# A Growth Experiment Template > A growth experiment template for SaaS teams with hypothesis, metrics, design, duration, results and decision fields, plus tips for running experiments well. Source: https://comppound.com/blog/a-growth-experiment-template (published 2026-07-28) Growth experiments are how SaaS teams learn what actually drives acquisition, activation, retention and revenue. Each experiment is a structured attempt to test an idea: change one thing, measure what happens and decide what to do next. Run consistently, experiments build a body of knowledge about your customers and product that no amount of opinion or best-practice reading can replace. Without structure, though, experiments quickly degrade into random changes with unclear results. Someone tweaks a headline, traffic happens to rise that week, and the team declares victory without knowing whether the headline had anything to do with it. Another person tests a new pricing page but forgets what they were trying to learn. Six months later, a new team member proposes the same experiment that already failed, because nobody wrote down the result. A simple experiment template fixes most of these problems. It forces clear thinking before anything launches, keeps documentation consistent, helps the whole team learn from every test and prevents repeated experiments. This article presents a template, explains each part and walks through an example. ## Why a template helps The first benefit is clearer thinking. Filling in a template before launching an experiment requires you to state what you observed, what you believe will happen and how you will know. Many weak experiment ideas fall apart at this stage, which saves the effort of running them. The second is consistent documentation. When every experiment is recorded the same way, anyone can read any experiment and understand it quickly. That makes reviewing results and comparing experiments far easier. The third is better learning across the team. Experiments are valuable mainly for what they teach, and a written record lets that learning spread beyond the person who ran the test. The fourth is avoiding repeated experiments. A searchable log of past experiments, with their results and learnings, prevents the team from testing the same idea twice without knowing it. ## The template The template has fifteen parts. Some take a sentence; a few deserve a paragraph. The example running through each part is a test of a simplified signup form for a SaaS product. ### Experiment name Give the experiment a short, descriptive name that someone can recognize at a glance in a list. "Simplified signup form test" works. "Test 14" does not. ### Owner Name the person responsible for running the experiment, analyzing results and writing up the learnings. Shared ownership usually means no ownership. ### Area Identify which part of the growth model the experiment targets: acquisition, activation, retention, monetization or referral. Tagging experiments by area makes it easy to see whether the team is focusing on the right parts of the funnel and to find related experiments later. ### Problem or opportunity Describe the observation that led to the experiment. This should be grounded in data or evidence, not just a hunch. In our example: a large share of visitors who start the signup form abandon it before finishing, and session recordings show visitors hesitating at the company size and phone number fields. Writing down the observation forces you to check that a problem actually exists before trying to solve it. ### Hypothesis State what you will change, what you expect to happen and why. A clear format is: if we make this change, then this metric will move in this direction, because of this reason. For the signup form: if we remove the company size and phone fields, then signup completion will increase, because users will face less friction and fewer concerns about being contacted by sales. The "because" part is important. It records the reasoning, which is what you actually learn about when the result comes in. If the hypothesis is confirmed, the reasoning is supported. If it is not, the reasoning needs rethinking. ### Primary metric Choose the single metric that determines whether the experiment succeeded. For the signup form, that is signup completion rate, the share of visitors who start the form and finish it. Having one primary metric prevents the common problem of looking at many metrics afterward and declaring success based on whichever one happened to improve. ### Secondary metrics List other metrics you will monitor to understand the experiment's broader effects. For the signup form, these might include activation rate among new signups, lead quality and how often sales follows up with new signups. A shorter form might bring in more signups who are less likely to become customers, and secondary metrics reveal that. ### Guardrail metrics Name metrics that should not get worse as a result of the experiment. Removing the company size field might make it harder for sales to prioritize leads, so qualified demo requests are a sensible guardrail. If signups rise but qualified demo requests fall sharply, the experiment may not be a win overall. Guardrails protect against optimizing one number at the expense of the business. ### Experiment design Describe how the experiment will run. Will it be an A/B test, where visitors are randomly shown one version or the other? A before-and-after comparison? A qualitative test with a handful of users? Describe the variants, the audience included and how traffic will be split. For the signup form, an A/B test with visitors randomly split evenly between the current form and the shorter one is a natural choice if traffic allows. ### Sample size and duration Estimate how many visitors or users the experiment needs and how long it will take to reach that number. This depends on your current conversion rate, the size of the change you hope to detect and your traffic. Online calculators can help. Decide this before launching, and avoid stopping the experiment early just because the results look good after a few days. Early results often swing and settle. If your traffic is too low to reach a meaningful sample in a reasonable time, consider a different design, such as a qualitative test or a larger change that would produce a bigger effect. ### Expected impact Make a rough estimate of the potential impact if the hypothesis is right. This does not need to be precise. Its purpose is to help prioritize, by distinguishing experiments that could make a meaningful difference from those that would only nudge a minor metric. ### Priority score Use a simple framework to compare experiments and decide which to run first. A popular option is ICE, which scores each experiment on impact, confidence and ease, typically from one to ten each, then combines the scores. The signup form experiment scores well on ease, since removing fields is quick, and reasonably well on impact and confidence given the session recording evidence. Priority scores are rough judgments, not precise calculations. Their value lies in making the team discuss and compare ideas explicitly. ### Launch date Record when the experiment starts. This matters for interpreting results, especially if something else changes during the test, such as a marketing campaign or a product release. ### Results When the experiment ends, record what happened to the primary metric, secondary metrics and guardrails, along with statistical confidence where it applies. Be honest. Record inconclusive results as inconclusive and negative results as negative. Those are often the most useful records of all. ### Learnings and next steps Finally, write down what the team learned and what happens next. Did the result support the reasoning in the hypothesis? What does it suggest about your users? Will you roll out the change, run a follow-up experiment or abandon the idea? For the signup form, suppose completion rose and qualified demo requests held steady. The team might roll out the shorter form and plan a follow-up experiment collecting company size later in onboarding, so sales still gets the information without adding friction at signup. ## Keep experiments in one shared log Store every experiment, using the template, in a single place the whole team can access and search. A shared document, a spreadsheet or a project management tool all work. Review the log regularly, both to prioritize upcoming experiments and to revisit what past experiments taught. Over time, the log becomes one of the most valuable resources a growth team has, because it captures accumulated knowledge about what works for your specific product and customers. ## Run experiments at a sustainable pace It is better to run a few well-designed experiments consistently than many poorly designed ones. Pick a cadence the team can sustain, such as one or two experiments in progress at a time, and protect it. Make sure experiments running at the same time do not interfere with each other, for example two tests on the same page that could affect each other's results. ## Avoid the common mistakes The mistakes that weaken growth experiments are well known. Teams launch without a clear hypothesis, so they cannot interpret results. They choose the primary metric after seeing the data. They stop tests too early. They ignore guardrails and harm the business while improving one number. They run tests with too little traffic to detect anything. They fail to document results, especially negative ones. And they never review past experiments, so lessons are lost. ## Bringing it together A growth experiment template turns experimentation from guesswork into a learning system. Record the name, owner and area. Ground each experiment in a real observation. State a clear hypothesis with its reasoning. Define primary, secondary and guardrail metrics. Design the test and estimate how long it needs. Prioritize with a simple framework. Then record results and learnings honestly in a shared log. Teams that do this consistently build a growing understanding of what drives their growth, which is the real goal of every experiment. If search is one of your acquisition channels, the same discipline applies to SEO tests, and it is part of how our [SEO growth service](/services/seo-growth) decides what to change on a site. ## Frequently asked questions ### What should a growth experiment template include? Problem, hypothesis, primary and secondary metrics, guardrails, design, duration, results, learnings and decision. ### What is a good hypothesis format? "If we [change], then [metric] will [change], because [reason]." ### Should I document failed experiments? Yes. Failed experiments provide valuable learning and prevent repeated tests. ## About Comppound Comppound runs SEO and AI search for B2B SaaS companies: a person sets the plan and reviews samples and every high-risk page, and AI speeds up production. - [Get the free report](https://comppound.com/ai-visibility-report): When buyers ask ChatGPT about your category, are you in the answer? - [Book a call](https://comppound.com/contact): Most of this works better when buyers can find you first