts
- Customer-anniversary moments
- Account-milestone thank-yous
- Annual employee recognition
- Year-end donor stewardship
One audience. One creative. One moment. The starting point for every brand, establishes a baseline for how your gifting program lands today, before you start optimizing it.
Underlying research → · ARF measurement standards, ESOMAR transparency code
Do segments respond differently?
Cohort comparison- Enterprise vs. mid-market vs. new client
- Engineering vs. sales recognition
- New vs. renewing customers
- Region-by-region campaigns
- Major-donor vs. sustaining-donor stewardship
Same creative. Different audience segments. Shows you which parts of your audience your program is reaching, and which it's not. Cuts your engagement signal across the segments you already think in: tier, region, tenure, department.
Underlying research → · Sweeney (2002) on k-anonymity cell-suppression thresholds
Which reveal performs better?
A/B · parameterized 2–5 variations · Bonferroni correction (a statistical adjustment for running several comparisons at once) at k≥3- CEO-voice vs. team-voice testing
- 30s vs. 60s video length
- Personal note vs. company note
- 3+ creative concept comparison (k≥3 variations)
- CTA wording: “See the impact” vs “Read our note”
- Reveal design: warm vs clean surface
Two to five creative variations. Randomly assigned, so the difference you measure is causal, not a guess. The cleanest way to learn which creative actually drives more action. Pick 2 variations for a head-to-head; 3–5 variations for a ranked leaderboard across several concepts. Message testing, voice, length, framing, opening line, CTA wording, is the most common use.
cohort_ column to balance within — assignment then balances independently inside each group, so every segment splits evenly too, not just the audience as a whole. Either way the assignment is re-derivable from the stored recipient tokens, so an independent reviewer can recompute it and confirm nothing moved after results started arriving. Sample floor n ≥ 100 per variation (n ≥ 200 at 2 variations, n ≥ 500 at 5 variations). A/B's endpoints are read on the people who scanned, not on everyone you mailed — the variation isn't seen until after the scan, so scanning can't select on treatment. That means your response rate sets the resolution: at an assumed 25% scan rate, 100 recipients per variation is 25 scanners per variation, and the detectable difference at k=2 is ≈40pp. The read sharpens as you scale: ≈28pp at 200 per variation, ≈20pp at 400, ≈14pp at 800, ≈10pp at 1,600. Scan higher and the same resolution costs fewer recipients; the study builder recomputes it live against the rate you enter. Wilson score-difference confidence intervals at every k; Bonferroni correction applies automatically at k≥3 so the chance of a false winner across all the pairwise comparisons stays controlled, and the uncorrected per-pair MDE is disclosed alongside.
Underlying research → · Webb & Sheeran on the intent-action gap, randomized-experiment design
Which of 2 to 5 gifts (or pieces) works best?
Gift A/B · parameterized 2–5 SKU randomized comparison- Premium gift vs branded swag, which drives more retention?
- Three finalist gifts, ranked across one audience
- Branded notebook vs branded mug, which gets the action?
- Two cost-tier options, which has the better return?
- Negotiating with procurement on a gift line-item between two finalists
Same reveal page, same QR card, comparable carrier. Two to five physical SKUs ship to evenly randomized groups of the audience, each in the packaging its own SKU requires. Allocated by SweetNote so the comparison is causal, the lift between SKUs is the causal effect of the SKU choice, not anything else.
cohort_ column to balance within and assignment balances independently inside each group, so every segment gets an even spread of SKUs too. Two attestations gate launch (execution-identical: same QR card, equally prominent across SKUs, comparable carrier, ≤48h send window — only the SKU and the packaging it requires differ; plus randomization-honor), so the only difference recipients experience is the SKU. Sample floor n ≥ 100 per SKU (n ≥ 200 total at 2 SKUs, up to n ≥ 500 at 5 SKUs); detectable difference ≈20pp at the floor, ≈14pp at 200 per SKU, ≈10pp at 400. Unlike creative A/B, this reads on everyone you sent (intent-to-treat), not on the scanners — the gift changes who scans, so conditioning a gift comparison on scanners would compare arms whose scanner populations are no longer alike. The practical effect is that a Gift A/B reaches a given resolution on roughly a quarter of the recipients a creative A/B needs. Same reveal across every SKU; brand handles SKU routing at fulfillment using the per-recipient variation_assignment column SweetNote writes to the CSV. Same Wilson + score-based difference-CI math as A/B; labels in the dashboard and deck are the brand's SKU names.
Underlying research → · Peck & Childers (2003) on haptic product judgments; randomized-experiment design for the causal verb.
Which piece gets opened?
Gift Lift · one physical difference, randomized 50/50- Does the gift in our donor thank-you actually drive the next gift?
- Is the swag in our welcome kit worth the line-item?
- Does the sample in our DTC moment drive conversion?
- Is the gift-with-purchase doing what the SKU costs?
- Justifying a physical-gifting line-item to a CFO
SweetNote's only study design that isolates the gift itself from the rest of the brand moment. Same QR card, shown the same way in both groups, same reveal, same sender, same send window. Half the recipients also receive the gift, randomized 50/50 by SweetNote. The only thing that changes is the gift, including the packaging it comes in (a box for the gift group, an envelope for the card-only group; the box is part of what a gift is). The lift between the card-only and card+gift groups is the causal effect of sending a gift, not the channel, the card, or the brand voice.
cohort_ column to balance within and assignment balances independently inside each group, so each segment is split 50/50 too. Hard floor n ≥ 100 per group (n ≥ 200 total); 4-tier capability disclosure: causal-robust at n ≥ 1,000/group (detects a ≈6pp lift at 80% power), causal-standard at n ≥ 600/group (≈8pp), descriptive-only at n ≥ 200/group, directional-pilot at n ≥ 100/group. Two attestations required at scoping: execution-identical (identical QR card, equally prominent in both groups; same reveal; comparable carrier and ≤48h send window — the only intended difference is the gift and the packaging it requires) and randomization-honor. Post-send screenshot audit. Same digital reveal for both groups keeps the comparison clean.
Underlying research → · Krishna (2012) on sensory-marketing effects; randomized 50/50 field-experiment (intent-to-treat) design.
At a glance.
Any design, one line each. Open its tab above for the full method.
| Design | The question it answers | Evidence | Minimum recipients | Premium |
|---|---|---|---|---|
| Snapshot | "How is the moment landing?" | Descriptive read | 50+ | $0 |
| Cohort | "Do segments respond differently?" | Descriptive read | 50+ per segment | $1,500 |
| A/B | "Which reveal performs better?" | Causal test | 100 per variation | $3,000 + $1,000/variation above 2 |
| Gift A/B | "Which of 2 to 5 gifts (or pieces) works best?" | Causal test | 100 per gift | $3,000 + $1,000/SKU above 2 |
| Gift Lift | "Which piece gets opened?" | Causal test | 100 per group | $4,000 |
Every design adds the $500 base fee plus the published graduated per-recipient rates, runs the same 30-day engagement window, and delivers the same live dashboard, settled report deck, and data tables. Outcome Attribution is included with any design, opt-in at scoping. Price yours →
The signals every study captures.
The same headline signals on every study, in funnel order: the tiles your dashboard leads with. Each opens into richer depth metrics (the watch curve, time-to-scan, reaction mix), with opt-in Outcome attribution on top. Surveys ask what people remembered; we measure what they did.
- 01ReachDid they open it?
- 02EngagementDid they press play?
- 03AffirmationDid they react?
- 04ActionDid it drive a next step?
- 05SpeedHow fast did they open it?
- 06RescannedDid they come back?
Wire a payment processor to add an opt-in 7th signal, Outcome (did they convert?), reported as aggregate causal dollar lift on randomized studies: the incremental dollars the send caused, measured against a randomized control, not a last-click number that also credits the conversions you’d have won anyway.
Same signal. Three ways to read it.
Illustrative example · Action signalThe design decides which question it answers.
Snapshot
“How is the moment landing?”
Your rate reads on its own as a baseline. Benchmark read: the live cross-tenant comparison appears once your comparator — the network excluding you — holds 10 other brands that have run in your category; no manufactured comparator is shown before then. The network is filling now, so a first study should expect to read as its own baseline. On every design, benchmark rates are published rounded to the nearest whole percentage point, never with the underlying counts, and the comparator’s size is published only as a band: 10–19, 20–49, 50–99 or 100+ other brands, and 25–99, 100–499, 500–1,999, 2,000–9,999 or 10,000+ recipients.
Cohort
“Do segments respond differently?”
One rate, cut by a cohort_ column: each cell its own 95% CI. Below n = 25 the counts stay visible and the rate and its interval are withheld. Descriptive read: segments weren’t randomized, so we report what we saw, not cause.
A/B · Gift A/B · Gift Lift
“Which reveal performs better? Is the gift worth it?”
Compared across randomized variations. Causal read: random assignment earns the cause-and-effect verb; a win counts only when the interval clears zero. (Shown here on Action so the panel compares the same signal in all three reads — Gift Lift's own headline runs on Reach, its primary endpoint, with Action as the supporting causal read.)