Growth experiment template: how to run and track marketing tests

What is a growth experiment?

A growth experiment is a small, time-boxed test of one assumption about how people find, try, or buy your product. You write down the assumption, the action you will take, what you will count, and the decision rule before you start. When the window closes, you record the result and the decision. Running growth experiments means doing this repeatedly, from one shared log, so each test is chosen because of what the last one showed.

A single well-designed test is covered in how to test a startup idea. This guide is about what comes after: how to keep a list of experiments, choose the next one, track them while they run, and look at several results together without fooling yourself.

Why a single test is not enough

One test answers one narrow question for one audience and one message. Founders often run a test, get an unclear result, and then try something unrelated the following week. After a month they have four half-remembered results that cannot be compared, because nothing was written down in the same way.

A growth experiment log fixes that. Every test uses the same fields, so you can see which audience replied, which message was ignored, and which decision you made. The log is also where you notice that you have been changing three things at once, or that a channel keeps getting "one more try" without ever meeting its rule.

The growth experiment template

Use one row per experiment in a spreadsheet, or one page per experiment in a document. Fill in every field above the line before the test starts. Fill in the rest when the window closes. Do not edit the decision rule after you have seen any results.

Two fields do most of the work. Measure keeps the denominator next to the count, so "four replies" always reads as "four replies from 30 messages." Next step links each experiment to the one after it, so the log shows a chain of decisions instead of a list of unrelated tries.

Fill the backlog with assumptions, not tactics

A backlog of growth ideas easily turns into a list of tactics: "try TikTok," "run a giveaway," "post on Product Hunt." A tactic does not tell you what you will learn. Rewrite each idea as an assumption about a specific audience before it goes in the log.

• "Try TikTok" becomes "Freelance designers who post work on TikTok will click a link to a pricing calculator from a short demo."
• "Run a giveaway" becomes "People who enter a giveaway for our template will open the follow-up email about the paid version."
• "Post in communities" becomes "Founders asking how to price a first product will reply when we answer their thread in full."

Good sources for assumptions are the guesses in your one-page startup marketing plan, the words people used in customer interviews, and the decisions from your last experiment. If an idea cannot be written as an assumption, it is not ready for the log yet.

Choose the next experiment

Many growth teams score their backlog with ICE, a method popularized by Sean Ellis: rate each idea for impact, confidence, and ease, then sort by the combined score (Breakout Growth episode on ICE). It is a fast way to compare many ideas. With a small team and little data, the scores are mostly guesses, and it is easy to give your favorite idea a high number.

For a solo founder, a shorter rule usually works better. Ask three questions in order:

1. Which assumption, if it is wrong, would change what I do next month? Test that one first, even if it is not the most exciting.
2. What is the cheapest action that would show a behavior? A message, a page, or a priced offer, not a new feature.
3. Can I finish it within two weeks? If not, cut the audience or the action until you can.

Run one experiment at a time per audience. If two experiments reach the same people in the same week, you cannot tell which one caused a reply.

Track experiments with a weekly review

Pick one fixed time each week, such as Friday afternoon, and spend 30 minutes on the log. A short, regular review is what turns separate tests into a loop.

1. Update the counts of every running experiment from the source: your sent folder, analytics, or payment records.
2. Close experiments whose window has ended. Write the result first, then compare it with the decision rule, then write the decision.
3. Write the next step for every closed experiment, even if the decision is to stop.
4. Choose what starts next week with the three questions above, and fill in its fields before you begin.

Keep the review honest by recording what did not happen. Silence is a result. Eight of ten people not replying belongs in the log as clearly as the two who did.

Growth experiment example: four weeks of one log

Here is what a log can look like after four weeks for a founder selling a proposal follow-up tool to freelance consultants. These are illustrative numbers, not results from a Marketing Loop customer.

Read the log top to bottom and a story appears. The first message reached too broad an audience. Narrowing to consultants who send many proposals more than doubled the replies, with the same message. Those people were willing to talk. The LinkedIn post brought visits but few signups, and one post is not enough to drop the channel, so the next step is a second post with a worked example before deciding.

Notice what the log does not claim. Seven replies from twenty messages is a useful signal for this audience and this message. It does not estimate demand across all consultants, and four calls do not show that anyone will pay.

Read results across experiments, not one at a time

Once you have five or more closed experiments, look across them before choosing the next one.

• Group by audience. If one audience keeps meeting its rule and another never does, that matters more than any single message.
• Group by type of behavior. Replies are cheaper than calls, and calls are cheaper than payments. A log full of replies and no priced offers means the hardest question has not been asked yet.
• Look for repeated words. Objections and phrases that appear in several results belong in your next message or page.
• Count the "one more try" decisions. A channel that has been on "gather more evidence" three times needs a stop or a real change.

What a growth experiment log cannot do

A log keeps you honest about what you observed. It does not make small samples representative, and it does not turn a set of encouraging replies into product-market fit. Each result is evidence for one audience, one message, and one window.

It also cannot fix a test after the fact. If you change the audience, message, and price at the same time, the log will record the result, but neither it nor you can say which change explains it. Change one thing per experiment whenever you can.

Keep the log next to the decisions

The hardest part of running growth experiments is not designing the next test. It is remembering why you chose it. A spreadsheet works if you review it every week and never skip the decision field.

We are building Boring Marketing Loop to keep that chain in one place. Each test records its message or link, the sample size, an end date, and the signal you chose before it started. After the result, you decide to continue, change direction, stop, or gather more evidence, and the next test is planned from that decision. We use the same fields in our own marketing test log.

Marketing Loop is in development. If you want to know when it is ready, join the build updates.

FAQ

What is the difference between a growth experiment and an A/B test?

An A/B test shows two versions to comparable groups at the same time and compares the results. A growth experiment is any time-boxed test of an assumption with a decision rule. An A/B test is one kind of growth experiment, and it needs more traffic than most early startups have.

How many growth experiments should a solo founder run at once?

Usually one or two, and never two on the same audience at the same time. More running experiments mean less time to do each one properly and more overlap between their results.

How long should a growth experiment run?

Long enough for the people you reached to respond, and short enough that you still remember why you started it. One to two weeks fits most outreach and content tests. Set the window before you start and close the experiment when it ends, even if the result is unclear.

Should I use a spreadsheet or a tool to track experiments?

A spreadsheet is enough to start. What matters is that every experiment uses the same fields and that you write the decision rule before the test. Move to a tool when the spreadsheet stops helping you see the chain from one decision to the next experiment.

Start your log this week

Copy the template into a spreadsheet. Write your three most important assumptions as backlog rows, choose one with the three questions, and fill in its fields before you send anything. Put the weekly review in your calendar, and close the first experiment on the day its window ends.