Growth marketing: an experiment operating system

Growth marketing is often sold as a collection of clever tricks. In practice, the teams that grow steadily are not the ones with the best tricks but the ones with the best routine: they decide what to test, run tests cleanly, record what happened and act on it, every week. This guide describes that routine as an operating system you can install in a marketing team of any size, with the roles, rituals and documents it needs.

What growth marketing actually is

Growth marketing applies experimentation to the whole customer journey, not just acquisition. Instead of asking "which channel should we spend more on", it asks "where in the journey are we losing the most value, and what is the smallest change that could fix it?" That may be an ad, but it may also be onboarding, pricing presentation, an email sequence or a sales hand-off. The method matters more than the channel.

Three ingredients make it work: a single metric that represents value for the business, a steady flow of testable ideas, and a measurement setup that can tell a real effect from noise.

Component 1: one north-star metric and its inputs

Pick one metric that best represents value delivered to customers and captured by the business, for example weekly active accounts, qualified pipeline or repeat orders. Then break it into the inputs you can influence: traffic, activation rate, conversion rate, retention, order value. Every experiment should target one input and state how it is expected to move the north-star metric.

This sounds bureaucratic, but it prevents the most common failure: a test that "wins" on clicks while the business metric does not move.

Component 2: a ranked experiment backlog

Keep all ideas in one backlog, each written as a hypothesis: "Because we observed X, we believe changing Y for audience Z will improve input metric M." Score each idea on expected impact, confidence and effort, and pull from the top. Simple scoring is enough; the point is to make trade-offs visible and stop the loudest opinion from winning by default.

Good sources of ideas include analytics funnels, session recordings, sales call notes, support tickets and customer interviews. Ideas copied from other companies' case studies are the weakest, because their context is not yours.

Component 3: tests that can actually be read

A test is only useful if its result can be trusted. Before launch, write down:

• The primary metric and the minimum change that would matter commercially.
• The audience and how traffic is split. In Google Ads, for example, custom experiments share a campaign's traffic and budget, and a cookie-based split keeps each user in either the original or the experiment, which reduces contamination.
• The duration, fixed in advance and long enough to cover full weekly cycles.
• The decision rule: what result means ship, what means stop, what means iterate.

Clean event tracking underpins all of this. Google Analytics 4 provides recommended events for common actions such as sign-ups, lead generation and purchases; using them consistently means every test reads from the same definitions. Our analytics and data work often starts by fixing these definitions before any test runs.

Not everything can be A/B tested. Low-traffic pages, brand campaigns and pricing changes may need before-and-after comparisons, holdout regions or qualitative research. Label those results as weaker evidence rather than pretending they are controlled tests.

Component 4: a weekly growth meeting

The operating rhythm is one short meeting a week with a fixed agenda:

• Review the north-star metric and its inputs against last week.
• Review tests that finished: result, decision, and what was learned.
• Check tests in flight for problems, not results; peeking and stopping early produces false winners.
• Pull the next tests from the backlog and assign owners.

Record every decision in a shared log. Over months, the log becomes the most valuable asset: a record of what works for your audience that no competitor can copy.

Growth marketing agency or in-house team?

An in-house team owns context and speed; an agency brings tooling, specialist skills and pattern recognition from other accounts. Many companies combine the two: the internal team owns the north-star metric and the backlog, while a partner supports measurement, testing infrastructure or specific channels. If you hire a growth marketing agency, ask to see its experiment log format and decision rules, not just its case studies, and agree up front who owns the data and the learnings.

Our approach follows this operating model. We do not promise a growth rate; we set up the metric, the backlog and the measurement so that every week produces a decision you can defend.

Sources

• Set up a custom experiment — Google Ads Help
• Recommended events — Google Analytics Help