> ## Documentation Index
> Fetch the complete documentation index at: https://docs.simulant.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Run a test

> Test creative and messaging at population scale — set up variants, audience, context, metrics, and questions, then read the results.

A **test** runs a structured study across a whole cohort. Use it to pre-test creative, compare messages, or rehearse a launch — and get both the numbers and the reasoning behind them.

<Frame caption="The Tests list — every study with its format, headline sentiment, and status at a glance.">
  <img src="https://mintcdn.com/simulant/j7X70BC9NUZuF_VF/images/app/tests-list.png?fit=max&auto=format&n=j7X70BC9NUZuF_VF&q=85&s=de932f4cd31219877dec647caedb8bf1" alt="Tests list showing campaign message tests with sentiment scores and statuses" width="1428" height="779" data-path="images/app/tests-list.png" />
</Frame>

## Build a test

Go to **Tests → New test**. The builder is organized into sections; work down the form and use the live panel to preview your setup.

<Frame caption="The test builder — identity, creative, variants, and audience on the left with a live preview and estimate on the right.">
  <img src="https://mintcdn.com/simulant/j7X70BC9NUZuF_VF/images/app/test-builder.png?fit=max&auto=format&n=j7X70BC9NUZuF_VF&q=85&s=21d5ac4dbf2aab5ee4869292b2f44994" alt="Test builder for a cost-of-living TV spot with variants A and B and a Young Voters cohort selected" width="1428" height="779" data-path="images/app/test-builder.png" />
</Frame>

### Identity

Name the test and state a **hypothesis** — what you expect to happen and why. Stating it up front keeps the read-out honest and gives you something to check the result against.

### Creative

Choose the format:

* **Video** — react to a video cut.
* **Image** — react to a static asset.
* **Messaging** — react to copy you type or upload.

Add **variants** to compare options side by side — up to four, labelled A through D. Attach a creative or message to each. See [Creatives](/guides/creatives).

### Audience

Pick the **cohort** the test runs against, or create a new one inline. Every variant is shown to the same audience, so differences in the result come from the creative, not the crowd. See [Build a cohort](/guides/cohorts).

### Context

Set the real-world environment the population is reacting within:

* **Current** — today's prevailing issues, mood, and events.
* **Neutral** — strip out context for a clean baseline.
* **Custom** — specify the **issues**, **mood**, **events**, and **competitive** pressure yourself.

Context is saved with the test so a run can be reconstructed and validated later.

### Metrics

Select the quantitative measures you want scored, such as **Recall**, **Sentiment**, and **Purchase intent**. Each is reported out of 100 and can be cut by segment.

### Questions

Add the questions each persona answers. Question types include:

* **Open-ended** — free-text responses that become verbatims.
* **Rating 1–5** — a numeric scale.
* **Multiple choice** — fixed options you define.

Drag to reorder questions. A good test mixes a few open-ended prompts with structured ratings.

### Panel and behavior

Tune how the run executes:

* **Sample size** — how many personas to draw from the cohort (capped at the cohort's `n`), or **Max**.
* **Randomize** — vary the order variants are shown to reduce order bias.
* **Attention checks** — include checks that keep responses honest.
* **Agent behavior** — toggles such as allowing **refusals**, requiring **grounded** answers, capturing **verbatims**, and probing with **adversarial** follow-ups.

## Run it

Click **Run test**. Simulant questions the population and assembles the results; you'll be notified when the run finishes. You can also **Save draft** to come back later.

<Note>
  Run time scales with sample size, number of variants, and question count. Larger panels take longer but tighten the confidence of the quantitative measures.
</Note>

## Read the results

Open a finished test to see:

* **Headline metrics** — recall, sentiment, intent, and clarity, each scored out of 100 against a benchmark.
* **Sentiment by segment** — where the creative resonates and where it falls flat, split positive / neutral / negative.
* **Verbatims** — what individual personas said, filterable by tone. Read the reasoning, not just the score.
* **Themes** — verbatims clustered into the patterns that recur across the panel.
* **Recommendations** — the actions the result points to.

<Frame caption="Test results — headline metrics, sentiment by segment, and verbatims you can filter by tone.">
  <img src="https://mintcdn.com/simulant/j7X70BC9NUZuF_VF/images/app/test-results.png?fit=max&auto=format&n=j7X70BC9NUZuF_VF&q=85&s=c5e80ceda5e3eb418e17823979367cb4" alt="Test results for the cost-of-living spot showing recall, sentiment, intent, clarity, segment breakdown, and voter verbatims" width="1428" height="779" data-path="images/app/test-results.png" />
</Frame>

<Frame caption="The Themes tab clusters verbatims into the patterns that recur across the panel.">
  <img src="https://mintcdn.com/simulant/j7X70BC9NUZuF_VF/images/app/test-themes.png?fit=max&auto=format&n=j7X70BC9NUZuF_VF&q=85&s=b4a20626a46764909ef9990e2eed534f" alt="Themes view showing recurring patterns such as cost-of-living resonance and doubts about funding" width="1428" height="779" data-path="images/app/test-themes.png" />
</Frame>

Use **Share** to bring in teammates or **Export** to take the results into a deck or report.

<Tip>
  Comparing variants? Look past the headline number to the segment cuts and themes — a variant that wins overall can still lose the audience you actually care about.
</Tip>

## Next

<CardGroup cols={2}>
  <Card title="Run a focus group" icon="messages-square" href="/guides/focus-groups">
    Hear the same cohort deliberate in conversation.
  </Card>

  <Card title="Compare and track" icon="chart-line" href="/guides/insights">
    Find patterns across every test you run.
  </Card>
</CardGroup>