Good marketing looks like magic. Mine looks like a plan you can run.
I'm Jordan Green. I design growth plays — lifecycle, paid, content — and build the AI machinery that executes them week after week. This wall holds five of those plays. Press Run the play on any board and watch the plan become the number.
Repeat purchase rate stuck at 19% for three quarters.
Test offer depth
flows rebuilt →
every board below actually runs — try one
Five plays. Receipts attached.
Each board is a real engagement pinned the way I actually plan: the brief, the moves, and the machinery. The results are already earned — you just get to run the play and watch them land.
pins drag on desktop, if you like a messier wallLifecycle email overhaul
DTC skincare brand · 8 months · retention
Animates the plan executing. All results are readable below regardless.
Repeat purchase rate stuck at 19% for three straight quarters.
Email was one weekly blast to 140,000 subscribers — the same message for a first-time buyer and a five-jar loyalist.
Map the lifecycle
Cohort all 140k customers by purchase cadence and product type. Find where people actually fall off.
9 segments. The day-40 drop-off was the whole story.
Rebuild the flows
Welcome, replenishment, winback — triggered off real behavior, not the calendar.
11 flows, 42 emails, all live by month three.
Time the replenishment
Send when the jar runs low, not when the send calendar says so. Cadence modeled per product.
Day-42 trigger for the hero SKU. Open rates doubled.
Test offer depth
Hold discounts back entirely for two segments and see who actually needs them.
They barely noticed. Margin stayed home.
19% → 26.2% of customers buying again inside 90 days.
CountedUp from 9%. The blast calendar got quieter, not louder.
CountedAll of it earned on timing and relevance, none of it bought.
Counted- Klaviyo flows×11
- dbt cohort modeldaily
- Segment eventswired
- GA4 holdout groups2
AI lead-scoring pipeline
B2B SaaS · n8n + Claude · 6 weeks to live
4,200 inbound leads a month. Three SDRs cherry-picking by company name.
Half the pipeline was luck. Good leads sat for hours while reps worked whoever looked familiar.
Write the rubric with sales
Scoring criteria drafted in a room with the reps, not handed down from marketing.
Sales signed it. That's what made the rest stick.
Build the n8n intake
Every lead enriched, deduped, and queued for scoring inside a minute of hitting the form.
60-second SLA, held through a 3× traffic spike.
Claude scores — and explains
0–100 against the rubric, plus a two-line reason a rep can actually read before dialing.
The reasons are why reps trust it. Nobody trusts a bare number.
Route with receipts
Hot leads hit Slack with the score, the why, and a one-tap claim. Everything logs to HubSpot.
Auto-routed since week six. Zero spreadsheets survived.
- n8n workflows×7
- Claude APIscoring
- Clearbit enrichinline
- HubSpot + Slackrouted
Reps work the right leads first instead of the familiar ones.
CountedMedian first touch on hot leads. Speed is the whole game.
CountedCheaper than a rep's first sip of coffee, and it never sleeps.
CountedThe content engine
B2B fintech · human-in-the-loop AI · 9 months
Ninety posts a quarter, without sounding like a content farm.
The founder wanted real thought leadership at real volume with a team of two writers. Pure-AI output had been tried — it read like everyone else's.
Extract the voice
A voice guide distilled from their 40 best-performing posts — cadence, stance, banned phrases.
Readers stopped guessing which posts were assisted.
Draft with AI, gate with humans
Claude drafts against the voice guide. Nothing publishes without an editor's hands on it.
100% human edit rate, forever. Non-negotiable.
One idea, five formats
Every pillar piece splits into channel-native cuts — post, thread, newsletter section, two shorts scripts.
Research once, publish five times. The math that makes 90 possible.
The weekly kill list
Search Console and social data reviewed every Friday. Formats that don't move get retired.
Volume without the kill list is just noise. We killed plenty.
Shipped, on voice, every quarter since launch.
CountedIn nine months, with zero paid distribution behind it.
CountedSame two people. The engine does the heavy lifting.
Counted- Claude projectsvoice-tuned
- Airtable pipelineidea→live
- Human edit gate100%
- Search Console loopweekly
Paid creative testing framework
E-commerce apparel · Meta + TikTok · 90 days
CAC climbing 8% a quarter. Bids were fine — creative was stale.
Six ads a week, all variations of the same tired angle, all judged on vibes in a Slack thread.
Map twelve angles
Every believable reason to buy, ranked by evidence from reviews, support tickets, and comment sections.
The winning angle was #9. Nobody had ever tested it.
Modularize the ad
Hooks × proof blocks × CTAs as swappable parts, so one shoot yields dozens of variants.
One shoot day → 40+ variants. Editors loved it. Eventually.
Kill/scale rules, pre-agreed
Spend and CPA thresholds set in advance. Every Monday the rules decide, not the room.
Vibes lost their seat at the table. CAC noticed.
Feed winners back
Every winning variant's hook and angle seeds the next batch. The system compounds.
Three evergreen winners still running six months later.
- Meta + TikTok2 ch
- Naming conventionenforced
- Auto-reportingLooker
- Weekly kill/scaleritual
In 90 days — up from six a week judged on vibes.
CountedSame budget, same bids. Creative was the lever all along.
CountedCreative velocity ×4 with the same team and shoot budget.
CountedMarketing ops warehouse + attribution
Multi-channel retailer · 6 channels · 4 months
Six channels, six dashboards, six versions of the truth.
Monday meetings were archaeology — an hour arguing about whose numbers were right before anyone could ask what to do about them.
One warehouse, one schema
Every channel's spend and revenue lands in BigQuery, named the same way, on the same clock.
Six pipelines live in month one. Naming fights: settled.
Model spend → revenue
dbt models trace every dollar out to every dollar back, at the campaign level.
Tested, documented, and boring — exactly as models should be.
Blended + incrementality views
Last-click for speed, blended for honesty, holdouts for the arguments money starts.
Branded search was 30% less heroic than it claimed. Reallocated.
The Monday ritual
One dashboard opens the weekly meeting. Decisions in the first ten minutes, not the last five.
The archaeology hour became a ten-minute read-out.
Every team argues from the same numbers now. Arguments got better.
CountedMostly branded-search over-credit and one channel double-counting.
CountedNot quarters. Confidence is the real deliverable here.
Counted- BigQuerywarehouse
- Fivetran6 pipes
- dbt modelstested
- Looker Studio1 board
How I run plays
Diagnose
Find the number that's wrong and the reason it's wrong. No play survives a bad diagnosis, so this stage gets the most stubbornness.
Design the play
One page: the lever, the moves, what we expect to happen, and what we'll kill if it doesn't. If it needs a deck, it isn't ready.
Build the machine
Automations, prompts, pipelines, dashboards — so the play runs every week without heroics, and keeps running after I leave.
Count it
Results get stamped when finance would agree with them. Vanity metrics don't make the wall — that's what keeps the wall honest.
Half strategist, half machinist.
I've spent eight years running growth for DTC and B2B teams — first as the person writing the emails, now as the person building the systems that write back. I live in the overlap between creative direction and ops engineering: comfortable arguing about a subject line at ten and debugging a webhook at eleven.
The war-room format is genuinely how I work. Every engagement starts as a board — brief, moves, expected numbers — and ends with receipts stamped on. If a play doesn't earn its ticket, it comes down off the wall. That rule has cost me some flattering case studies and kept every one you just scrolled past honest.
I take on two or three plays at a time, hands-on, from diagnosis to the dashboard that proves it worked.
Got a number that's stuck?
Tell me what it is and what it should be. I'll bring the board, the moves, and the machinery — and we'll run the play together.