Working with Hypotheses

Creating, editing, and managing ideas for experiments.

What is a Hypothesis in AB-Labz Workbench

A hypothesis is a structured card with a description of a proposed change in the product. It includes the formulation of the idea, expected effect, target metric, and experiment planning parameters.

Hypotheses without results are located in the Hypotheses section. After completing the experiment and saving results, the hypothesis automatically moves to the Archive.

Creating a Hypothesis

Open the Hypotheses section and click "Add Hypothesis". In the modal window, fill in the fields:

Required Fields

Name (up to 255 characters): Brief description of the change. For example: "New CTA button on homepage", "Simplified registration form".

Hypothesis Text: Detailed formulation of the assumption. Recommended format: "If [change], then [expected effect]".

Example: "If we add a bright CTA button to the homepage, then the number of registrations will increase".

Optional Fields

Key Metric: The indicator by which the experiment's success will be evaluated. You can select a metric from the catalog via the dropdown list. If the metric is not created in advance, leave the field empty — it can be added later.

Expected Change (%): Projected effect as a percentage relative to the current value. For example: 10 means +10% to the baseline metric value. This field is used for sample size calculation.

Required Sample: Minimum number of users required for the experiment. Calculated using the sample size calculator tool and filled in manually.

External ID: Experiment identifier from your split system. Used to link the experiment in AB-Labz Workbench with the experiment in your system (also used in API).

Tags: Arbitrary labels for grouping hypotheses (up to 3 tags, each up to 10 characters). For example: onboarding, ui, payment.

After filling in, click "Save". The hypothesis will appear in the list with "queued" status.

Viewing and Filtering

The Hypotheses section displays all active hypotheses without results, sorted by creation date (newest first).

For each hypothesis, you can see:

• Name and author
• Key metric (if specified)
• Expected change
• Start and end dates (if scheduled)

Click on a hypothesis card to open detailed view with full information.

Editing a Hypothesis

All organization members can edit any hypotheses.

Open the hypothesis card and click "Edit". You can change any fields, including name, text, metric, and additional parameters.

Typical editing scenarios:

• Clarifying wording after team discussion
• Adding a key metric if it wasn't specified during creation
• Updating required sample after recalculation
• Changing tags for better grouping

Changes are saved immediately after clicking "Save".

Scheduling Dates

To schedule an experiment, open the hypothesis card and click the "Schedule" button. In the modal window, specify the start and end dates of the experiment. After saving, the hypothesis will appear in the Experiments Calendar, where you can track all scheduled tests and avoid overlaps.

Recommended order:

1. Create a hypothesis with basic parameters
2. Calculate sample size in the sample size calculator tool
3. Determine experiment duration based on traffic
4. Add dates to the hypothesis
5. Check calendar for conflicts

Dates can be changed before and during the experiment through the planning interface — they are only needed for planning and do not affect data analysis.

Saving Results

After completing the experiment, results are saved from the Data Preparation section:

1. Upload experiment data to the Data Preparation section
2. Run analysis and review results
3. Click the "Save Results" button that appears after analysis
4. In the modal window, select one of the options:
   • Save to hypothesis — attach results to an existing hypothesis from the list
   • Create experiment — create a new hypothesis on the fly and immediately attach results to it

Important: If you analyze data by different segments (for example, by country or device), save each result separately to the same hypothesis. After saving, the hypothesis will show toggles to view all segment slices.

After saving results, the hypothesis automatically moves to the Experiments Archive and disappears from the Hypotheses section.

Deleting a Hypothesis

Hypotheses can be deleted by:

• Hypothesis author — only their own
• Administrators — any in the organization

Open the hypothesis menu ([...]) and select "Delete". Confirm the action in the dialog box.

Deletion is irreversible. If a hypothesis has saved results, it's better to leave it in the archive than delete it — this preserves experiment history.

Calendar Integration

Hypotheses with start and end dates are automatically displayed in the Experiments Calendar.

The calendar helps:

• Visualize the team's experiment load
• Avoid parallel launch of conflicting tests
• Plan resources based on schedule

Best Practices

Formulate hypotheses specifically: Not "Improve conversion", but "If we change the button color from blue to orange, then the number of registrations will increase".

Use links: The hypothesis field supports pasting links. Attach links to research to justify the hypothesis or to designs in Figma/Miro so the team can visually assess changes. This may seem like overcomplication, but over time, when you study the experiments archive, you'll thank yourself for the context.

Specify key metric: This focuses the team on the main success indicator and simplifies subsequent analysis.

Document results in detail: Even failed experiments are valuable. Write not only "didn't work", but also why, what hypotheses it refutes, what to try next.

Use tags: Group hypotheses by direction (onboarding, payment, retention) to analyze work efficiency by topic.