• May 30 2024

Multi-Search: A Game-Changer in E-Commerce

Brendan O’Shaughnessy, Chief Commercial Officer

Welcome to the second in a series of blog posts about our new multi-modality search and discovery engine, Multi-Search. In this blog, we’ll share some new concepts that we’ve been working on to improve search for shoppers and highlight how we plan to make Search a much more interactive function that can be used in different contexts across e-commerce sites.

To properly assess the benefits of Multi-Search, you have to consider the limitations and challenges of conventional search as it applies to online shopping today. Whether you are looking for a special new outfit for an evening out, a formal occasion, or just something to lounge around in, we all know the limitations of today’s keyword-based text search. Consider that many brands have in the region of 20,000 products, and many retailers have upwards of half a million products, a simple text search will return many results that we, as shoppers, are left to sort, filter, and browse through until we can find the product we’re looking for. It’s not ideal.     

Recently, my wife had a semi-formal work event to attend, and she wanted a new outfit for it –  something elegant, simple, and in a particular color tone. Having visited several of her go-to retail sites, she struggled to find a dress that ticked the boxes. It wasn’t that there weren’t choices available, but finding each one was so time-consuming, and being able to review the selection of options that she did see to make a decision was even more difficult. I’m glad to say my wife eventually found something she liked enough to buy, but the experience felt more difficult and time-consuming than it should have – she might have been quicker in popping down to the local mall.

This online experience of struggling to find a particular or similar product to what you have in mind is quite common. One reason for this is the gap between how our brains work and how typical search engines work. The example of a search for “sleeveless blue evening gown” on the website of your favorite luxury brand is unlikely to generate any results (I’ve tried… nada) because while we understand the concept of the phrase, the search engine only sees individual words and can’t make sense of the phrase.

That’s simply because common keyword search engines don’t have any way to establish the intent of any given search. We have all experienced this frustration when we conduct an online product search and are led to a dead end; for many online stores, this can represent up to 30% of the search journeys and leads shoppers to abandon those websites. 

At ViSenze, we are addressing these problems through a combination of our hybrid search engine and visual recognition models, and our solutions are transforming the quality and relevancy of search results compared to keyword search engines. Researched, designed, and built over the last 18 months, Multi-Search can combine different inputs such as images and text – to deliver much better results based on what you’re looking for. This isn’t just an improvement; it’s an entirely different way of thinking about search and opens up new ways and placements in which shoppers can use search.  

How else does Multi-Search differ?

Multi-Search offers a better alternative solution for the Search Bar, but it can do much more than that. Let me give you an example of one such placement, ‘out of the [search] box.’

Trigger Search

We’ve already explained how Multi-Search is a multi-modality hybrid search engine, which means you can conduct one search using an image and text together. How would you use this? Well, one way is as a second-step refinement or filter to a typical image search – an example I shared in my first blog on this topic. That was an example of a multi-modality search, but it depended on the consumer having an image of something they wanted to search for. 

But every e-commerce site has hundreds, thousands, and even tens of thousands of professionally shot images that feature products. And up to now, they have not been searchable. Multi-Search’s semantic and multi-modality capabilities come to the fore by making every image on an e-commerce site a searchable asset. For the first time, shoppers can see a product online and search for something similar in another color or be inspired by a product image and use it as a reference to find a matching accessory. This means that we are not limited to trying to adequately describe what we want or fumble about looking for a product SKU number or a product name to reference; we can simply select any image on a retailer’s site and use that to search for a related product or variation. 

We call it Trigger Search, and this is what it looks like;

A simple icon tap on any product image on a PLP or PDP will open a new modal search window comprising the selected image and search bar with prompts. Type in your chosen search term (in this example, the shopper is looking for a similar dress in pink and types “this in pink”) and you’ll quickly be presented with great search results.

This powerful new search tool for retailers is an example of how our multi-modality capabilities are a clear differentiator when compared with other neural or semantic search providers. There are 3 key differences: 

  • Firstly, the natural phrasing of the search term “ this in pink” is supported, delivering a more natural way of searching
  • Secondly, Multi-Search’s multi-modality capabilities can contextualize the search with the selected image and understand that the search phrase refers to the type of dress in the chosen image. As a result, it can understand the shopper’s intent and deliver highly accurate and relevant results.
  • Trigger Search also illustrates how Multi-Search is not limited to the Search bar at the top of the page. It supports contextualized search from anywhere across an e-commerce site, and that’s truly transformational for shoppers and retailers. 

Why Now, and Why ViSenze?

At ViSenze, we’ve been at the forefront of AI visual search and recommendations for over a decade. Our expertise in vector-based search technology, combined with the latest developments in Large Language Models (LLMs), is behind our ability to offer a highly differentiated search engine that better understands consumer needs.

The arrival of these technologies marks a significant shift in consumer expectations. Shoppers can look forward to new and better ways to search and find products faster, more easily, and with better results. For retailers who embrace this next evolution of search, it’s about more sales and increased customer loyalty.

The Future of Multi-Search

This is just the beginning. Tune in for more updates on Multi-Search as we formally launch it at the NRF Asia show in Singapore from June 11-13. We’ll continue to share more exciting new placements and ways to use multi-modality search to help shoppers find better products faster. 

In the meantime, to explore how Multi-Search can transform your e-commerce experience, click below to book a demo.

Join us on this exciting journey and discover how Multi-Search can elevate your online shopping experience to the next level.