Brand database · briefs/wam/BRAND.md · source of truth

WAM Mints — Brand Database (BRAND.md)

WAM Mints — Brand Database (BRAND.md)

The single source of truth for every durable fact about this brand. Any note David sends

(inventory, suppliers, accounts, contacts, decisions) is filed HERE the same turn it arrives —

before any context compaction. Memory files only POINT here. Never store secrets (tokens, keys,

passwords): reference the env file that holds them. Keep entries dated. Hub renders this file.

0 · Identity & entities

FieldValue
Legal entity / ownerTO FILL
Key people & rolesTO FILL
Support email / phoneTO FILL
Brand relationshipTO FILL (own brand / client / partner)

1 · Products & SKUs

Live Shopify product Honey Lemon Ginger Mints (honey-lemon-ginger-mints, Product 8144505634953), option "Size".

Read from the store 2026-10-05. 48 mints a pocket tin, 150 a resealable pouch. Every variant 0.8 oz, taxable,

inventory tracked + DENY (stops selling at zero).

SKUVariantVariant IDPriceMints
WAM-MT-HLG-3TIN3 Tins44347541684361$25.99 (from 2026-10-24; $24.99 from 10-19, $23.99 before)144
WAM-MT-HLG-5TIN5 Tins44347541717129$37.99 (from 2026-10-24; $36.99 from 10-19, $35.99 before)240
WAM-MP-HLG-PCHPouch45184616300681$24.99 (from 2026-10-24; $23.99 from 10-19, $22.99 before)150
WAM-MR-HLG-RFLL1 Pouch + 1 Tin45184617021577$29.99198
WAM-MS-HLG-SUBS2 Pouches + 1 Tin45184616824969$52.00348

Package-test product (hidden, 2026-10-05): wam-mints-packages, Product 8888721735817, same title + same SKU

codes at the test prices — 3 Tins 48517130518665 · 5 Tins 48517130846345 · Pouch 48517130879113; since 2026-10-19

$25.99 / $37.99 / $24.99 (price test round 2; was $27.99 / $39.99 / $22.99, then $24.99 / $36.99 / $23.99); inventory not tracked. Record: products/wammints/package-test/README.md. Other products in the store:

limited-edition-kpops-honey-lemon-ginger-mints, honey-lemon-ginger-mints-copy-individual,

watermelon-cucumber-mints (draft).

2 · Inventory & supply chain

SKUUnits on hand (Shopify)Location / 3PLLot / mfg dateShelf lifeCOGS (landed)MOQ / reorder lead timeUpdated
WAM-MT-HLG-TIN (single tin, the one tin item ordered)1,264?56 days PO → warehouse (David); price break at 60,000 tins a PO (David)2026-10-13
WAM-MT-HLG-3TIN (3 Tins)421 (= 1,264 ÷ 3: packed from single tins)not ordered separately2026-10-13
WAM-MT-HLG-5TIN (5 Tins)252 (= 1,264 ÷ 5: packed from single tins)not ordered separately2026-10-13
WAM-MR-HLG-RFLL / WAM-MS-HLG-SUBS (bundles)1,264 each (1 single tin + 1 or 2 pouches)not ordered separately2026-10-13
WAM-MP-HLG-PCH (Pouch)4,746?56 days · its own item (David)2026-10-13
WAM-MP-HLG-KPOP (K-POPS pouch, limited)534?not reordered (David)2026-10-13

Notes: lead time 8 weeks, PO to delivered to warehouse (David, 2026-10-09).

Confirmed by David, 2026-10-13:

Shopify's 3 Tins count was a separate 545 on 2026-10-12. On 2026-10-13 it reads 421 = 1,264 ÷ 3, now derived from

single tins like 5 Tins and the bundles.

Shopify stock is tracked + DENY; this

login can't read locations or incoming stock. The counts imply the warehouse (or its stock-sync app) builds 5 Tins and

both bundles from single tins (267 = 1,338 ÷ 5; bundles = single tins), while 3 Tins carries its own count — to

confirm. Read 2026-10-28 21:05 PT: single tins 536 · 3 Tins 245 · 5 Tins 107 (= 536 ÷ 5) · both bundles 536 · Pouch 4,694.

3 Tins is not 536 ÷ 3 (178), so it carries its own count again. The PDP box opens on 1 Pouch from 21:09 PT (§6). Channels drawing on it: website, Skio renewals, TikTok shipped by seller, drafts; TikTok shipped by TikTok

(FBT) and Amazon (~25 orders/day) ship from their own warehouses as far as known. Inventory forecast

(tools/inventory-forecast/, Launch Center /wam/inventory/, 2026-10-12): with no open POs on file, 3-tin packs

(12.8/day) run out ~Nov 24 and single tins (22.8/day) ~Dec 10, while a PO placed today lands Dec 7; both are past

their reorder date (lead + 14-day buffer). Daily cap until new stock ≈ 7.4 website orders/day (today 17.6), limited by

3-tin packs. Fulfillment partner, shipping SLA, packaging on hand: TO FILL.

Read 2026-11-06 17:43 PT: 3 Tins 64 · 5 Tins 17 · single tins 82 · both bundles 82 · Pouch 4,628 · K-POPS 523. Units a day over the 14

days to Nov 6: 3 Tins web 6.1, TikTok 3.0, Skio renewals 1.7 · 5 Tins web 4.1, TikTok 0.6, renewals 0.2 · single tin TikTok 1.7. Once the site

stops selling tins (§6), 64 packs last about two weeks with TikTok on, about five with it off.

Amazon, per listing — units ordered, Seller Central → Business Reports → Detail Page Sales and Traffic by Child Item, read 2026-11-15.

First sale Aug 17, 2026; nothing sold before. Both tin listings held the Buy Box 96–99.8% every month, so they never ran out; the fall since

September is traffic (3 Tins sessions: 1,943 in Sep, 1,140 in Oct, 374 in Nov 1–15). Amazon is stocked by sending units in (David, 2026-11-15).

Listing (ASIN)Aug 17–31SepOctNov 1–15Best 30 days
3 Tins (B0GQ5K1P2Y)274482325111485 (Aug 31 → Sep 29)
5 Tins (B0GQ5G7G9X)171818412190 (Sep 7 → Oct 6)
Pouch (B0G66Z16J9)171112233101240 (Aug 17 → Sep 15)
1 Tin + 1 Pouch (B0GQ579FCM)1026201227
1 Tin + 2 Pouches (B0GQ54R8X1)4281015

One month at peak, to send to Amazon (David's question, 2026-11-15): about 500 3 Tins and 200 5 Tins — the best 30-day windows (485 and

190, windows a week apart) rounded up. Packed from single tins, that's 2,500 tins (1,500 + 1,000), plus about 45 if the two tin-and-pouch

bundles go too. The last 30 days ran at about half the peak: 261 and 57 (Oct 12 → Nov 10). Whole-account sales peaked at $21.8K for 30 days (Aug 17 → Sep 15).

The FB spend behind the peak (MER data, Meta account act_909212504920044, read 2026-11-15): September $19,537, about $651 a day, steady all month

($556–$783; late August ~$661). October fell to ~$442 a day and Nov 1–14 ran ~$456, and Amazon's 3 Tins fell by a third over the same step (482 → 325).

Amazon tins per $1,000 of FB spend, same 30-day windows: 3 Tins 23–28 in every window since launch (20.6 in the latest, Oct 12 → Nov 10), so they track

FB almost one-for-one; 5 Tins 4.5–9.9 (peak mid-Sep → early Oct, falling since). Amazon's own ads moved with FB ($74–85 a day in Sep, ~$51 now).

At $2K a day of FB (David's question, 2026-11-15), 3.07× September: planning number about 1,200 3 Tins and 470 5 Tins a month (cost per sale up

~25% from tripling the spend, Meta's usual price for scale) ≈ 6,000 single tins. 1,490 / 590 if every dollar worked like September; ~1,060 / 420 at 40% dearer.

3 · Offer & pricing (locked)

4 · Accounts & IDs (no secrets)

SystemID / handleNotes
Domain(s) / DNSTO FILL
Netlify site (id · team)TO FILL
29next store / campaignTO FILL
Meta BM · ad account · pixelTO FILLCAPI token → env file only
GA4 property · measurement id · GTMTO FILL
Klaviyo orgTO FILL
adscalerads client · Airtable baseTO FILL
Cookie / event prefixTO FILL
Amazon (US) · Seller Central "WAM! Mints"parent ASIN B0GR75TMWL; 3 Tins B0GQ5K1P2Y · 5 Tins B0GQ5G7G9X · Pouch B0G66Z16J9 · 1 Tin + 1 Pouch B0GQ579FCM · 1 Tin + 2 Pouches B0GQ54R8X1; Ads entity ENTITY12FFWYKD8ROIDFirst sale 2026-08-17. Saved session for automated reads: ~/.local/share/mer-amazon/state.json (never print it). Business Reports retires end of 2026 (Seller Central banner, seen 2026-11-15).

5 · Compliance ceiling

6 · Decisions & taste log (dated, David's verdicts = ground truth)

Not significant, but both buying signals point the same way, and the call was his.

How it shipped: box-test/ship_box.py --push makes layout/theme.liquid add wam-box-b server-side on the real

PDP only. That is the exact code path B ran. Mida was concluded ~2 minutes after the push, so no visitor saw the

old box in between.

Verified:

- box-test/verify_rollout.py ALL PASS: four browser/device configs, the copy product untouched, four carts,

WELCOME20, the sticky bar.

- Attribution matches the pre-ship arm-A baseline event for event (Meta eventIDs 2/1/2; GA4; TikTok; Klaviyo;

metrics.wammints).

- iPhone Simulator: 5 Tins Monthly $32.39 into the cart; one-time prices right; no jump on tap.

- Observer audit healthy (32 timers/sec, flat).

- PageSpeed unchanged within noise: median 42 → 38; the after-runs ranged 32–41; CLS 0.259 both; LCP ~20 s

before and after.

Found: Mida's anti-flicker still holds every page white with no test running: ~1 s extra on a 4G phone (PDP

painted 1.2 s, white until 2.3 s). He kept it for the next test (entry above).

7 · Asset map

8 · Live surfaces

9 · Open questions / asks

10 · Change log


Imported status record (2026-09-02) — distribute these facts into the sections above as the brand is touched

From memory wam_ads_pipeline.md

WAM Mints is a full OCC brand with a VIDEO-first ads pipeline (2026-07-13, 75596518).

Related: [[wam-klaviyo-status]] (flows/org), [[occ-master-dashboard]], [[jt-ad-pipeline-skill]].

From memory wam_campaigns_session_brief.md

WAM Mints — Campaigns + Content Calendar (Session Brief)

This is the canonical handoff. Read this first. The flows + popup are DONE in another session and are explicitly OUT OF SCOPE here (read-only context). This session owns: editorial calendar planning + one-shot Klaviyo campaigns (newsletters, drops, promos, holiday sends).


Scope

IN

OUT (do not touch — already shipped or owned elsewhere)


Production state (as of 2026-05-19) — read-only

Klaviyo account: WAM!™ Mints · org S47K8v · sender WAM!™ Mints / support@wammints.com.

LIVE flows (sending today — DO NOT touch unless David explicitly asks):

idtriggercadence
WelcomeRZ8cagAdded to Email List VQv2SEv3.1
#02 Site abandonUMQuaiseg TME5ML2-email, 7-min entry
#03 Browse abandonUsPFJYseg WKuNc73-email
#04 Cart abandonSAgiURseg SAAeLi4-email
#05 Checkout abandonRMYKBsseg VyqXaD4-email

All cascade flows exit on Placed Order WBVQaM == 0 since-flow-start.

DRAFT flows (built, awaiting David's activation gates — DO NOT modify):

idtriggergate
#08 Review/UGCTekexBmetric UcuU3F Fulfilled Order +10dreviews-widget URL needs real Shopify anchor
#09 WinbackYwYCHEseg Szr3Ma (lapsed 60d)WAMBACK20 needs to be a real Shopify discount

Lists: Email VQv2SE (~844, also the popup target) · Preview TXrT2M · SMS X6x87U.

Collision rule for campaigns: any campaign sent to a segment that's ALSO a flow trigger risks double-mailing profiles already in that flow. Default: send campaigns to VQv2SE (the main email list) minus the trigger segments, OR create dedicated campaign segments. NEVER blast a cascade trigger segment without a Smart-Sending check.


Brand voice + HARD RULES

Read first: email/wam/brand-info.md (in repo). It carries the 6 HARD RULES verbatim — they're enforced across every email.

One-line summary of the 6 rules (do not paraphrase in copy — defer to brand-info.md for exact language):

1. No medical / therapeutic claims. Ingredients + "tingle" only.

2. The two named humans are Anderson .Paak (founder, "out on a world tour" — never name a tour/album/group like Silk Sonic/MGM) + Dr. Jonathan Salinas (Harvard- and UCLA-trained laryngologist). Honest credits, no name-dropping of labels/management/entertainment companies.

3. Tagline: "Confidence is meant to be shared™" — used sparingly.

4. Slug for the only product: honey-lemon-ginger-mints (not wam-mints-honey-lemon-ginger).

5. No fake scarcity / fake discounts. Any code mentioned must be real in Shopify with a real expiry. (WELCOME20 20% off + free shipping is the cascade code; WAMBACK20 is winback-specific, currently pending.)

6. No real label / management / entertainment-co names. "Artists out on world tours" / "performers" / "touring crews" — never "MGM Entertainment", "Silk Sonic", any tour or album name.

Voice differentiation — IMPORTANT:


Klaviyo campaign API (operational basics)

The flow lessons mostly apply (and you'll waste hours if you ignore them):

Authoritative source: pull the OpenAPI directly. /usr/bin/curl https://raw.githubusercontent.com/klaviyo/openapi/main/openapi/stable.json → read components.schemas.CampaignCreateQuery + CampaignDefinition etc. Brute-probing the API is a trap.

Carry-over lessons from the flow sessions (codified — don't re-discover):

1. urlencode not url_encode. Klaviyo Django filter is urlencode (no underscore). One bad filter fails the WHOLE template render — silent failure (HTML looks perfect). Affected HF flows; WAM bodies already correct. If you write new Liquid CTAs, use urlencode.

2. Verify every email via POST /api/template-render/ (not visual inspection). It reproduces Klaviyo's UI Preview and catches what visual review misses. Hard rule: render-test before scheduling any send.

3. Templates are PATCH-able as standalones (PATCH /api/templates/{id}); flow-OWNED clones are read-only (PATCH 404). For campaigns the templates you create are standalones, so PATCH works.

4. Segment defs are PATCH-able in place (PATCH /api/segments/{id} → 200, keeps id). Profile-property not-equals excludes profiles where the property is unset — use OR(existence:not-set, not-equals X) if you need the unset cohort included.

5. curl -g for any URL containing [ ] (Klaviyo fields/page params). Python stock urllib has no CA bundle → use /usr/bin/curl.

6. Klaviyo metrics endpoint quirk: /metrics/?page[size]=100 may return 0 (silent); /metrics/ (no params) returns all. Use the former with additional-fields instead, or just paginate without page-size.

7. MCP klaviyo_get_email_template can read any template by id (incl. flow clones). Good for content-verification.

Campaign-specific notes (not yet exercised in WAM, gather as you go):


Credentials + environment


Pending operator gates (informational only — David owns)

Listed so you don't get confused if you stumble into them. None block campaign work.


Cross-references (deep context if you need it)


Open repo state at handoff

Two new files in email/wam/, uncommitted as of this brief's write time:

If David tells you to CMP, follow the rebase-discipline rule (rebase → surgical-add only your files → commit → force-with-lease push the feature branch, never main; creative/dashboard/index.html is hook-dirty and must NEVER be committed).


First actions for this session

1. Read this brief (you just did).

2. Read email/wam/brand-info.md — voice + the 6 HARD RULES verbatim.

3. Pull current campaign inventory to know what's already been sent: GET /api/campaigns/?filter=equals(messages.channel,'email') (or use the MCP). Lays groundwork for the calendar.

4. Ask David: what's the calendar window (month / quarter / through year-end)? What recurring slots (weekly newsletter? monthly drop?). What one-shots (holidays / collabs / restocks)?

5. Audience strategy first, copy second. Map: who's the campaign audience, who's getting it via what flow already, where's the collision risk.


Hard discipline (apply throughout)

From memory wam_klaviyo_status.md

⚠️ WAM brand-copy HARD RULES (David, 2026-05-15)

Five rules, all enforced in email/wam/brand-info.md (inline) + applied across the full 13-email cascade:

1. No "every night." Anderson .Paak is NOT "on stage every single night" — not a nightly headline set. Never write "every night" / "on stage every single night."

2. Never name the headliner — no "Bruno" / "Bruno Mars" anywhere. The brand must not piggyback that name in copy. ONLY say "on a world tour" / "out on a world tour" / "currently on a world tour". (David, 2026-05-15: "Let's not mention Bruno. Let's just say on a world tour." This SUPERSEDES the earlier same-day framing that still named Bruno.)

3. Full brand name with trademark, everywhere: WAM!™ Mints (renders WAM!™ Mints). Never a bare "WAM!". 41 normalizations applied across the cascade.

4. Never seed the "marketing gimmick" thought. Do NOT say the product "wasn't dreamed up in a marketing meeting" or any phrasing that plants the idea it could be a marketing ploy — never voice the negative even to deny it. (Removed from #04 d0.) Also removed the standalone "That's not a marketing line." sentence from #04 d2.

5. The product is "the propolis mint" — NO tingle/sensation claim, EVER. The mints do NOT give a tingle/buzz/sensation. David (2026-05-15): "Instead of mints with a tingle, let's say the propolis mint. The mints don't give you a tingle." Never write "tingle" / "the tingle" / "mints-with-a-tingle" / "one big tingle" / "a bright little buzz" / any spilanthes sensory-payoff hook. Name ingredients plainly. Sub-rule: performers carry it in a "bag", never a "mic bag" — David: "Singers don't have a mic bag. You can say bag in general." 10 tingle spots + 5 mic-bag spots scrubbed from the cascade 2026-05-15. Also: the #04 d3 closing P.S. ("Made by the guy out on the road and the doctor who got tired of watching performers white-knuckle through menthol…") was removed entirely per David screenshot ("remove this").

6. Never name "MGM Entertainment" — or ANY specific real label / management / residency / entertainment company (David, 2026-05-18: "Lets not mention MGM Entertainment"). Same class as rule #2 (Bruno): the brand must not piggyback a real entity's name. Dr. Jonathan Salinas is described ONLY as "a Harvard- & UCLA-trained laryngologist" / "in-house physician for artists out on world tours" — never tied to MGM (or any named company). Root-cause seed was brand-info.md line 6 ("…in-house physician for MGM Entertainment artists") → scrubbed to "…for artists out on world tours" + rule (6) added to the brand-info.md HARD RULES block so future copy isn't re-seeded. Body fix: 03-browse-abandon-d1.html:116 (only body occurrence). 0 "MGM" across all bodies/popups/subject-lines/brand-info ✓.

2026-05-18 edits (David screenshots, WAM session)

Root cause of #1 was a misleading line in brand-info.md ("Anderson is on stage every night"), now replaced with all 6 rules. Same root-cause class hit #5: brand-info.md line 13 ("Acmella oleracea — the tingle") + line 84 campaign ("One big tingle") seeded the tingle concept — both now scrubbed/overridden so future copy sessions aren't re-seeded. Cascade fully scrubbed 2026-05-15: 0 "Bruno"/"marketing meeting", 0 "tingle", 0 "mic bag", 41× "WAM!™ Mints", 0 bare "WAM!" (residual grep tingle\|[Mm]ic bag\|marketing line over 0[2-5]-*.html = NONE ✓).

OPEN FLAGS (David decisions):

🟢 2026-05-18 — WAM CASCADE #02–#05 SET LIVE by David (MILESTONE)

David flipped all 4 cascade flows draft→live himself ("Great I set them all to live!"). LIVE + sending to real WAM subscribers now: #02 UMQuai · #03 UsPFJY · #04 SAgiUR · #05 RMYKBs (+ welcome RZ8cag). Activation gate CLEARED. Verified via API (all status=live).

⚠️ NEW OPERATIONAL HARD RULE — editing a now-LIVE WAM cascade flow: flow content is API-immutable → any copy/slug/timing change still = delete+recreate. But these are LIVE: the safe order is now create the NEW flow (draft) → verify → PATCH it live → THEN delete the old live flow (minimize the unprotected window). Deleting a live flow also drops any profiles currently in-flight in it (they won't resume in the new flow). Acceptable for low-volume early days but never delete-first on a live flow. (This session's #03 slug-fix used delete-first while it was still DRAFT — fine then; going forward they're LIVE.) Re-flagging: account default sender still bare "WAM! Mints" (per-email from_label correct); logo lockup (logo-wordmark.png sub) still wants David's eyeball.

🟢 2026-05-19 — #08 Review/UGC + #09 Winback FLOWS BUILT (DRAFT) — supersede skeletons RZucbb/QSFBuc

Built in Klaviyo from the 5 drafted bodies, mirroring the LIVE cascade pattern (sender WAM!™ Mints/support@wammints.com, smart-sending on, 7-min entry on segment-triggered). All standalone templates created → render-gated (200) → flows POSTed → structural+render-verified → old DRAFT skeletons deleted (create-before-delete).

NEW flow id (DRAFT)triggerprofile_filtercadenceseedsclones
#08 Review/UGCTekexB (was RZucbb)metric UcuU3F Fulfilled Order (Shopify)none (Smart Sending)delay 10d → D0 → delay 5d → D1XG6Pdn/SRcWXaWPbZyz/SBjbUY
#09 WinbackYwYCHE (was QSFBuc)segment Szr3Ma WAM Lapsed (no order 60d)Placed Order WBVQaM count==0 since flow-start (exit on re-purchase — audience IS past customers so alltime would empty it; flow-start is correct exit-on-new-conversion semantics)delay 7m → D0 → delay 3d → D1 → delay 3d → D2Y9Rm9d/XwQiwA/V7pnpuXTXA73/TG6Lts/XgWxuq

⚠️ ACTIVATION GATES (David, before flipping draft→live):

1. #08 reviews-widget URL — bodies' CTAs use https://www.wammints.com/products/honey-lemon-ginger-mints?...#reviews (placeholder anchor). Confirm the Shopify PDP reviews app's actual anchor/URL; if different, update bodies + recreate templates. (Wrong anchor = CTA scrolls nowhere → review-rate kill.)

2. #09 WAMBACK20 must be a REAL active 20%-off Shopify discount code with a real expiry (honest-discount HARD RULE, same class as WELCOME20/RECOVER83/RECOVER88 — false code destroys list). Body promises 20%; currently a placeholder. David creates the code in Shopify, then flips live.

Build lessons applied: WAM bodies are pure HTML (zero Liquid → render-test passes trivially, no url_encode/template-render trap risk like HF). Schemas honored — entry_action_id+temporary_id+links.next chaining; minute time-delays must omit delay_until_weekdays (only days-unit accepts it). Klaviyo metrics list endpoint returned 0 with page[size]=100 but 56 with no params — query quirk worth remembering. Used direct GET /metrics/{id} to confirm UcuU3F = Fulfilled Order (Shopify integration) before wiring.

2026-05-18 — 08-review-ugc + 09-winback BODIES DRAFTED (flows NOT built yet)

David scoped "iron out" → just 08-review-ugc + 09-winback (the 2 skeletons in his screenshot); I draft per the email-copywriting skill lifecycle mapping + WAM brand HARD RULES 1-6, he reviews on the preview + screenshot-iterates (like the cascade), THEN I build the flows. 5 bodies written (email/wam/bodies/, commit ada524d, preview live on wam-cascade-preview.netlify.app): 08-review-ugc-d0 (~10d post-Fulfilled-Order review/UGC ask) + 08-review-ugc-d1 (last-nudge) — REQUESTS reviews, never fabricates (register_wam audit: 0 real reviews exist); 09-winback-d0 (pattern-interrupt "your bag's been quiet", no offer) + 09-winback-d1 (the offer) + 09-winback-d2 (friendly breakup/permission = Sunset self-clean). WAM cascade HTML skeleton + friend-texting voice, Klaviyo-CDN imgs, correct slug honey-lemon-ginger-mints; brand rules verified 0-violation (Python audit — note: keep banned tokens OUT of doc-comments too; they false-flag audits + the WAM root-cause discipline says don't let trigger words ride the pipeline — comments now just cite brand-info.md).

Build-time flags (resolve before/at flow build — flows are NOT built yet):

1. 08 review CTA = …/products/honey-lemon-ginger-mints#reviews PLACEHOLDER — confirm the actual Shopify reviews app/anchor (Judge.me/Loox/Okendo/native) with David.

2. WAMBACK20 (09-winback offer) = PLACEHOLDER — must be a REAL active Shopify discount before 09 goes live (honest-discount HARD RULE, same class as cascade WELCOME20 / HF RECOVER). 09-d1/d2 use Shopify /discount/WAMBACK20?redirect=/products/honey-lemon-ginger-mints form.

3. 09-winback trigger: the QSFBuc skeleton triggers on Placed Order — WRONG (winback = lapsed/inactivity). Build on a LAPSED segment ("WAM customers, no order in ~60d") — that segment likely needs creating (no ready lapsed segment in the live list). 08-review-ugc trigger = Fulfilled Order + ~10d entry delay (correct).

4. When building: same proven pattern as cascade, but 08/09 skeletons are DRAFT (RZucbb/QSFBuc) so delete-first is fine (NOT live like #02-#05); fix the 09 trigger; reuse Klaviyo-CDN imgs (already uploaded).

2026-05-18 — WAM CASCADE FLOWS #02–#05 BUILT IN KLAVIYO (built DRAFT; now LIVE per above)

David: "lets start putting this into Klaviyo." The #02–#05 flows previously existed only as metric-triggered placeholder skeletons (fake subjects "Forgetting something?" etc., SMS/conditional-split stubs — NOT the real cascade copy; the #02 skeleton even had a trigger bug, firing on Viewed-Product which is #03's signal). Replaced all four with the real edited bodies, segment-triggered (David's pick), proven HF REST pattern (standalone templates → POST flow → GET-verify 7/7 → delete skeleton; create-new-before-delete).

CURRENT WAM CASCADE FLOW IDS — all DRAFT, nothing sends:

FlowIDTrigger segmentEmails (cadence from entry)Skeleton deleted
#02 Site AbandonedUMQuaiTME5MLD0 +7min · D2 +48h (2)~~QVx7Gw~~
#03 Browse AbandonedUsPFJYWKuNc7D0 +7min · D1 +24h · D3 +72h (3)~~TJ7jq5~~
#04 Cart AbandonedSAgiURSAAeLiD0 +7min · D1 +24h · D2 +48h · D3 +72h (4)~~YeuagH~~
#05 Checkout AbandonedRMYKBsVyqXaDD0 +7min · D1 +24h · D2 +48h · D3 +72h (4)~~Rc28M4~~

2026-05-18 (cont.) — WRONG PRODUCT SLUG fixed (#03 only)

David: CTA link wrong — wammints.com/products/wam-mints-honey-lemon-ginger → /products/honey-lemon-ginger-mints (the LIVE welcome v3.1 already used the correct slug; confirms it). Wrong slug was in only the 3 #03 browse-abandon bodies (d0/d1 product CTA, d3 discount/WELCOME20?redirect=/products/… form) — 1 each; #02/#04/#05 + popups clean. Literal slug replace, 3 #03 seeds re-synced, flow #03 recreated QX6W7a→UsPFJY (7-min entry + spec preserved), seed content-proven, preview redeployed+verified. Commit f70fed2. #03 ID is now UsPFJY (supersedes QX6W7a; #02 UMQuai/#04 SAgiUR/#05 RMYKBs unchanged).

2026-05-18 (cont.) — IMAGE-HOST FIX: cascade images were broken in Klaviyo

David caught: all images broken-link in the #02–#05 Klaviyo flows. Cause: the 13 cascade bodies referenced images via relative ../assets/X — resolves only on the preview Netlify site (assets deployed alongside), but Klaviyo renders the template standalone with no base URL → every image broken. The initial build cloned the broken-path bodies verbatim. 🔑 REUSABLE TRAP (same class as the HF preview-Liquid trap): a static preview making images look fine ≠ they work in Klaviyo. Klaviyo email bodies MUST use absolute, externally-hosted image URLs — NEVER relative paths. Audit <img src> for ..//relative BEFORE building any Klaviyo flow. Sub-trap: Klaviyo's image API rejects SVG ("only JPEG, PNG and GIF") — logos/icons must be raster; the SVG logo ref was repointed to the raster logo-wordmark.png (⚠️ that's the plain wordmark, not the exact …hlg-wordmark lockup — David to eyeball in preview/flows; swap = 1 image upload + remap).

Fix (David's pick: Klaviyo media library — matches the working LIVE welcome which is Klaviyo-CDN-hosted): uploaded the 11 unique images to Klaviyo's media library (WAM org S47K8v) via POST /api/images import_from_url from the public preview URLs → https://d3k81ch9hvuctc.cloudfront.net/company/S47K8v/images/…. Rewrote every ../assets/X across the 13 cascade bodies to the Klaviyo CDN URLs (repo commit f158444), re-synced the 13 seed templates, delete+recreated all 4 flows from the patched seeds (content API-immutable). Seed content-proven (0 ../assets, Klaviyo-hosted imgs); preview redeployed + verified (imgs 200 from Klaviyo CDN, now matches what Klaviyo actually sends). The 13 bodies are now permanently absolute-URL (the local email/wam/assets/ + the preview assets/ dir are no longer used by the cascade — relative-path fragility eliminated). git stash+rebase recovery note: an unrelated hook-modified creative/dashboard/index.html was dirty and blocked a scoped stash → stash-ALL then surgical-add only the 13 (never git add -A); rebase replayed clean, both edit layers (screenshot edits + image rewrites) verified intact, 0 conflict markers.

Status (last updated 2026-05-15): #04 cart-abandon 4-day flow built + client-preview deployed; popup chain restored; cascade segments live

WAM Klaviyo (WAM!™ Mints, org S47K8v) backs the wammints.com Shopify storefront. After today's session WAM has the full popup → list → welcome flow chain working again, plus 4 cascade abandonment segments wired to real Shopify integration metrics, plus Klaviyo onsite JS finally installed for browser-fired events.

Live infrastructure

Klaviyo account: WAM!™ Mints

Lists (post-cleanup 2026-05-14 — David deleted Newsletter + VIP Members)

IDNameMembersNotes
VQv2SEEmail List844Active subscriber base — popup target + welcome flow trigger source
TXrT2MPreview List1Klaviyo default
X6x87USMS List0Not yet activated
~~V63GEZ~~~~Newsletter~~~~deleted~~Was popup + welcome trigger target — broke chain on deletion (see below)
~~QSaT3q~~~~VIP Members~~~~deleted~~

Cascade segments (4, all rebuilt 2026-05-14 with real metric logic + 7d windows)

IDNameMembers todayDefinition
TME5MLWAM #02 Site Abandoned (7d)0Active on Site (WsFjtZ) > 0 in 7d AND Viewed Product (SeNvQq) = 0 in 7d
WKuNc7WAM #03 Browse Abandoned (7d)0Viewed Product > 0 in 7d AND Added to Cart (Shopify XMHbWw AND API UAH5Yq) = 0 in 7d
SAAeLiWAM #04 Cart Abandoned (7d)1Added to Cart (either source) > 0 in 7d AND Checkout Started (VmTJVx) = 0 in 7d
VyqXaDWAM #05 Checkout Abandoned (7d)3Checkout Started > 0 in 7d AND Placed Order (WBVQaM) = 0 in 7d

#02 + #03 = 0 today is normal — Klaviyo onsite JS was just turned on 2026-05-14, so Active on Site + Viewed Product events haven't accumulated identified-profile volume yet. Counts will grow over 24-72h as identified visitors browse.

Segments — ACTUAL live state (verified via API 2026-05-18, supersedes the stale list below)

The "12 from earlier scaffold-segments-api.py run" were ALL DELETED — direct GET on QZ5S6d/R6EnuS/RSVVNh/TAJyXn/TWUAuy/UzVD7b/Y6a9LK/VQUYpG/WDcxH7/Xu6mdp/XyHByM = HTTP 404 (gone). (Memory-is-point-in-time lesson: the scaffold segments did not persist / were cleaned; never assume an old segment list — GET /api/segments/{id} to verify before reusing.) Current WAM segments = ONLY:

~~### Other segments still live (12 from earlier scaffold-segments-api.py run) — STALE, all 404, see above~~

Key Shopify ↔ Klaviyo metric IDs (discovered today)

Flows (11 in WAM Klaviyo, mostly DRAFT skeletons from earlier scaffolding)

Popup

Klaviyo onsite JS (turned on 2026-05-14 by David)

TRAP — deleted Klaviyo list silently breaks popup + flow trigger

Established 2026-05-14 incident.

What happened

David cleaned up WAM Klaviyo by deleting the Newsletter list (V63GEZ) and VIP Members list (QSaT3q) — they looked like clutter. But:

1. The exit-intent popup was hardcoded to subscribe to V63GEZ (KLAVIYO_LIST_ID = 'V63GEZ' in the JS in theme.liquid). Every popup signup since the deletion silently failed — Klaviyo's Client Subscribe API returns 202 even when the list is missing, so the popup showed the success state but the email went nowhere.

2. The 01-welcome (v3.1 LIVE) flow trigger was "Added to Newsletter list" — also referenced the deleted V63GEZ. Even after the popup was repointed, new signups to Email List wouldn't have triggered the welcome email until the flow trigger source list was also updated.

How long this was broken

Unknown — between the deletion timestamp and 2026-05-14 ~14:50 PT when David noticed counts on the cascade segments were 0 and asked which list the popup wrote to. Some unknown number of popup signups landed nowhere during the window.

Fix (took ~30 min total)

1. Updated popup source email/wam/popups/wam-welcome-popup-shopify.html to use VQv2SE (Email List) — committed f260772

2. Pulled live theme.liquid via Shopify CLI, patched KLAVIYO_LIST_ID = 'VQv2SE', pushed back via shopify theme push --theme=149521301641 --allow-live --only=layout/theme.liquid

3. Verified live wammints.com serves the new list_id

4. David updated 01-welcome flow trigger via Klaviyo UI: source list Newsletter → Email List

Action items for ALL future sessions

HARD RULE: never delete a Klaviyo list without first auditing what references it. Before deleting any list, grep:

```bash

1. Search the repo for the list_id

grep -rn "<LIST_ID>" --include=".html" --include=".py" --include=".js" --include=".md" --include="*.yaml"

2. Pull the LIVE Shopify theme + grep theme.liquid for the list_id

shopify theme pull --theme=<LIVE_THEME_ID> --store=<STORE> --only=layout/theme.liquid

grep "<LIST_ID>" /path/to/pulled/layout/theme.liquid

3. Check Klaviyo flows whose trigger_type='Added to List' for the deleted list

(Klaviyo's flow-trigger details aren't easily exposed via API — verify in UI)

```

If ANY reference exists, repoint to a kept list FIRST, then delete the source list. The cost of a 5-min audit is much less than silently breaking the entire popup → welcome flow chain for an unknown number of signups.

Generalize: any time a brand admin deletes a Klaviyo list, Shopify metafield, 29next coupon, GTM tag, etc. that's referenced by automation, the deletion can produce silent failures. Default to "rename + leave" unless you're certain nothing references it.

TRAP — netlify deploy silently hit the WRONG site for a whole session

Bit 2026-05-15. The client preview is wam-cascade-preview.netlify.app (siteId 6dfa0488-8bf0-485c-9e2b-2505a70e11fc) — the URL in this note + manifest, the link the client holds. But /tmp/wam-cascade-preview/.netlify/state.json had drifted to siteId aebc4378-50a4-4782-82ff-ac7c65f530da = a stray auto-named site dancing-gaufre-9c1686.netlify.app. Every "preview redeployed + CDN-verified" this session (round-2 copy corrections AND the UGC video) landed on dancing-gaufre; the client URL stayed stale (UGC 404, tingle residual) until David asked "why didn't you deploy to wam-cascade-preview?".

Root cause: trusted "✔ Deploy is live" + the directory being named wam-cascade-preview, and verified on whatever URL the deploy command echoed back — never asserted it was the intended site. The actual target lives only in .netlify/state.json, not the folder name. This is the same class as the attribution "don't trust 'it deployed fine'" trap and the adscaler "READ the active account context FIRST" rule.

HARD RULE — before AND after every netlify deploy: assert the target site is the intended one.

SKILL: IG-reel frame → email UGC video-card (reusable, all brands)

Codified 2026-05-15 building #03 d1's social-proof video block (commit 5a16976). David: "insert this video thumbnail in the ugc one about social proof" + Reel URL. Reusable pipeline + the traps that bit:

1. Identify the email first — "the ugc one about social proof" = the cascade's dedicated social-proof email (WAM #03 d1 "Who actually keeps these"). Confirm target + image-source method with AskUserQuestion before touching the browser.

2. Pull the frame via the user's logged-in Chrome (IG WebFetch = login wall; oEmbed dead). Chrome MCP: list_connected_browsers → ask user which browser → select_browser → tabs_context_mcp(createIfEmpty) → navigate. IG reel needs a user gesture: a computer left_click on the video, then ~5s, before <video> readyState reaches 4. Drive playback by JS: v.muted=true; v.play(); v.currentTime=T; v.pause() then wait ~2s for the seeked frame to decode (readyState dips to 1 mid-seek — wait it out, re-assert before capture).

3. Frame grab = canvas, NOT screenshot. canvas.drawImage(video) then toDataURL('image/jpeg') — IG reels stream via same-origin MSE blob, so the canvas is NOT tainted (cross-origin would throw SecurityError). This yields the pure video pixels — none of IG's DOM chrome (no play button / username bar / like rail), far better than a screenshot+crop.

4. TRAP — two exfil paths are blocked, one works:

- javascript_tool returning base64 → [BLOCKED: Base64 encoded data] (privacy guard, any size). JS returns also truncate ~1.5KB.

- Chrome MCP computer screenshot save_to_disk:true → the saved path is not surfaced and the file is not in any Bash-reachable dir (ls/find may stat it once then it's gone; sandbox).

- WORKS: in-page <a download> of the canvas blob (needs explicit user OK — it's a file download). It lands in the user's Chrome Downloads. If that Chrome is a different machine / the sandbox blocks the cross-tree read (it does for ~/Downloads), have the user drop the file into the repo email/<brand>/assets/ (David put it in canonical /dev, not the worktree → cp it across; git hash-object both to prove byte-identity before any rm+pull).

5. Bake the play button into the JPG with Pillow (/tmp/wam_play_overlay.py): resize to email weight (600w keeps 9:16 ≈ 77KB @ q84), soft radial dark scrim for contrast, brand-color disc + cream ring + drop-shadow + cream triangle (nudge apex right ~10%). No CSS/overlay = Outlook-safe. Commit BOTH the raw frame and the -card composite; email <img> references the -card.

6. Block placement: linked card after the social-proof pattern copy, before the founder/authority block (real-user proof → authority proof escalation). Whole <img> + a "▶ Watch on Instagram · @handle" caption both link the reel (no UTM on IG links — ignored, noisy).

7. Brand-rule the chosen frame: the reel's burned-in caption is part of the asset — reject any frame whose caption violates a HARD RULE (skipped t=35 "EVERY SINGLE NIGHT" → violates WAM rule #1; used t=24 "YOU KNOW LIKE", clean). First-party check: confirm whose account the reel is (here @wammints verified = WAM's own → no third-party creator rights/likeness issue, not Anderson/headliner).

/dev sync caveat: a user-added untracked asset that's byte-identical to the committed blob blocks git pull --ff-only ("untracked would be overwritten"). Prove identity with git hash-object, then rm the untracked dupe + pull (restores identical tracked bytes, zero loss).

Outstanding work for WAM

Body content for cascade flow emails (4 still empty)

The flows 02-site-abandoned, 03-browse-abandoned, 04-abandoned-cart, 05-abandoned-checkout are scaffolded but have no email content. Same approach as HF: each gets a tailored email per the cascade-by-intent design (HF's cart-abandoners-e1.html / leads-education-e1.html are good reference patterns — though WAM's brand DNA, tone, and product context differ entirely).

Welcome series E2 + E3

01-welcome E1 (v3.1 LOCKED) is the only welcome email shipped. E2 + E3 templates exist as placeholders in WAM Klaviyo (01 Welcome · E2 · Brand story (placeholder) + 01 Welcome · E3 · Best-sellers (placeholder)).

Popup signup recovery (probably impossible)

The popup signups that failed during the broken window (between Newsletter deletion and 2026-05-14 fix) cannot be recovered — Klaviyo's Subscribe API returns 202 even on failure, so the failed POSTs aren't logged on the receiving side. Going forward fixed.

When extending to NS / SkinScience / etc.

Same pattern applies:

September 2026 (built 2026-08-25, NOT pushed): six sends live in

email/wam/campaigns/2609* — Vol. 16–21, continuing the A→B→C→D pillar rotation.

Review page: wam-cascade-preview.netlify.app/september-review.html. Klaviyo has

zero September campaigns and stays that way until David judges. Send time is

always 14:00 UTC = 10:00 ET (16:00 UTC for the Sat 9am PT promo). Vol numbers

are sequential per SEND including holiday promos (Aug 18/19 both got Vol. 14 — a

duplicate, don't copy that).

LABOR20 is a BLOCKING operator gate for the Sep 5 promo: 20% + free shipping,

expires Mon Sep 7 11:59pm PT. The email links carry &discount=LABOR20, so a

missing code silently walks people to full price.

Testimonial emails pull REAL reviews from Stamped.io — public widget API,

storeUrl=wam-mints.myshopify.com, no auth. Filter: reviewVerifiedType==2, 5★,

drop medical-adjacent language (the brand's own AVOID list includes dry mouth), drop

emoji, and drop any review containing WAM!™ — the older ones do, and quoting

verbatim would break the ® rule. Never invent a review or a reviewer's job title.

From memory wam_smile_loyalty_status.md

The WAM!® Fam — WAM!® Mints rewards program, Smile.io backend. LIVE as of 2026-07-18.

Live: page at /pages/join-the-wam-™-fam (handle has a ™ — cosmetic, works

fine, deliberately NOT renamed because the nav item points at it), linked in

Main menu as "Rewards". Smile program activated; launcher live site-wide.

Repo: products/wammints/loyalty/ — lp/index.html (also standalone at

wam-fam-rewards.netlify.app), shopify/ (section/template/css pushed to live

theme #149521301641), SMILE-SETUP.md (config checklist).

Economics (live, verified): currency Notes · earn 5/$1 base, 100 join,

250 birthday, 100 review, 50 IG, 50 TikTok, 150 share · redeem 500=$5,

1000=$12, 400=free ship, 2000=free tin, 3000=free pouch · tiers by

CALENDAR-YEAR spend Rising Star $0 / Headliner $100 / Encore $250 with

5/7/10 Notes per $1 · referral $10/$10 · points expiry OFF. Plan = Growth

($199/mo, was in free trial — confirm it bills as Growth, the Shopify app card

showed a stale $79).

Two traps worth remembering:

1. Smile lazy-loads in TWO hops — app embed → smile-loader.js → (on

window.load) smile-lite → defines loadSmileSdk() → calling it defines

window.SmileUI. A synchronous if (window.SmileUI) check at click time is

ALWAYS false, which silently sent every CTA to /account/register (the store

login) instead of the rewards panel. Warming on window.load is ALSO too

early. Must poll for loadSmileSdk. See the handler in lp/index.html.

2. The LP carries its own sticky header (it's served standalone too). The

Shopify converter MUST strip it or the page renders two stacked navs.

Hero reels: 8 real UGC clips from the ad pipeline. Sources live OUTSIDE this

repo at dev/creative/wam/videos/ (1080x1920, 38–121MB). Encoded to 5s silent

360x640 loops, 1.6MB total, in lp/reels/ and as theme assets

wam-fam-reel-01..08.mp4. They carry burned-in ad captions, so the cards have

no captions of their own (would double-stack).

Open / optional:

See [[shopify_embedded_app_iframe_automation]] · [[wam_klaviyo_status]] ·

[[occluded_tab_testing_trap]] (media won't decode in the hidden automation tab —

verify video in the Browser pane).

Rendered 2026-11-16 from the git-tracked file. Edits happen in the repo (Claude files every fact David sends the same turn); this page is the read view.