Fiscal year 2026 · Seventh annual edition
Growth that holds up under audit.
Jordan Green runs growth marketing and AI operations for DTC and B2B teams. This report states seven years of that work the way results deserve to be read: line by line, at cost, with the workings shown.
Attributable impact across the five line items†
Leads scored per month by pipelines currently in production
Automations running unattended, monitored, and documented
Consecutive years without a missed reporting period
† Twelve-month attributable impact, client-reported and holdout-checked where noted. Basis of preparation in Note 1.
Marketing has a reputation for creative accounting. I’d rather do creative work, and account for it.
Most portfolios are highlight reels. This one is a filing. Every engagement on the following statement carries an account code, a stated impact, and notes on exactly how the result was produced — the campaign you could put on a billboard and the pipeline you could put in a code review. I’ve spent seven years insisting those are the same discipline.
The practice splits roughly down the middle. One half is classical growth marketing: positioning, lifecycle, paid creative, the patient work of finding words that make strangers care. The other half is machinery — n8n flows, Claude-powered scoring, warehouses and attribution models — built so the creative work compounds instead of evaporating at the end of each quarter.
Open the line items below. Each one expands into the full working papers. When you’ve reviewed all five, the books balance — a small ceremony I’ve found more honest than a testimonial carousel.
Five engagements, stated at realised impact. Select any line item to open its working papers — the story on the left, the auditor’s notes on the right. Reviewed lines receive a tick; the total accrues below.
An eight-figure skincare brand was acquiring customers beautifully and then going quiet on them — one welcome email, a monthly newsletter, and a prayer. We rebuilt the entire lifecycle around how people actually use the product: replenishment flows timed to each SKU’s usage cycle, a post-purchase education series, and win-backs written in a voice mined from 2,300 five-star reviews.
Sixty-one emails shipped over eight months. Repeat purchase rate moved from 21% to 29% — a 38% relative lift, measured against a 10% holdout that never saw the new flows. Finance signed the number, which remains my favourite review.
A B2B SaaS sales team was hand-triaging every inbound lead and losing the good ones in the pile. I built a scoring pipeline that reads each lead the way their best SDR would: enrichment data, intent signals, and — the part rules-based scoring always fumbles — the free-text “what are you trying to solve?” answer, graded by Claude against a written ICP rubric.
4,200 leads a month now arrive scored 0–100 with a two-line brief for the rep, routed in under four minutes at roughly two cents a lead. Same traffic, same team: 22% more SQLs, because the right leads finally got called first.
Ninety posts a quarter sounds like a content farm. It isn’t — that’s the point of the design. AI drafts from briefs generated out of SERP gap analysis; two human editors own the point of view, the claims, and the final cut. Nothing ships unread, and anything the editors wouldn’t put their name on dies in the queue. About a third does.
Eighteen months in, organic traffic sat at 3.1× baseline and — the number I watch more closely — organic-sourced pipeline grew faster than traffic did. A decay model flags aging pieces for refresh before rankings slip, so the library compounds instead of rotting.
Paid creative was being produced the expensive way: big swings, long waits, opinions settling arguments that data should have. I replaced it with a modular framework — hooks × angles × formats, every asset named so its lineage is machine-readable — and a weekly kill/scale readout the whole team could challenge.
214 variants over two quarters. Losers died within 72 hours by automated rule, never by meeting. Winning DNA got recombined instead of retired. Blended CAC fell 27% while spend grew — the order those two things happened in is the entire trick.
The marketing team answered “what’s working?” with six dashboards that disagreed with each other, politely. I piped all six channels into one warehouse, enforced UTM governance with a linter (yes, really), and built a single attribution model with its assumptions written at the top of the file where everyone can argue with them.
One model everyone argues with beats six models nobody trusts. Monthly reporting went from forty hours of spreadsheet archaeology to six, and budget meetings now start from the same number — which turned out to be the actual deliverable.
of $7,540,000 stated — open each line to review
† Impact figures are twelve-month attributable values agreed with each client’s finance team, converted to USD. Notes n.2 and n.3 describe the creative and technical accounts respectively. The preparer rounds down on principle.
Basis of preparation
Engagements are stated on an outcomes basis. Work begins with a two-week discovery, ships weekly thereafter, and is measured against holdouts or pre-agreed baselines — never against “vibes at the QBR.” Where AI performs a task, a named human owns the output. No figure appears in this report that the client’s finance team hasn’t seen first.
Creative accounts
| 2.1 | Positioning & messaging | Narrative, category framing, the words on the homepage |
| 2.2 | Lifecycle & retention | Email/SMS programmes, replenishment logic, win-backs |
| 2.3 | Paid creative systems | Modular testing frameworks, hooks, UGC direction |
| 2.4 | Brand voice | Style systems mined from customers, enforced in prompts |
| 2.5 | Landing & CRO | Page architecture, offer testing, plain-English copy |
Technical accounts
| 3.1 | Orchestration | n8n — 22 production flows, alerting, run-books |
| 3.2 | Applied AI | Claude API: scoring, drafting, enrichment, evals |
| 3.3 | Data & warehouse | BigQuery, dbt, Fivetran; SQL written by hand |
| 3.4 | Attribution & analytics | GA4, Segment, position-based models in version control |
| 3.5 | Platform APIs | Meta, Google, Klaviyo, HubSpot — automated rules & sync |
Related-party disclosures
The preparer is also the subject of this report, a conflict of interest managed through holdout groups, versioned assumptions, and a stubborn preference for numbers a CFO would initial. Readers seeking an independent view are directed to Section 04.
Opinion: unqualified.
The reviewers found the results fairly stated, in all material respects.
In the reviewers’ judgment, the statements herein present fairly the growth practice of the preparer. Their remarks, reproduced without adjustment:
Jordan is the only marketer I’ve worked with who could present to the board and then go fix the dbt model the numbers came from. Both, in the same afternoon.
The lead scoring pipeline paid for itself in the first month. What I didn’t expect was the documentation — my team could run it the day the engagement ended.
We hired a growth marketer and got a data team, an automation engineer, and the best copywriter in the building. The invoice said one person. I checked.
The next fiscal year is unwritten. Let’s state it well.
I take on a small number of engagements a year — growth marketing with the machinery to prove it worked. If your numbers deserve better bookkeeping, the ledger is open.
