What the numbers looked like before. What we built.
Three recent engagements, told the way they happened: what the books said, what the source data said, what got built, and what changed. Business details are removed; the numbers are real.
DTC food manufacturer · Shopify + marketplace · ~100k shipments a year
The bakery that was told it lost money.
The problem
Four years of books said margins were 58% and 2025 ended in an $84K–$111K loss. The owner was making pricing, staffing and channel decisions off those numbers. None of them were right.
What the source data said
- COGS had been split equally across nine product lines for years, so product margins were meaningless. Actual gross margin was about 33%, not 58%.
- Inventory on the balance sheet covered finished goods only. Raw materials, packaging, shipping supplies and dry ice were expensed on purchase, overstating COGS and understating income.
- $200K–$350K a year of cost had been pushed to dry ice as a flat 20% share. $111K of card processing fees were never recorded. $1.4M of discounts sat in revenue instead of contra-revenue.
- Two products lost money on every unit shipped. 52 low-volume SKUs produced 7.8% of revenue. The marketplace channel alone generated roughly $544K of gross profit nobody could see.
What we built
- A 13-week diagnostic rebuilt from source data: 450,000+ Shopify orders, 56,000+ UPS shipments, 52,000+ marketplace orders, product costing for every line, and the full general ledger.
- A Financial Controls Framework and a locked chart of accounts: revenue by product type, COGS by category, processing fees as expenses, discounts as contra-revenue, customer money held as liabilities.
- A nine-account inventory system (raw materials through finished goods) with weekly COGS at standard cost, month-end true-ups to physical counts, and variance thresholds that decide when to recount.
- A Shopify-to-ledger revenue model built from payout exports so every deposit matches one balanced entry to the penny. Centralized bill pay with a weekly owner approval call and rolling cash projections through peak season.
Where it stands
The business was profitable. Weekly gross margin by product now exists, shipping moved from a loss to a consistent profit on cost-plus pricing, a supplier switch is saving ~$51K a year, and the owner runs the holiday season off a daily cash view instead of the line of credit. Ongoing fractional CFO engagement; accountant of record.
Legal services firm · two brands · $3.7M a year in paid lead generation
Five systems, no shared key, and a 60% leak nobody could see.
The problem
The firm bought demand efficiently (13,000+ signed clients at about $298 each) but could not tell which of those clients ever produced a fee. Phone, CRM, fee, accounting and ad platforms handed off overnight with nothing linking a call to a case to a dollar. Marketing decisions were made on vendor reports.
What the source data said
- At one brand, about 60% of clients signed over three years were withdrawn or closed with no fee, half within seven months. At the sister brand the figure was 4%. The leak was not in targeting; it was in what happened after signing.
- One supplier carried 45% of all paid lead spend and billed above its signed contract on both brands. No contract committed volume and no renewal date was on file.
- The firm’s own search channel cost 1.6–1.8x as much per signed client as the call vendors, and roughly a tenth of its media was buying searches for the government agency itself. Spend had been cut from $1.69M to $0.64M with no decision record behind the cuts.
- A later performance collapse ($104K to $37K a month, impressions down 75%) was blamed on platform policy. The change log showed four unauthorized Target CPA cuts. Fees lag signing by 13–21 months, so none of this was visible in revenue yet.
What we built
- A data bridge joining the five systems outside the firm’s own tools, with a defined "signed" and "retained" that tied fee receipts to the books within a fraction of a percent.
- Cost per signed client by channel and by billing line, cohort retention by signing year, and a fee-lag model so the first year of any plan is judged on signed clients and retention, not fees.
- A 57-page diagnostic deck and long-form findings document, built with a marketing-strategy partner, that quantified about $160K of pipeline at risk and led directly to a signed engagement.
Where it stands
Phase I accepted; a multi-month reporting-layer build and ongoing CMO-level retainer followed, with Workflow Seams owning data and systems.
Delivered in partnership with Ballyhoo, a marketing strategy agency.
Video production company · ~$1M revenue · owner-led
From gut feel to a biweekly numbers meeting.
The problem
A growing production company with strong work and no finance function. Projects were priced and staffed from experience, bonuses and distributions were decided on cash in the bank, and the owner carried the whole picture in his head.
What the source data said
- Revenue up about $200K year over year with no structured way to see which project types, clients or divisions produced the margin.
- Tax reserves, bonus pools and owner distributions competing for the same cash with no rule for which came first.
- Big decisions (a retirement plan, a second vehicle, a minority stake in a rental partner, a three-division rebrand) arriving faster than the numbers to evaluate them.
What we built
- An operations system in Notion the team actually runs on: projects with income, wages, expenses and margin per job, gear rental transactions with rate and discount tracking, and a transaction database tied to finance parameters.
- A standing biweekly CFO meeting: revenue and AR pipeline, quarterly projections, bonus and reserve decisions made against a rule, tax reserve funded first.
- Decision memos on the real questions: a Roth 401(k) deferred when the $20K–$30K match cost outweighed the owner benefit; an acquisition stake evaluated on rental economics; division pricing tiers reassessed against past wins and losses.
Where it stands
On track to pass $1M for the first time with a tax reserve surplus and a funded bonus pool, and an owner who makes capital decisions in a meeting with numbers instead of alone with a bank balance. Ongoing fractional CFO engagement.
Want to know what your source data says?
Every one of these started with a 15-minute call and a question the owner couldn't answer from their own reports.