# A language app compares two opening arguments

Companion to /learn/ios/product-page-optimization/

## Outcome
Prepare an Apple product-page experiment with a clear creative question and release plan.

Fictional example, not a client result.

## Scenario
PhrasePilot is a fictional travel-language app. The current opening shows a lesson menu; the proposed treatment shows a supported restaurant conversation. The team wants to know whether a practical situation helps the intended audience understand the app. It is not testing whether language learning as a category is attractive.

## Reasoning
Confirm that the intended asset and page are eligible in the current App Store Connect setup before designing the experiment. Separate the creative hypothesis from the platform configuration. Review the audience, locale, treatment assets and competing release activity. A beautifully designed treatment is unusable if it belongs to a surface the chosen experiment cannot evaluate.

## Action
Prepare a product-accurate conversation screen and a concise caption. Check the treatment on the relevant devices and obtain approval before launch. Save the control and treatment versions, record the configuration, and use the platform’s result definitions in the review. If the evidence is inconclusive, explain that rather than presenting the visual preference of the team as a test outcome.

## Trap to avoid
Do not treat an experiment estimate as a guaranteed forecast for every future visitor. The tested audience, timing and treatment matter. A rollout decision should preserve that context and identify what needs monitoring after publication.

## Advanced considerations
Assign a release owner before the experiment begins. Someone must confirm whether the chosen creative was actually applied and whether related localized assets remain coherent. A test that produces a decision but no verified handoff creates a gap between learning and execution. Keep an archive of the rejected creative too; it records what was tested and prevents accidental reuse without context.

## Exercise
The marketing team changes the acquisition campaign during the experiment. What belongs in the result review?

## Suggested answer
Record the campaign change, dates and potential audience implications. Inspect available segmentation and the experiment context. Do not silently claim the observed result describes a stable audience if the conditions materially changed.

## Your worksheet

### Current Apple setup reference and eligibility
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Creative question and supported product screen
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Control and treatment asset versions
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Locales, devices and approvals
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Metric definition and review plan
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

### Result interpretation and deployment record
Answer:
Evidence:
Uncertainty:
Owner:
Next action:

## An experiment handoff for PhrasePilot
The fictional language app compares its lesson-menu opening with a supported restaurant-conversation opening. The example keeps creative approval, experiment readiness and publication as separate decisions so a good-looking asset is not mistaken for a completed experiment.

Illustrative ownership and acceptance gates, not an App Store Connect interface replica.
Gate | Required record | Responsible role / exit condition
Creative ready | Control and treatment files, product proof and locale | Creative lead: approved, accurate and readable assets
Setup ready | Actual supported page, assets and selected configuration | Experiment owner: console eligibility and review requirements checked
Interpretation ready | Metric definition, planned review and concurrent releases | Analyst: comparison can answer the stated question
Decision recorded | Result, uncertainty and reasoning for chosen action | Decision owner: rollout, retain or investigate explicitly selected
Release verified | Exact applied asset bundle and observation date | Publisher: confirm the intended version became visible

### Resolve the audience assumption before launch
A travel-language audience in your brief is not automatically the audience isolated by the experiment tool. Record the actual listing and eligible population, and consider how broadly the result can be interpreted. If the intended audience cannot be separated, narrow the conclusion rather than describing the test as precise audience targeting.

### Keep a release conflict log
Record campaign shifts, product releases and metadata changes that could complicate interpretation. Avoid unnecessary unrelated changes during a focused comparison, but do not delay an important accuracy correction merely to preserve an experiment. When conditions change, document the consequence and decide whether the original question remains answerable.

### Verify the decision after applying it
Archive the tested treatment, result context and rollout instruction. After publication, compare the visible assets with the intended version in the relevant context. A decision slide is not evidence that the treatment was applied correctly. Continue observing relevant signals without treating every post-release fluctuation as confirmation or reversal of the experiment.

## Continue reading
- /learn/conversion/ab-test-hypotheses/
- /learn/conversion/screenshot-strategy/
- /learn/measurement/metadata-change-log/