# Switching site search to Google Vertex AI: +20.6% conversion rate

- Outcome: Won
- Page type: Collection page (Ecommerce)
- Elements: Filters, sorting & site search
- Test run by: [36 Agency](https://liftarchive.com/practitioners/36-agency/)
- Commentary by: LiftArchive (our analysis, not theirs)
- Original source: https://www.shoplift.ai/success-stories/how-nimstrata-helped-the-bees-knees-increase-conversion-rate
- Page: https://liftarchive.com/test/bees-knees-ai-search-vs-boost/

Replacing the Boost Commerce search app with Nimstrata's Google Vertex AI search on search and collection pages raised conversion rate by 20.6%. The Shopify agency 36 ran this test with Nimstrata for The Bee's Knees British Imports, and it was published as a Shoplift success story. The test ran for five weeks and reached 99% confidence.

## Result

**+20.6% conversion rate**

- Conversion rate: +20.6%, reported at 99% confidence.
- The case study says the AI's results improved over the test as it learned from shopper behavior.
- No revenue or order value figures were shared.

## Control

Search results and collection pages powered by the existing Boost Commerce search and filter app.

## Variant

The same pages powered by Nimstrata's Retail Cloud Connect app, which uses Google Vertex AI Search for Commerce and filters that adapt to each query.

## Why it likely won (our reading)

- The old search often returned 'No Results Found', which sends shoppers away; better matching keeps them browsing.
- Filters that change with the query show only useful options, so shoppers reach the right product faster.
- A grocery import store has many niche items with odd names, where smarter search helps most.

## Caveats

This is a comparison of two vendor apps, reported by the vendor side. It tests the whole search engine, not one design element.

## How to run this test

- Hypothesis: If we replace our current site search with a search engine that matches intent better, then conversion rate will rise because fewer searches end with no results or the wrong products.
- Primary metric: Conversion rate
- Guardrail: Searches ending with no results
- Guardrail: Revenue per visitor
- Guardrail: Page load time on search and collection pages
- Who should test it: Stores with large or varied catalogs where many visitors use search or filters. It probably won't matter much for small catalogs where most shoppers buy from one or two product pages.

---

Credit: 36 Agency ran this test and owns the result. Summary and commentary by LiftArchive (https://liftarchive.com/test/bees-knees-ai-search-vs-boost/).
Usage: Free to use, quote and build on, with credit and a link: name the practitioner or organization that ran the test and link to its LiftArchive page. The results belong to the practitioners who ran and shared them; the summaries and commentary are LiftArchive's.