Answer
Arata is the design partner furthest along: the only one with real accounts connected, Meta Ads and Shopify, so every rule that protects a live client starts here and its connections are never touched.
Today
Profile
| Fact | Source |
|---|---|
Slug arata-in, store arata.in |
memory brew-live-stores |
| A hair care brand; the code calls it a hair fall brand | brew/ai-questions.js:4 |
| The first brand carried into Brew on 15 Oct; it was the hand made pack the read was modelled on, now an ordinary brand pack | brew/SCOPE.md:43-55 |
| Claimed for the brand by our team's staff address; anyone with an arata.in email joins the team on their own | brew/auth.js:106-107 |
| Home is production, feed.joinsquareup.com, since 25 Oct; staging is for testing only | memory brew-prod-backend brew/deploy.sh:19-20 |
| A real design partner prospect, one of five; never wiped or reset | memory brew-live-stores |
Connections
| Account | State | Source |
|---|---|---|
| Meta Ads | Ad account act_2942894532542334, connected 22 Oct. The grant is ads_read, business_management and pages_show_list; the app is on development access until Meta App Review | memory brew-meta-connector-qa memory brew-meta-app-review |
| Shopify | Connected 23 Oct through the brand's own app, route own-oauth; the offline token never expires | memory brew-shopify-connector-qa memory brew-prod-backend |
| Protection | PROTECTED_CONNECTIONS defaults to arata-in: disconnect, re-select and reconnect are refused unless an admin forces it |
brew/connectors/protected.js:2-3 brew/connectors/index.js:267 |
| Both sites | The same tokens read on staging and production; nothing was refreshed or revoked in the copy | brew/db/partner-copy.js:7-8 |
| Everything else | No GA4, Google Ads, Klaviyo, Judge.me or Slack connection on record | brew/docs/audit-2026-10-24-design-partners.md:3 |
What the accounts showed
| Figure | When | Source |
|---|---|---|
| About 1,200 orders a day, about 35,000 in 30 days | 23 Oct | memory brew-shopify-connector-qa |
| 97% of orders come through the GoKwik checkout app; Shopify's 4,411 abandoned checkouts are only carts that reached Shopify's own checkout | 23 Oct | memory brew-shopify-connector-qa brew commit 19c87c6 |
| Cash on delivery: 2% cancelled, 37% shipped and still unpaid; never called failed | 23 to 24 Oct | brew commit 04571a2 brew/connectors/shopify.js:203-214 |
| 11 of 12 zero stock products keep selling at zero | 23 Oct | memory brew-shopify-connector-qa |
| One campaign's Meta frequency 4.05, where an older reading said 1.7 | 23 Oct | memory brew-shopify-connector-qa |
| Live Meta ads shown on a card: 1,577 | 22 Oct | brew/card-nums.js:2 |
History
| Date | What happened | Source |
|---|---|---|
| 15 Oct | A 50 Play rewrite stopped when the AI credit ran out; 7 Plays published by hand | memory brew-arata-rewrite-2026-10-15 |
| 15 Oct | Arata agreed its team notes stay public on the demo site | memory brew-audit-2026-10-15 |
| 16 Oct | Final run: 56 of 56 Plays | memory brew-go-live-2026-10-17 |
| 19 Oct | Our team's deck for Arata became context the Plays work to | memory brew-absorb-all-learnings |
| 20 Oct | All Plays redone on Opus, 54 of 56, about $65 to $70 | memory brew-arata-reload-2026-10-20 |
| 20 Oct | Pro against Opus compared on 54 pairs: Opus clearly better, confounded by inputs | memory brew-model-compare-2026-10-20 |
| 21 Oct | Sloosh video tests run on Arata; its viral final cost 80.8 credits | memory brew-viral-remakes |
| 22 Oct | Meta connected; the connector and the privacy of its numbers fixed the same day | memory brew-meta-connector-qa |
| 23 Oct | Shopify connected; 11 store Plays rewritten by hand for $3.40 | memory brew-shopify-connector-qa |
| 23 Oct | Studio tested live on Arata: listing set, 360 degree turn, packshot | memory brew-jam-v2 |
| 23 to 24 Oct | Round 2 QA report on Arata: 32 pass after round 4 | memory brew-qa-round2-2026-10-24 |
| 24 Oct | Design partner audit of all five | brew/docs/audit-2026-10-24-design-partners.md |
| 25 Oct | Copied to production, last of the five; its home briefly moved back to staging for a 26 Oct onboarding, then to production again | memory brew-prod-backend brew commit deb7d90 brew commit 119dc8a |
| 25 Oct | Feed drafts on Arata were collages; every product picture is now judged before it shows | memory brew-draft-quality |
| 25 Oct night | Re-check on production: 25 pass, 4 fixed, 3 pass with a note | memory brew-qa-round2-2026-10-24 |
| 26 Oct | Param's screenshots of a ₹2,182 combo card led to five jam and brief fixes | memory brew-jam-rca-2026-10-26 |
| 26 Oct | A WhatsApp broadcast draft led to a claim check rule | brew/claims.js:201 |
Open issues
- "Fix the pack prices" repeats across about six cards; no dedupe across Plays yet memory brew-arata-reload-2026-10-20
- Some cards speak in a we and our voice memory brew-qa-round2-2026-10-24
- MitoActive shows no Shopify units for 30 days, possibly a preorder memory brew-qa-round2-2026-10-24
- The first load after a release is slow memory brew-qa-round2-2026-10-24
- Meta App Review is blocked on business verification and two screencasts, which are recorded on Cozy, never Arata memory brew-meta-app-review
Next step
No commercial next step, contact name or agreement is on record. See Open.
Tomorrow
| Item | Version | Tag |
|---|---|---|
| Arata stays the first brand every fix is checked on, and its connections stay protected in code | v1 | Stays |
| A named contact, a weekly check in and a feedback log entry per call | v1 | New |
| A written design partner agreement covering data use, the demo and the end of the partnership | v1 | New |
| First brand offered posting with a yes, once v2 ships | v2 | New, proposed |
| First brand offered results, since it is the only one with Shopify and Meta connected | v3 | New, proposed |
How we do it
- Never disconnect, reconnect or re-select Arata's Meta or Shopify, on either site. Disconnect tests run on Cozy Snugglet.
- Read only audits with the stored tokens are fine:
node tools/meta-audit.mjs arata-in
node tools/shopify-audit.mjs arata-in
node db/connector-ping.mjs arata-in
- Any cleanup that touches
private/vault/meta.json,connections.jsonor the disconnect action skipsarata-in. - Before a long paid run on Arata, check the AI credit and ask other sessions not to deploy.
- Check a fix on Arata first, then on the other four partners.
Numbers
| Metric | Definition | Where measured | Status |
|---|---|---|---|
| Connected accounts | Real accounts reading, not simulated | Connector ledger, db/connector-ping.mjs |
Today: 2 |
| Team members | People from arata.in who joined | private/members.json |
Proposed, not counted yet |
| Hand offs a week | Briefs shared or sent from Arata's jams | Jam records | Proposed |
| Paid spend on Arata | AI, Studio and finals cost per month | brands/arata-in/spend/ |
Proposed |
Open
- Who at Arata do we work with? No contact name is on record. Proposal: add the main contact, their role and how they prefer to talk to this page.
- What is the next step with Arata? Proposal: a 30 minute call this week to walk the Plays they rated highest, then agree a design partner agreement from Contracts.
- Does Arata stay on the demo site? It agreed on 15 Oct that its team notes stay public there, but it now has real accounts connected. Proposal: keep only its public pack on the demo and ask Arata to confirm in writing.
- Is our team's deck for Arata a paid engagement? The sources do not say. Proposal: record whether Arata pays our team separately, so the Brew partnership stays independent.