Growth Experiments Template (Free, with ICE + ROTI Scoring)
A growth experiments template is a structured format for writing a single testable bet. The standard fields are observation, change, predicted effect, audience, and success criteria. Filling them in before you build forces the hypothesis to be falsifiable, the metric to be specific, and the bar for shipping to be set in advance, so the result of the test produces a decision instead of an argument.
This page gives you GrowthLab's free 5-field template with an example filled in, compares it to the other widely-used formats (Lean Startup hypothesis, PIE, ICE-scored), and shows how to score what to test next using ICE and ROTI. Copy the template, paste it into your doc, or use it directly inside GrowthLab.
The GrowthLab Growth Experiment Template
Five fields, in this order. Filling all five before you build is the single biggest predictor of whether the result will be useful afterward.
1. Observation
The data or insight that motivates the test. Grounds the bet in evidence rather than a hunch. If you cannot point to a number, a customer quote, or a clear pattern, the hypothesis is a guess.
Example: Analytics show 40% of new signups drop on the email-verification step.
2. Change
The single, specific thing you will do differently. One variable. Vague changes produce vague results.
Example: Replace email verification with a magic link for new signups.
3. Predicted effect
Which metric you expect to move, and in which direction. Pick one primary metric before you launch. Watching ten metrics invites cherry-picking after the fact.
Example: Lift signup completion rate.
4. Audience
Who sees the change, with the constraint that makes the test reproducible. Be explicit about traffic source, device, or segment if it matters.
Example: New visitors landing on /signup over the next two weeks.
5. Success criteria
The result that would justify shipping the change. Set the bar before launch so a flat or negative outcome becomes a clean learning, not an argument about what counts.
Example: A significant lift in signup completion rate over a two-week test, at standard significance.
Example: the five fields, assembled as a full hypothesis
Because analytics show 40% of new signups drop on the email-verification step, we will replace email verification with a magic link for new signups, and expect signup completion rate to lift among new visitors over the next two weeks. We will ship if the lift is significant at standard significance.
Other template formats compared
Three other formats you will see in the wild. Each captures something slightly different, and a strong program often combines a hypothesis template with a scoring lens like ICE or PIE.
| Format | What it captures | When to use |
|---|---|---|
| GrowthLab 5-field | Observation, change, predicted effect, audience, success criteria | Default. Forces falsifiability and a pre-set ship bar in five fields anyone on the team can fill in. |
| Lean Startup hypothesis | "We believe [change] will result in [outcome] for [audience]. We will know we are right when [signal]." | Quick framing for early-stage teams. Lighter than 5-field but easier to leave assumptions implicit. |
| PIE (Potential, Importance, Ease) | A prioritization scoring lens, not a hypothesis structure | Use alongside a hypothesis template to rank a backlog. Adjacent to ICE. |
| ICE-scored | Add Impact, Confidence, and Ease scores (1 to 10) to any hypothesis | Use after the hypothesis is written to rank it against the rest of the backlog. ICE is the standard for fast prioritization. |
How to score what to test next
A hypothesis template tells you how to write a single bet. A scoring lens tells you which bet to run next. The pair: ICE before, ROTI after. ICE estimates Impact, Confidence, and Ease on a 1 to 10 scale to rank candidates; ROTI grades the time-payback of a finished experiment so your next ICE scores are smarter. See the ICE scoring and ROTI definitions for the formulas.
Growth experiment loop, from template to learning
The template is step one of a loop that compounds. The full sequence, in order, is what turns a stack of finished tests into a system that gets sharper over time.
1. Write the hypothesis using the template
Fill in all five fields. If any field is empty, the experiment is not yet ready to run.
2. Score with ICE
Rate Impact, Confidence, and Ease on a 1 to 10 scale. Rank the candidate against the rest of the backlog.
3. Run the test
Ship the variant, hold it until the predetermined sample size or duration is reached, and watch only the primary metric.
4. Review with ROTI
Weigh the result and the learning against the real time the test consumed. Mark the pattern so similar future tests inherit the lesson.
5. Capture the learning
Store the hypothesis, result, and one-sentence takeaway in a searchable library, so the team stops repeating dead ends.
6. Compound into the next batch
Use what you just learned to adjust the Confidence and Ease scores of related ideas still in the queue. The loop tightens with every turn.
Frequently asked questions
What should a growth experiment template include?
At minimum: an observation grounding the test in evidence, the specific change you will make, the metric you expect to move and in which direction, the audience who will see the change, and the success criteria that would justify shipping. Setting the success criteria before launch is the field most often skipped and the one that protects you from explaining away a flat result later.
What is the difference between a hypothesis and a growth experiment template?
The hypothesis is the testable statement. The template is the format you write it in. A strong template forces the hypothesis to be falsifiable by requiring an observation, a specific change, a primary metric, an audience, and a pre-set bar for success. Without a template, hypotheses tend to drift into goals or opinions that cannot be cleanly proven or disproven.
Is there a free growth experiments template?
Yes. The five-field template on this page is free to copy and use directly. GrowthLab is also free to start as a managed tool that captures the same template, runs ICE and ROTI scoring on it, and stores the result in a searchable learnings library.
What is the best format for a growth experiment hypothesis?
A reliable structure is: because we observed X, we will change Y, expecting metric Z to move for audience A, and we will ship if result R. It names the evidence, the change, the predicted effect, the audience, and the bar for success. Lean Startup phrasing works for the same idea in fewer words; pick the version your team will actually fill in completely.
Should I score growth experiments with ICE or RICE?
ICE is faster and fine when candidate experiments affect similar audience sizes. RICE adds Reach (how many users are affected) and divides by Effort, which helps when ideas touch very different audience sizes. Start with ICE and graduate to RICE only when reach differences distort your ranking.
Can I use this template inside an experiment management tool?
Yes. GrowthLab uses this exact five-field structure for every experiment, AI-drafts the first version when you describe the idea, runs ICE and ROTI scoring on top, and stores every result in a learnings library. The template works in a doc or a spreadsheet too; the tool just removes the friction of scoring and remembering what you tried.
Go deeper
- What is ICE scoring?
- What is ROTI?
- What is an experiment hypothesis?
- Experiment prioritization guide
- Growth experimentation: the 2026 ultimate guide
- Best experiment management tools (2026)
About GrowthLab
GrowthLab is an experiment management tool where AI drafts the hypotheses, ICE and ROTI prioritize them, and every learning compounds into the next batch.
