Phase 1 · Project Plan · 2026-05-06

How do we get reviews-per-bride from 1.1 → 5+?

The pre-2018 mechanic produced 50,000 reviews a year. We removed it, and per-bride reviews collapsed. Phase 1 restores the multi-supplier capture pattern through three triggers feeding one conversational agent — without rebuilding the dashboard the CEO killed.

OwnerAlec Butterworth SponsorCEO (in-person approval) Started2026-05-04 StatusExploring → Building

The goal, decomposed

Same brides, more reviews each. Not "find 30,000 more brides." Both the CEO and Leanne (WedXM lead) independently named this lever.

Today
~10,000
reviews per year · 1.1 per bride
Pre-2018 baseline
~50,000
5–7 reviews per bride · mandatory
Phase 1 target
5×
restore the multi-supplier pattern

The system

Three input triggers. One conversational agent. One review-submission path.

Trigger 1

Bride email cascade + SMS

Day 3 email → Day 6-7 nudge → Day 10-12 SMS. Cascade stops on first click.

Trigger 2

WedXM supplier-nominated rail

Vendor marks wedding complete → SMS to bride references vendor by name → link.

Trigger 3 · NEW

Registered-bride channel

Past wedding + booked suppliers in dashboard + no review yet → automated outreach.

→

Conversational AI Agent

One endpoint at /reviews/agent/<token>

Pre-fills booked suppliers where data available · multi-supplier per session · captures rating + spend + her words → formatted review

Reviews submitted to admin Validation gate enforces "format her words, never generate" · ≤2★ routes to human review queue

Phase scope · now vs deferred vs killed

Phase 1 — Now

Three triggers + one agent

  • Email cascade rebuild + deliverability
  • WedXM SMS rail
  • Registered-bride channel
  • Conversational AI agent
  • Multiplicative incentive (5 = 5 tickets, 10 = 15…)
Phase 2 — Deferred

Claim-your-storefront

When a non-advertiser supplier gets a review, they receive a free invitation to claim their EW page. Self-serve onboarding. Restores the 2018 warm-lead pipeline without the structural mistake (sales-team outbound from review-derived contacts) that killed it.

Confirmed by Leanne as a real undersupplied revenue channel.

Killed by CEO

Post-Wedding Dashboard

Photo gallery (Pixieset partnership), AI wedding recap, anniversary planner, friend referrals, verified-bride badge. CEO call: "won't give as much value to the business or it's not really worth all the time and effort."

Architecture · locked decisions

Caller
Cloudflare Worker · single-tenant · EW-controlled
AI
Anthropic Claude API · Sonnet for reasoning · Haiku for parsing
SMS
Twilio · Alpha Sender ID = "EasyWeddings"
Email
Existing HubSpot · rebuilt templates
DB access
Admin REST API · NOT direct MySQL · IP allowlist
Auth
API key in header · service-account scoped
No new data store
All reads/writes flow through admin's existing data layer
No Make.com
Per established EW rule (codified 2026-05-05)

Stakeholders

Status colour: green = aligned & active, amber = waiting on reply, gray = not yet contacted.

CEO
Sponsor + approver
In-person approval received
Leanne
WedXM operational lead
Met 2026-05-06 · full ally · multi-supplier insight independently confirmed
Jeremy
Engineering — admin / API
API access ticket submitted via report-issue form
Ian
IT · Twilio Alpha Sender + Spam Act consent
Slack draft sent
Vitali
Engineering — review system
Will receive elasticity SQL + Trigger 3 sizing queries
Rebecca Crozier
Day 3 email content owner
Awaiting reply
Grace Pham
Day 12 email content owner
Awaiting reply

Current blockers

External dependencies that gate live operation. Build work continues against stubs in parallel.

Jeremy

API access ticket

5 endpoints + auth + rate limits. In his dev queue.

Twilio · AU Carriers

Alpha Sender ID approval

3-8 week lead time. ACMA mandate from 1 July 2026.

Matt / Ian

Spam Act consent confirmation

Does registration ToS cover SMS marketing under inferred consent?

Vitali

Trigger 3 population sizing

Decision gate: 500/yr threshold (minor) vs 2K+/yr (major channel).

Operational checklist with dependencies

While the API ticket sits in Jeremy's queue, a lot of work is shippable today — deliverability rehab, agent prompt design, Worker scaffold against stubs. The interactive checklist tracks state.

Open the To-Do →