TrustInsider
English
Analyse your brand

User manual

Design and read an experiment

Separate experiments on your landing pages from comparisons of AI answers. Define the change before reading the outcome, and always check coverage and limits.

Section 1

Prepare a landing-version comparison

The comparison uses a published page as control and a different version of it as treatment. A clear hypothesis explains what you intend to test.

Suggested form values are not a universal recommendation. Check that the hypothesis, duration and planned sample fit your case; a saved draft is not a started experiment.

  1. Select the landing page and check its published control version.
  2. Choose the alternative version and describe the hypothesis and planned changes.
  3. Enter the required baseline rate, minimum effect and duration; save the draft design.
  4. Review the design and use the dedicated confirmation before starting.
The plan for a comparison between two versions of a landing page, with the command to save it before it starts. (fictional brand, demo data)
Demo data · Ilmarde Audio

Section 2

Define the AI-answer comparison

The design starts from a saved brand panel. Separate treated and control questions, and specify the intervention and comparison periods. Control questions help separate the effect from general changes.

The comparison hypothesis follows the chosen design. Saving does not start a collection or produce new observations: evaluation needs data matching that panel and those periods.

  1. Choose the panel and read its country, language and version.
  2. Assign questions to the two groups without placing the same question in both.
  3. Choose assistants available in the panel, the intervention factor and its description.
  4. Set the before period, intervention date and after period, then confirm the design.
The design of a comparison on AI answers: treated and control questions, observed assistants and periods. (fictional brand, demo data)
Demo data · Ilmarde Audio

Section 3

Read outcome, sample and interval together

The summary uses available outcomes. Missing data or an inconclusive comparison do not become an improvement or a measured zero.

Before-and-after alone does not establish the cause of a change. A positive treatment is not automatically applied or published; that remains a separate decision.

  1. Open the result and check its status, available observations and group coverage.
  2. Read the difference and interval when present, alongside comparison conditions.
  3. Check that required time and data have matured before concluding.
The outcome of an experiment on answers: the estimated effect with its interval, the treated and control chart and the table. (fictional brand, demo data)
Demo data · Ilmarde Audio

Section 4

Manage status without losing context

Available actions depend on the experiment’s status, which tells you what stage the comparison has reached. Finalisation may remain blocked until the displayed conditions are met.

Cancelling an experiment and choosing content to publish are different decisions. Keep the original hypothesis and comparison conditions in view when interpreting the result.

  1. Before starting, cancelling, finalising or deleting a landing experiment, read its confirmation.
  2. If Finalise is blocked, check status and maturity rather than treating the block as a negative outcome.
  3. If the version changed or the outcome is uncertain, refresh before retrying.
An experiment running on a landing page: the sample collected, the interim result and the commands that ask for confirmation. (fictional brand, demo data)
Demo data · Ilmarde Audio

Related paths

Definitions, FAQs, guides and methodology: from here you reach the material you need to read the results and explain them to your team in the same words.

From question to action

Learn how to use TrustInsider: choose the brand and the questions, read the collected answers, update your content and compare the results.

Read

Find the right answer

Help with access, credits, collections, integrations and data: guides, FAQs, the site assistant and support requests from your workspace.

Read
Apply the method to your brand