Smart search suggestions and predictive results: +18.39% purchases
Commentary byLiftArchiveOur analysis, not theirsAdding smart suggestions and predictive results to site search lifted purchases by 18.39% for Andaaz Fashion.
ConvertPolo ran this A/B test for the ethnic fashion brand, and it was published as a VWO case study. Search engagement rose 28%.
- Purchases: +18.39% for the variant, tracked through order confirmations.
- Search engagement: +28%, meaning more shoppers used the search tool.
- No sample size, duration or significance level was shared.
The change
Variant adds suggestions and predictive product results as shoppers type.
Simplified mockup drawn by LiftArchive from the shared screenshots.
Control
The search bar only worked if shoppers typed exact product names. It gave no suggestions, image previews or category filters.
Variant
Search showed smart suggestions and predictive results as shoppers typed, with better navigation to relevant products.
Why it likely won
Our reading of the result, not the practitioner's claim.
- Shoppers rarely know exact product names. Suggestions forgive typos and vague terms.
- Predictive results let people jump straight to a product page, which shortens the path to buying.
- Ethnic wear has many specific terms, so help with naming and spelling may matter even more here.
Caveats: The case study gives no sample size, duration or significance. Results are reported by the agency and testing vendor.
Who ran it
ConvertPolo ran this test for Andaaz Fashion and published the results (VWO case study). The numbers are theirs; the summary and commentary on this page are ours.
Read the original case study on ConvertPolo →
How to run this test on your store
- Hypothesis
- If we add autocomplete suggestions and predictive results to site search, then more searchers will reach a product and buy, because they no longer need to know exact product names.
- Primary metric
- Purchase conversion rate of visitors who search
- Guardrails
- Zero-result search rate
- Search exit rate
- Mobile page speed
- Who should test it
- Stores with a large catalog where many visitors use search, especially when product names are hard to spell or guess. It matters little if few visitors search.
Questions
Does autocomplete in site search increase sales?
It often helps because shoppers find products faster and hit fewer dead ends.
In this test it lifted purchases by 18.39%, though results vary by store.
What should predictive search show?
Usually suggested terms, a few matching products with images and prices, and relevant categories. Make sure it loads fast on mobile.
Similar tests
@oliverkenyon
Blend CommerceBlend CommerceBlend CommerceBlend CommerceBlend Commerce