The teams we work with are usually busy running what already works — the channels that got them here, the campaigns that hit last quarter's number. That work is necessary and it consumes everyone. What it doesn't do is answer the question of what comes next.
The difference in practice
A marketing agency takes a channel and executes against it. More ads, more content, more email. The output is activity, measured in deliverables.
Growth engineering starts somewhere else — with a hypothesis about where the leverage actually is, and then builds the mechanism that exploits it. The output is a system the company owns.
If the engagement ends and the growth stops, it wasn't engineering. It was labor.
What an experiment actually requires
A growth experiment isn't "let's try LinkedIn ads." It has four parts:
- A hypothesis — a specific claim about what will move a specific metric
- A success threshold — decided before the test runs, not after
- A kill criterion — the condition under which you stop
- A timebox — fixed, so the test can't quietly become a program
Miss any one of those and you don't have an experiment. You have an initiative, and initiatives don't produce learnings — they produce sunk cost.
Why this matters for what we build
Framing the work as engineering changes what you're willing to ship. You stop optimizing the funnel by a few percent and start asking which mechanism doesn't exist yet.
That's a harder question. It's also the only one whose answer compounds.