Why AI-powered search engines are the future


One of the most important characteristics of a robust ecommerce site, whether in the promotional product industry or beyond, is the quality of its search function. After all, if you can’t find what you’re looking for, how can you make a purchase? Hamish Ogilvy, CEO and co-founder of Search.io, a company that uses machine learning to improve internet search results, explains the power of high-quality on-site search engines.

Hamish Ogilvy, CEO and Co-Founder of Search.io

Q: Why is an on-site search feature so important for e-commerce?

A: Ecommerce site research has the power to improve the visitor experience, retain customers, and increase on-site conversion rates. It actually has a disproportionate influence on the success of buyers and sellers. Visitors who use search can generate around 30-60% of all e-commerce site revenue, research shows, and on-site searchers convert 1.8 times more than non-searchers.

On-site conversion rates hover around 3% industry-wide, but Amazon.com enjoys a conversion rate that is five times the industry average (even higher for Prime members). Research is vital in the world’s largest market to find anything, so naturally Amazon has invested heavily in search engineering for 20 years.

The good news for brands and retailers is that with the availability of new standard search solutions, they don’t need to invest as much money or resources to develop their own capabilities.

Q: How does a fast and accurate search improve the online customer experience?

A: Online retailers know they have about 15 seconds on average to build customer loyalty before they bounce off an ecommerce site. Buyers do research, but most site search solutions don’t work the way people would expect.

Amazon did a study in which it demonstrated that every 100 millisecond lag in a website’s response time costs millions of dollars in lost revenue.

Speed ​​and accuracy are important to visitors. I should also note that this includes mobile buyers. More than half of e-commerce customers buy through phones where the user experience is different, but vitally important.

Question: Your site mentions that many brands still rely on 20-year-old search technology. Why is it so restrictive for them? What are the issues with the old search features?

A: Search engines can’t look at something the same way people do. It is difficult for a search engine to understand the user’s intention. For example, if someone searches your site for a “shower curtain rod,” a search engine examines every word in the query and must decide where to send you. Older search engine technology that only uses keywords may see “shower” and send you to the wrong place. These older technologies force site owners to write detailed rules to overcome these challenges.

For a retailer, that means investing hundreds of hours of writing rules, synonyms, and technical hacks to make simple searches work. And the job is never done. When you make changes to your product catalog or add new products, you need to write more rules. Because it’s so time consuming and complex, most e-merchants end up focusing on the top 20% of their catalog, leaving the remaining 80% unoptimized.

Q: Can you explain what an artificially intelligent search engine is and what it can do compared to traditional search functions?

A: There are two things AI is doing right now to help online sellers. One provides more relevant results, and the other reclassifies the results based on the data.

AI-powered search engines understand context and intent. They operate most of the time without human intervention (definition of rules and exceptions). Instead, they “learn” from search trends and user interactions on the site.

For online sellers, this translates to a lot less work on their part. You don’t need to fill your product pages with keywords, synonyms, tags, and metadata. The search results are working.

This resolves relevance. But a lot of products can be relevant, so in which order should you rank them? This is where machine learning can also help by using signals to rank products in the order most likely to convert. We may use data such as clicks, registrations, additions to cart, purchases, etc. It may also differ by region.

There are a lot of factors to consider, and this is where AI can really help. Ultimately, we give retailers control over deciding what’s most important to their business so they can configure the results to work for them.

Q: Can you give some examples of how AI interprets the meaning of search queries to deliver specific results?

A: Sure. A real world example is a company we spoke with that sells seasonal products – in this case it was Halloween clothing and costumes. If a customer is looking for, for example, Halloween socks, they will need to enter “Halloween socks” to find them. But with the AI-based research, they can type in “scary socks” or “horror socks” or something like that and get the same great results. Try this with a normal search engine; it will not work. The company didn’t have to do anything to activate the feature, as the AI ​​understands the concept of Halloween and automatically associates it with “horror” or “scary”.

Q: Would you like to add anything about the future of online research?

A: Consumers are increasingly sophisticated in their shopping habits and their demands for great shopping experiences. For us, research is really the starting point. Research can be harnessed to provide the kinds of interactions that were unthinkable just a few years ago. Items like personalization, merchandising, and better recommendations can all be provided through a better search experience. Often times, people don’t realize that collection pages and product filter pages (where there is actually no user query) are powered by search technology as well.


Rosemary S. Bishop