Image Similarity Search

The Image Similarity Search is a new innovative feature of the Prefixbox Product Discovery Suite’s Browse Category. It recommends products that look similar to another product, enabling shoppers to easily browse through the retailers' product catalogue, without interrupting the shopping journey.

This innovation benefits all the different subsectors across the eCommerce industry for multiple reasons:

• expressing one’s taste for beauty in words is nearly impossible , but using a visual image as a “search term” results in finding all the relevant products in the catalogue instantly
• webstores using specific terminology such as those in the DIY, furniture, etc. sectors often face the difficulty that shoppers do not know the exact name of products and thus cannot find what they are looking for, even if the product exists in the catalogue. With the Image Similarity Search feature, shoppers can find a similar product first, and then browse through the relevant images without having to learn the actual industry terminology.
• even when shoppers use the correct search term (or a synonym of it), there’s a good chance that they need to use filters to narrow in on the actual product that they want to buy. As these filters are based on pre-defined parameters, the user needs to spend more time refining them and even then they may not reflect the actual user intent. Seeing visually similar products in just one click streamlines the shopping journey.

‘A picture is worth a thousand words’ - has never been more relevant than in eCommerce product discovery. There’s no need to rely only on product metadata anymore to give your customers what they are looking for.

The technology behind the Image Similarity Search feature is completely different from any other search enhancing features, as it does not rely on search words, but only on product appearance. It means that it provides a brand new perspective in the world of search, lifting it to a brand new level by using visuals in addition to search keywords.

Features that support product discovery on the market include

• related products or recommended products, that recommend similar products based on a pre-set parameter
• users also bought, that builds on previous sales data
• or different filtering, faceted search solutions that again use pre-defined attributes for refining the search.

Image Similarity Search empowers users by showing visually similar products instantly, which can seamlessly guide them through the product catalogue without requiring shoppers to understand the logic behind the structure of the product catalogue.

We are unaware of any other search provider offering this unique capability.

Search Engine relevance Search UX Merchandising and Brand Promotions Merchandising campaign analytics Brand analytics Visual product discovery optimization Expanding our synonym database specific to languages and industries Improved intent refinement recommendations Autocomplete hover product preview

We have implemented the Image Similarity Search as a new feature of the Prefixbox Product Discovery Suite for Praktiker, which is a leading DIY store chain in Hungary. On a yearly average, the site has 600,000 visitors, 6.5 million page loads, and 800,000 searches per month.

Our aim with this new feature was to make it easier for shoppers to browse through the product catalogue and find what they are looking for even without knowing the actual industry terminology.

The first results of the A/B tests show quick engagement rates of the new feature: after clicking the “Show similar products button” the users quickly understood the function and kept on using it multiple times to scan through related products. On average, they have found the requested product in 3-4 steps, without having to stop the shopping journey to set filters or being misled by an incorrect search term.

The implementation phase only started at the beginning of November, but we can already see the increasing engagement rates of this new feature and its impact on better user experience and higher conversion rates.

The pilot version of the Image Similarity Search was first implemented in the webshop of the Praktiker (a DIY store). While this innovative feature is brand new for shoppers, their engagement increased very quickly, showing that they were curious to try this new feature, understood how it works and were satisfied with the search results.

The initial results of the A/B testing on the search results pages show that a slightly higher amount of people (13.52% of users) clicked on the “Show similar products” button as who clicked on “Related Products” (12.35% of users), the other popular feature for refining product search.

“We are thrilled to see the first results of the Image Similarity Search. As a DIY store, we have a wide customer base ranging from tradespeople to private households, and we are striving to make our products easy to find for all of them. This new innovation gives our users the tool to scan through our products effortlessly, significantly increasing the shopper experience that can be quantified in higher conversion rates.” Szilvia Burka - Marketing Director and Board Member at Praktiker

https://www.prefixbox.com/de/