You hired a growth marketer and got a campaign manager with a new title. The loop that separates growth from marketing is a weekly cadence with one metric and one kill decision. So why does your team have neither?
The kickoff deck had a loop drawn on it in three colors. A growth lead with a real track record had built it, the budget behind it finally looked serious, and for about six weeks it ran.
Four months later the same channels are running, the weekly growth meeting is a status update, and nothing on that slide ever had an exit. Every experiment in it was designed to be read, not decided. Nobody had written what result would stop the work, and nobody had the authority to say stop.
What Makes Growth Marketing Different (One Metric, Not a Remit)
The definition of growth marketing is everywhere. A generated answer writes it out, every vendor guide opens with it, and it is the same paragraph each time. Reading it will not help you run one.
What the guides do after the definition is more interesting. Most hand you a list: eight steps, four pillars, a catalog of strategies. A list gets worked through once and cannot be run every week. The few that do describe a loop stop at analyzing the result, and none says what happens to the experiment afterwards.
The useful distinction is operational. Growth works in experiments tied to a single number. Marketing works in channels tied to a plan. One gets judged by whether the number moved this week, the other by whether the plan was delivered this quarter. A demand generation strategy reports at the end of the quarter. A growth loop moves one number by Friday.
Testing is already normal. Ascend2 surveyed 402 marketers in 2025, all of them already running A/B tests, and 84 percent of that group test at least monthly. The constraint sits after the result comes in, not before it.
If a growth role has no single number attached to it, treat it as a remit rather than a function. A role that owns acquisition, retention, brand and lifecycle gets measured on activity.
The Weekly Loop: Hypothesis, Ship, Kill
A loop has four beats and a fixed clock, and the clock is what makes it cheaper than the alternative. That schedule is the part of a growth marketing process most teams skip: the calendar forces a decision, not the analysis.
Hypothesis comes first, written before the change ships: what we expect, and what result would prove us wrong. Ship means a change that reaches users inside the week, not a project that needs a release train. Read compares the result against the threshold on a date set in advance. Kill is the decision, and it happens whether the result was good or bad.
Wingify's State of Experimentation 2026, built on a survey of 108 organizations across 21 industries, reports that 56 percent of programs are trapped in an execution bottleneck. Those teams are not short of ideas. They are short of throughput past a fixed point.
A threshold does not have to be statistical. Activation moving the right way by Friday is enough to decide on, and so is a checkout step abandoned less often than last week. What makes either one a threshold is that it was written before the result arrived and someone other than its author can check it. Adopt one rule now: nobody revises the number after the experiment ships.
Here is the clock.
Week | Decision due | Owner |
|---|---|---|
0 | The one metric this cycle is judged on | whoever reports the number |
1 | The experiment enters the queue with its kill threshold written down | experiment owner |
2 | The change ships by Friday, or it gets dropped | whoever can put it live |
3 | Kill, keep or scale, against the threshold | the person with kill authority |
4 | The result goes into the record and the queue refills | ops or analyst |
The clock matters more than the rigor. A rough threshold you check on Friday beats a careful study that answers after the decision was made by default, and at the volume most teams run a clean statistical read is not available inside a week anyway. So the threshold stays directional: the number moved, it did not move, or it moved the wrong way. Friday produces one of those answers whether or not the test was big enough.
Where loops rot is at the fourth beat. Kill feels like admitting the hypothesis was wrong, so teams invent a reason to keep going: the traffic was unusual, the quarter was odd. None of it changes the result.
Sizing the Team You Can Actually Staff
You do not staff a growth loop by headcount. You staff it by decisions per week.
Three roles have to exist. Someone owns the number and can say what it means. Someone can ship a change into the live product. Someone can kill an experiment, including one they sponsored. On a small team one person can hold two. Nobody can hold all three.
The shipper is usually the role that is missing. If every change waits on an engineering sprint or an agency ticket, your constraint is access, not budget. Hiring a growth marketer into that setup buys you a person who cannot ship.
Teams of five hit a different wall: enough people to ship, and not enough distance between the person proposing an experiment and the person judging it. Name the judge from outside the loop, someone who does not own the outcome.
The calendar and handoff mechanics are an operations problem, written down in our guide to marketing operations. What the loop needs is different: the queue that fills those slots, and the decision attached to each one.
A workable starting size is one or two experiments in flight for a team of two. Reading and killing take longer than shipping, and a queue of eight is where nothing gets decided.
Choosing the One Metric Without Lying to Yourself
The metric is where growth programs lie to themselves, and rarely dramatically. It is a number that could not have moved inside the cycle, so every experiment ends as "inconclusive" and the team learns to shrug at the word.
Three tests, in order. Can this number move inside a week? Can it move because of something the team did, rather than because a quarter ended? Can one person act on it on Monday?
Lagging revenue fails the first test. Sessions and impressions fail the second, because they move for reasons the team does not control. A composite score fails the third, because nobody can say which part of it moved. What passes is narrower and less glamorous: activation on the first meaningful action in the product, completion of one step in a signup path, reply rate on one outbound sequence. All of it can move by Friday.
Pick one, then keep it for a full cycle. Changing the metric the week a test loses turns the scoreboard into a moving target. How those numbers get set and reported at team level is a discipline of its own, and the mechanics are in our guide to marketing KPIs.
The number also has to survive the quarter. Ask what you would stop doing if it doubled. If the answer is nothing, it is not a growth metric. It is a stat.
The Experiment Queue and How Items Die
An experiment queue puts an expiry date on every item. That single property separates it from a backlog, which is just a list of things people want. An automated marketing workflow can ship the change. It cannot decide whether the change was worth keeping.
Every item enters with five fields: the change, the metric it should move, the threshold that counts as a win, the expiry date, and the name of whoever decides. If any is blank, the item stays out.
Intake is where most queues lose their shape. Anyone can propose, but not everyone can add. A proposal becomes an item when its fields are filled and someone with capacity accepts it.
Items die in three ways. They hit the threshold and get kept or scaled. They reach the expiry date without a readable result and get killed anyway. Or the thing being tested stops mattering. All three endings are normal.
The base rate is the part almost nobody quotes. Analytics-Toolkit published a meta-analysis in 2022, now several years old, covering 1,001 A/B tests: 33.5 percent produced a statistically significant positive result, and the median lift across all tests was 0.08 percent. Set those against your own queue. Two items out of three will not win, and the typical item moves almost nothing, so a queue that keeps every inconclusive test alive spends its attention on the two-thirds that did not work.
What separates programs that keep going is less about budget and more about records. Speero's 2025 maturity research, built on more than 150 experimentation teams, found that 87 percent of mature programs standardize their process and test records. That comes to me through a secondary write-up rather than the report, and the growth claim attached to it is a correlation, not a cause.
Five lines are enough: what changed, what we expected, what happened, what we will do differently, who signed it. Ours sit in one project file, findable six months later. It is also the part of a data-driven marketing strategy that gets skipped, where the data is collected and the decision after it is not.
When Growth Marketing Is the Wrong Structure
Pages that ask whether growth marketing suits your business often mean whether you can handle growth: sales capacity, fulfillment, hiring. Real questions, and not this one. The question here is whether the loop can exist at all. Three tests, and they are structural rather than motivational. A product led growth strategy fails on different questions entirely, which is why borrowing another motion's checklist settles nothing.
Can one metric move inside a week? If the only number that matters is closed revenue on a nine-month sales cycle, the loop's clock is nine months. That is a planning cycle with a weekly meeting attached, and no amount of enthusiasm changes the clock.
Can a change reach users in under a week? If every change waits for a release train, a legal review or a migration, the clock is longer than the loop. The work still counts as marketing, but it cannot be run as an experiment queue.
Does one named person hold the kill decision, including the right to kill something they sponsored? If killing requires consensus, nothing dies.
Two yeses and one no is still a no. The loop breaks at its slowest step, and that is usually the third one, because authority is harder to hand over than access. The three tests also explain a pattern I have seen more than once: a company passes the first two, fails the third, and concludes that growth marketing does not work in its category.
When the answer is no, hiring a growth marketer makes it worse. You get a person whose job is to run a loop that cannot close, and the first thing they learn is to report activity instead of decisions. If the leak sits before anyone sees a number worth moving, that is a lead generation funnels problem.
Frequently Asked Questions
What is growth marketing in one sentence?
Growth marketing is a weekly loop that ties one metric to experiments that ship fast and get killed on a threshold written before they start. It needs a number owner, a shipper, and someone with the authority to stop the work.
How is growth marketing different from demand generation?
Demand generation builds and fills a funnel, and reports across a quarter. Growth moves one number inside that funnel this week. Teams need both, run by different people with different clocks.
How many experiments should a small team run at once?
Only as many as you can read and kill inside a week. For a team of two that is one or two in flight, because reading and deciding take longer than building.
Run the Loop Before You Hire for It
Start with four weeks and one metric, and let the first kill decision prove the loop is real. Allable keeps the queue, the experiment record and that decision in one workspace, so the loop has somewhere to live before it has a team.
