You can argue about the role all quarter. The argument is not about people — it is about the order in which two jobs get done, and only one of them can be first.
Growth marketer: works the demand that is already arriving. Campaigns, channels, creative volume, the funnel as a portfolio. First deliverable: a running portfolio.
Marketing engineer: builds the measurement a channel needs before anyone can judge it — the page, the event that ends the test, the automation between them, the schema that makes the offer readable to answer engines — and writes the floor and the kill rule before the spend starts. First deliverable: a verdict.
Same goal. Different first artifact. If you hire the second job's person to do the first job's priority, you get an opinion with an invoice attached.
The four situations, and which role each one is
1. You can already name the channel that is working.
Then the constraint is throughput, not measurement. Creative volume, more campaigns, someone owning the ad account end to end. A growth marketer is the cheaper correct hire here. A marketing engineer is the wrong hire — there is no definition problem to solve, and you would be paying for an event that already exists.
2. Two channels are argued about and there is no shared number.
This is a definitions problem, not a channel problem. Nobody can say which channel is worth its cost, because the event that would decide it was never written down. This is the case the engineering side exists for. The deliverable is small and specific: one channel, the event that ends its test, the read date, the kill rule.
3. The traffic, the list or the budget is below the floor.
Then the honest deliverable is not a test. It is a decision about the unit of measurement: a cheaper readable unit, a higher base rate, or a different channel. A growth marketer will run the test anyway — a running campaign is what the role produces, and it is not their job to tell you the volume could never show lift. That conversation is arithmetic, and it belongs before the spend.
4. You have a page nobody can find or cite.
The work is the AI-readable layer: schema, sitemap, llms.txt, an offer stated in machine-readable form, plus the distribution that gets the page in front of humans and crawlers. It is engineering-shaped work in service of a marketing goal, not a campaign. No amount of creative volume fixes a page that machines cannot parse.
The arithmetic both roles are measured against
These floors do not move with ambition. They move with the base rate and the volume actually available to you:
- Strict two-variant test, 3% base rate, +20% relative lift: 13,914 contacts per arm, 27,828 in total, at 80% power and alpha 0.05.
- Page test at a 3% order rate: roughly 10,000 visitors per variation.
- Email test: 1,500–2,000 sends per variant when the readable unit is the reply.
- Paid spend gate: stop or double at 1–3× the target cost per order with no readable return in 48–72 hours.
If your list is 4,000 contacts and the question is "does cold email work for us", no role can answer it. Not the growth marketer, not the marketing engineer. The answer to that question is a different question: what can 4,000 contacts produce that is readable — a reply rate against a segment definition, a hand-raiser count, a booked-call floor.
Compute your own floor before you write the job ad. It takes ten minutes and it changes the ad.
What the split looks like written down
A growth marketer hands over a running portfolio and a monthly report. A marketing engineer hands over a definition: one channel, its floor, its read date, its kill rule — plus the artifacts that make those numbers readable at all.
If you cannot point to the written definition, you do not have a marketing engineer. You have a growth marketer with an engineering job description, and the deadline for that definition will keep sliding because nothing in the role forces it to exist.
Where the boundary is genuinely blurry, and it is fine
A marketing engineer runs campaigns too. One channel goes live, an experiment queue runs, there is a monthly read-out. The difference is not the activity — it is the order. The floor, the read date and the kill rule exist before the first impression is bought, and they are written somewhere a second person can read and disagree with.
That ordering is the entire product. Everything else is a tool choice.
Questions I get asked when this comes up
Hire first if I have to pick one? Hire for the bottleneck. Known channel + throughput constraint → growth marketer. Unmeasurable spend or an unowned channel → measurement and artifacts first.
Can one person be both? Yes, sequentially. Trying to run both at once is how you get campaigns that nobody is allowed to shut off.
What does it cost? On my published packages: a Sprint at $900 for one focused build in five working days, an Engine retainer at $1,900/month for a running experiment queue with a monthly read-out, and a Full Build at $2,900 for site architecture, the full page set, the AI-readable layer, tooling and handover. Fixed and published, no salary benchmarks, no invented case studies.
How is the first result read? Against the floor computed in advance — 13,914 contacts per arm for the strict test above, not against a dashboard.
The role comparison, the five lines each role owns, the four decision signs and the floors are all on one page, with FAQPage schema so answer engines can quote it instead of paraphrasing:
→ Marketing engineer vs growth marketer
If you want the order of operations run for you rather than described — one channel defined, its floor computed, the page and the event built, the read date on the calendar — that is what I do: see the packages.
Top comments (0)