# Testing Decisive Recommendation Wording on Product Pages

Page: https://alidemirbas.com.tr/en/lab/ab-test-library/tests/testing-decisive-recommendation-wording-on-product-pages

## Hypothesis

If a single, well-founded recommendation appears next to otherwise neutral options, shoppers who are stuck between choices may narrow them down more easily. Purchase conversion on the product page could improve, though a recommendation that does not fit may feel pushy and work against it.

## When to run this test

Session recordings of shoppers switching back and forth between options, support requests asking which one to pick, or heavy drop-off at the selection step show people freezing in front of the choice. If the recommendation helps, shoppers in that arm hesitate less, make a choice and go on to buy. If the hesitation comes from unclear option descriptions, clarify those first; before launch, also verify that the recommended option genuinely suits the customer and was not picked for its margin alone. Since everyone sees the same recommendation, do not word it as if it were personalized.

## Primary metric

**Purchase conversion rate (CR)**: Compare purchases among eligible visitors assigned to neutral options or to one clearly stated recommendation.

## Other metrics

- **Revenue per visitor (RPV)**: Show how changes in purchase conversion affect total revenue.
- **Return rate**: Track returns over the same follow-up period; do not attribute every return to the recommendation.
- **Recommended option selection rate**: Report the share selecting the featured option among all assigned visitors and among those who make a selection.

## What to check before and during the test

- **Message**: Compare neutral option presentation with one explicit recommendation.
- **Fit**: Verify that the recommended option is suitable for the intended customer.
- **Rationale**: Fix the wording and any supporting rationale before the test.
- **Other choices**: Keep every alternative accessible and record selections across options.
- **Audience**: Review defined customer groups without assuming one recommendation fits everyone.

## What to watch when reading the result

- Even if purchase conversion rises, do not reach a verdict before checking return rate and revenue per visitor, and compare returns once the same follow-up period has passed in both arms. If sales go up and returns follow, the recommendation may be steering people toward an option that does not suit them, so do not call it a winner.
- The result covers this fixed message and this recommended option. A different option or a different tone is a separate test, and changing both at once leaves the effect impossible to attribute. Report the recommended option selection rate among all assigned visitors and among those who choose, and review the defined customer groups on their own.
- A small purchase gap does not mean the recommendation failed; a shift in selection share without a change in purchases may mean it only moves choices around. The [A/B test significance calculator](https://alidemirbas.com.tr/en/calculators/ab-test) shows whether the conversion gap could be chance.

## Frequently Asked Questions

### Which option can I recommend?

Recommend an option with a verified basis. Do not call a high-margin but unsuitable choice the best for the user, or imply personalization when everyone sees the same recommendation. Use purchase conversion as the primary outcome.

### How should I report the recommended option's selection share?

Report it among all assigned visitors and among people making a selection. Preserve access to every alternative and record the choices. Restricted access would test obstruction rather than recommendation.

### Can I keep the recommendation if conversion rises?

Check returns and revenue per visitor before deciding. Review predefined user groups rather than assuming the recommendation suits everyone.