Real work. Real results.

Reclaim Lost Revenue · Equip Layer · Atlanta, GA

The Operator

BriLaSha Beauty Bar is a beauty studio in Atlanta specializing in wood therapy body sculpting, permanent makeup, body contouring, and massage. The owner, Brittany, runs the business with online booking via Acuity (connected to Stripe), an Instagram presence, and a Mailchimp subscriber list.

The Problem

Like most beauty businesses, BriLaSha had a growing list of lapsed clients — people who had booked before but hadn't returned in months. The Instagram DMs had unanswered inquiries. The subscriber list was uncontacted. Revenue was walking out the back door.

What We Built

DYOE Way ran a full Reclaim Lost Revenue engagement using our Claude-powered campaign stack:

• Email segmentation — Divided the Mailchimp list into four cohorts: Warm (90–180 days since last visit), Mid (181–365 days), Cold (366+ days), and New Subscribers
• Claude-written campaigns — Two rounds of personalized outreach with segment-specific subject lines and copy (e.g., "We miss your face" for warm clients, "Missing you at the bar" for cold clients)
• Domain authentication — Set up Mailchimp domain auth for deliverability
• ManyChat DM automation — Configured an AI assistant ("Alaina") on Instagram/Facebook to handle booking inquiries, quote services, and route prospects to the scheduler
• Promo code system — Created Stripe/Acuity promo codes (WELCOMEBACK, WELCOMEBACKBROWS, NEWBRI10) tied to specific offers to track conversion from campaign to booking
• Website build — Deployed a professional main website on Netlify with service breakdown, pricing, and booking integration

Results — Round 1

Segment	Recipients	Open Rate
Warm (90–180 days)	19	22%
Mid (181–365 days)	63	30%
Cold (366+ days)	71	18%
New Subscribers	37	33%

190 total recipients across 4 campaigns. Open rates ranged from 18% (cold — expected) to 33% (new subscribers). The Mid segment at 30% was the strongest signal: people who hadn't visited in 6–12 months were the most reachable and the most likely to rebook.

Claude's Role

Every campaign subject line and body was written by Claude, reviewed by DYOE Way, and approved by Brittany before send. The segmentation logic was designed in Claude Code. The ManyChat AI assistant prompt was authored and tuned in Claude. The website was built and deployed from Claude Code sessions.

What We Learned

• Mid-range lapsed clients (6–12 months) are the highest-ROI reactivation segment for beauty businesses
• Cold clients (1–2 years) still open at 18% — worth reaching, but don't over-invest
• Promo codes on reactivation campaigns need to be segment-specific; a blanket 10% off doesn't motivate a client who spent $275 last time
• DM automation must sound human — "Girl, where have you been?!" outperformed every formal subject line

Tools Used

Claude API (Sonnet) · Claude Code · Mailchimp · ManyChat · Stripe · Acuity · Netlify

Product Build · Equip Layer · Pre-Launch Beta 2026

The Problem

Self-employed service providers — stylists, tax preparers, freelancers, anyone filing a Schedule C — do not have a finance tool designed for them. They use spreadsheets, shoeboxes, or nothing. When tax season hits, they scramble to reconstruct 12 months of transactions from bank statements, Cash App exports, and receipts stuffed in a drawer.

The result: missed deductions, wrong line-item classifications, and money left on the table. Every year.

What SplitLedger Does

SplitLedger is an AI finance co-pilot that parses bank statements, cross-references payment platform exports, classifies transactions to Schedule C line items, and surfaces missed deductions — all powered by Claude edge functions on Supabase.

• Statement parsing — Upload a bank statement (PDF or CSV); Claude's parse-statement edge function extracts and normalizes every transaction
• Cross-source reconciliation — Merge Cash App + Shopify + bank deposits; auto-dedupe self-transfers between accounts
• Related-party flagging — Auto-detect payments to/from family or shared names; prompt user to classify as business/personal/rent
• Expense line-item mapping — Match merchants to Schedule C lines (e.g., "Shopify subscription" → Line 25/27a, not Line 10)
• Missing-deduction prompts — Decision tree: "You have a car → did you track miles? You rent → related-party rent? Event receipts → booth fees?"
• Budget advisor — Claude-powered insights on spending patterns, savings goals, and cash flow

Architecture

Component	Technology
Frontend	React + TypeScript
Backend	Supabase (PostgreSQL + Edge Functions)
AI Layer	Claude API (Sonnet) via Supabase Edge Functions
Edge Functions	parse-statement, budget-advisor, dashboard-insights
Auth	Supabase Auth (Google/Apple OAuth planned for launch)
Source	github.com/bjyoe2016/Splitledger-

Feature Backlog — Derived from Real Tax Cases

The 10-item feature backlog was generated from a real tax filing session — a first-time Schedule C filer with income from Cash App, Shopify, and a bank account, plus related-party transactions with family. Every friction point we hit became a SplitLedger feature:

• Cross-source income reconciliation
• Related-party classification prompts
• Expense line-item mapping with validation
• Missing-deduction decision tree
• 1099 reconciliation dashboard
• Accountant handoff mode (one-click clean export)
• SSA credit awareness ("Filing earns you Social Security quarters")
• Receipt capture nudges during the year
• Related-party rental income cross-return link
• De minimis safe harbor election auto-generator

Status

App runs locally. All 6 database migrations applied. All 3 edge functions deployed. Next: end-to-end statement upload test, then web deploy, then public beta.

Tools Used

Claude API (Sonnet) · Claude Code · Supabase · React · TypeScript · GitHub

A Note on What You See Here

We only publish case studies with the operator's knowledge and consent. We only include numbers that actually happened. If an outcome is projected or in-progress, we say so. If a venture is pre-launch, the status says "beta." This page will grow as we ship more work and as clients consent to being featured.