{"id":55922,"date":"2025-01-29T20:55:39","date_gmt":"2025-01-29T20:55:39","guid":{"rendered":"https:\/\/www.bloomreach.com\/?post_type=library&p=55922"},"modified":"2025-02-04T01:35:41","modified_gmt":"2025-02-04T01:35:41","slug":"dont-build-your-search-strategy-on-lies-and-misinformation","status":"publish","type":"library","link":"https:\/\/www.bloomreach.com\/en\/blog\/dont-build-your-search-strategy-on-lies-and-misinformation","title":{"rendered":"Don\u2019t Build Your Search Strategy on Lies and Misinformation"},"content":{"rendered":"\n

There\u2019s been plenty of confusion and misinformation in the market around ecommerce search. Vendors make bold claims about which search technologies are on the way out and which ones are \u201cinnovative,\u201d but these claims are often misleading or outright false. A lot of technical jargon also gets thrown around, leading to confusion about what\u2019s current and what\u2019s not. <\/p>\n\n\n\n

Ultimately, customers are just looking for ways to drive real outcomes \u2014 revenue per visitor and conversion \u2014 and ensure that they\u2019re future-proofing their search investments. That\u2019s why I\u2019m here to dispel some myths around ecommerce search and share insights into where I believe the future of search is headed. <\/p>\n\n\n\n

Claim 1: Simply Having AI in Your Search Is Enough<\/h2>\n\n\n\n

Many vendors tout the latest AI technologies as their claim on why they are more modern than others. However, with how quickly AI has evolved, not all AI is built the same<\/strong> \u2014 search solutions created without next-generation AI<\/a> simply can\u2019t keep up with modern expectations. <\/p>\n\n\n\n

At Bloomreach, we have upgraded the Discovery core with GenAI including models from Google, OpenAI, and NVIDIA, which allows us to serve any queries \u2014 from very precise searches to natural language searches \u2014 and deliver unprecedented relevance.  <\/p>\n\n\n\n

Plus, we\u2019re constantly fine-tuning our AI with the latest advancements. If you\u2019re talking to a vendor who hasn\u2019t updated their search core in the last few months (or even weeks), they\u2019re already way behind. Without advanced AI built into the very foundation of their products, you\u2019ll need nothing short of a complete rearchitecture to keep customers coming back.<\/p>\n\n\n\n

Of course, the proof is in the numbers. We\u2019d be more than happy to put our search solutions to the test and prove to you<\/a> that our technology is the best in the market.<\/p>\n\n\n\n

Claim 2: Clickstream-Based Machine Learning Is a New Innovation <\/h2>\n\n\n\n

Clickstream-based machine learning (ML) analyzes a user\u2019s on-site behavior to personalize their search experience. We\u2019ve heard some recent claims about this being a new or pioneering innovation. Honestly, of all the claims we\u2019ve heard floating around, this one is by far the most egregious and ridiculous. <\/p>\n\n\n\n

Is the technology effective? Absolutely. Is it new? Absolutely not. <\/p>\n\n\n\n

It\u2019s funny \u2014 we must have missed the memo of others being pioneers of this technology when we literally patented* it years ago (starting in 2012)! See the footnote below for a list of relevant patents. Certainly not a leapfrog of innovation, but glad to hear others are playing catch-up!<\/p>\n\n\n\n

Bloomreach Discovery uses clickstream data (among many other data sources) and has long enabled features like personalized search results and tailored product rankings based on user behavior.<\/p>\n\n\n\n

\"Bloomreach<\/figure>\n\n\n\n

Claim 3: The Future Is Transformer-Powered Product Discovery <\/h2>\n\n\n\n

Finally, you\u2019ll hear vendors tout transformers as the future of product discovery. If that\u2019s the case, then I guess we\u2019ve been living in the future, as we\u2019ve been utilizing transformers in Bloomreach Discovery since 2022! <\/p>\n\n\n\n

Built on proprietary technology, our search can discern nuanced differences like \u201cmilk chocolate\u201d versus \u201cchocolate milk\u201d or \u201cdress shirt\u201d versus \u201cshirt dress.\u201d We augment this with transformers<\/a>, including BERT (bi-directional encoder representations from transformers) for natural language processing and ViT and DETR for visual search capabilities.<\/p>\n\n\n\n

BERT analyzes words in context, enabling our search to grasp nuances, idiomatic expressions, and subtleties across languages. Fine-tuned with over 15 years of commerce-specific semantic data, our solution excels in delivering precise, context-aware results tailored for ecommerce across a wide range of languages<\/a>.<\/p>\n\n\n\n

We also can\u2019t forget that large language models (LLMs) are transformer models. Bloomreach Discovery leverages LLMs in search recall to ensure all potential candidates are considered during recall generation, as well as for search optimization to refine the final result set. LLMs of course can be applied in myriad areas to improve both search ranking and relevance, and we continue to build on our innovation in this area. We would love to share our roadmap with you \u2014 just give us a holler!<\/p>\n\n\n\n

The Future of Search<\/h2>\n\n\n\n

Let\u2019s move beyond falsehoods and focus on what truly lies ahead for ecommerce search:<\/p>\n\n\n\n

Search Will Become More Conversational  <\/h3>\n\n\n\n

Shoppers will increasingly demand natural, personalized experiences. Conversational shopping agents like Bloomreach Clarity<\/a> will revolutionize ecommerce by scaling the personal touch of in-store sales associates to the digital realm.<\/p>\n\n\n\n

\n