A Game Test Report Template Your Publisher Will Understand

A test report has four parts in a fixed order: a one-paragraph summary, the setup, the results, and your interpretation. Reviewers read the summary and skim the rest, so the summary must contain the finding. A failed test written up clearly is more useful to a publisher than a vague success.

Why this matters

Reports are how a publisher decides whether to spend their own money testing your game further. They read many, quickly.

A report that requires effort to understand loses to one that does not, even when the underlying result is better.

Writing them consistently also improves your own decisions, because a structure forces you to separate what happened from what you think it means.

What publishers want to see, in order

Four sections. Do not reorder them.

1. Summary. What you tested, what happened, what you would do next. One paragraph.
2. Setup. Build, dates, spend, markets, creatives, audience.
3. Results. The numbers, with sample sizes and windows.
4. Interpretation. Your reading, your hypothesis, your proposed next step.

Keep it to one page plus an appendix. Anything essential that does not fit on the page is going in the wrong section.

Write the summary as though it is the only part that will be read, because for most readers it will be.

The summary paragraph

Three or four sentences, containing the finding rather than a description of activity.

A weak summary says what you did. A strong one says what you learned.

Weak: "We ran a four-day creative test on two hooks across one market with a small budget."

Strong: "Hook A produced installs at roughly half the cost of hook B over four days in one market. Day one retention was similar for both. We believe the mechanic reads more clearly in hook A and would take it into a wider test with two new variations."

The second version contains the setup, the result, and the next step, and it can be read in ten seconds.

The setup section

Everything needed to judge whether the result is credible.

• Build version and platform
• Test dates, and the number of full days
• Total spend and spend per creative
• Markets, listed individually
• Channel, campaign objective, and optimisation event
• Creatives by name, with a thumbnail or link
• Audience and placement settings
• Anything that changed mid-test, stated plainly

That last line is the one that builds trust. Volunteering that a creative was swapped on day two, or that tracking broke for six hours, makes everything else believable.

The results section

Numbers only. No interpretation here, which is what keeps the report honest.

Every row carries its sample size and its window. The field definitions are in The Metrics Sheet Publishers Ask For.

If a number's sample is too small to be meaningful, write "insufficient volume" rather than reporting it. Reviewers respect that considerably more than a figure that falls apart under a follow-up question.

Interpretation, and how not to oversell

This is where reports lose credibility, and the failure is always the same: claiming more than the data supports.

Three rules.

One hypothesis, not a list. "Day one is low, and the level two drop-off suggests difficulty" is credible. Four possible explanations reads as guessing.

Separate what you observed from what you infer. State the observation, then your reading of it, in that order and clearly labelled.

Name the limits. Small sample, single market, short window, one creative pair. Saying this yourself is far better than a reviewer noticing it.

Never extrapolate a small test to a scaled figure. A reviewer knows what happens at scale, per Why Your CPI Went Up After the Test Ended, and an inflated projection undermines the whole report.

What to include when the test failed

Failed tests are worth reporting, and how you write them up says a great deal about the studio.

Include four things.

1. The result, stated plainly at the top. No burying it in the third paragraph.
2. Your best single explanation, with the evidence that supports it.
3. What you would change, specifically enough to be actionable.
4. What you would not change, and why. This is the part that shows judgement.

That fourth point is unusual and it works. A team that can say "retention was poor but the hook clearly landed, so we are keeping the creative direction and rebuilding the first session" is demonstrating exactly the reading ability a publisher wants in a partner.

A clear no is a useful contribution. A vague maybe is not.

What we would do

Write the summary last, after the numbers are in, and rewrite it once. It is the highest-leverage paragraph in the document.

Keep every report you produce in the same format, in one place. Comparability across your own tests is worth more than any single report, and a run of them is a genuine credential.

Send reports on failed tests too. Studios that only surface winners are harder to trust than studios that report everything the same way, and publishers see enough submissions to notice the difference.

The short version

• Four sections: summary, setup, results, interpretation, in that order.
• The summary must contain the finding, not a description of the activity.
• Setup should include anything that changed mid-test, volunteered.
• Results carry no interpretation, and every number carries its sample and window.
• Give one hypothesis, label observation against inference, and name the limits.
• Report failed tests plainly, including what you would not change.

Rewrite your last test as this four-section page. If you want a read on a report before it goes out, send it over.

Related reading: The Metrics Sheet Publishers Ask For, The Publisher Submission Checklist, and Creative Testing on a $500 Budget.