Blog
Growth marketing: what it is and how to plan an experiment

Growth marketing is an approach to improving acquisition, activation and customer continuity through hypotheses, experiments and learning. The work might change an offer, a page or a welcome sequence, depending on where the problem lies. Start with a concrete decision: which part of the journey limits the outcome, and which change can we evaluate?
The term has no universal boundary with growth hacking or performance marketing. In his original July 2010 article, Sean Ellis describes prioritizing ideas, testing them and deciding which to keep. This is a practitioner’s perspective, not evidence that any experiment produces growth.
Choosing a first experiment
Suppose a newsletter receives 400 monthly visits to its signup form and gains 20 subscribers: 5%. These figures are hypothetical. The team thinks the form does not clearly explain what readers will receive. Before buying more traffic, it proposes clarifying the promise while keeping the rest of the page unchanged.
The hypothesis is: ‘Explaining the newsletter’s topic and frequency will help more visitors who want that content complete signup.’ It could be wrong: the issue might be interest in the topic, trust or a broken form. First check that signup works and talk to some readers. A copy test cannot fix a technical failure.
Prioritize this change based on its relationship to the observed problem, its cost and whether the result can be interpreted. A score using numbers chosen by the team may help structure a discussion; it does not independently estimate the chance of success.
What to define before starting
Choose a primary metric, such as confirmed signups divided by eligible visitors. Define a visitor, how to handle repeat visits and when confirmation is recorded. Keep a relevant guardrail too: form errors, unsubscribes or subsequent subscriber quality.
For an A/B test, assign variants randomly and persistently under comparable conditions, without changing the channel, offer and page at the same time. The NIST experimental handbook explains random assignment as part of a design for comparing a factor. Applying it to a website requires checking the implementation and ensuring a person does not switch variants with every visit.
Plan sample size, the effect you need to detect, acceptable uncertainty and duration before starting. There is no universal number of days or conversions. If traffic cannot answer the question in time, an interview or comprehension test may be a better next step, with different conclusions from an A/B test.
An example: 5% versus 6.5%
Imagine that over four weeks one variant gets 10 signups from 200 visitors and the other gets 13 from 200. Observed rates are 5% and 6.5%: a difference of 1.5 percentage points, equivalent to an observed relative increase of 30%.
That 30% should not be reported as a demonstrated improvement. There are only 23 signups, and the difference may reflect chance. An approximate normal interval for the difference, used here only as an illustration, runs from roughly −3.1 to +6.1 percentage points. With few events, use methods appropriate for proportions and the agreed analysis plan.
The team might retain the wording because interviews show it is clearer, while separating that editorial choice from a conversion-lift claim. To continue measurement, define a new test or a valid continuation rule. Looking every day until a favorable result appears changes the interpretation.
How this differs from publishing more content
A growth cycle connects an observation to a change, measurement and decision. It may conclude that activation or retention needs work before acquisition expands. Recording an inconclusive test also helps avoid repeating it under a new label.
For comparisons between organic posts, see testing social media content. For why acquisition may run ahead of evidence of value, read about product-market fit.
Plan your next post in HeyMark.
Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.