Free Subscription

  • Access 15 free news articles each month


Try one month for $5
  • Unlimited access to news,insights and opinions
  • Quarterly and weekly magazines
  • Independent research reports and forecasts
  • Quarterly webinars with industry experts
  • Q&A with retail leaders
  • Career advice
  • Exclusive Masterclass access. Part of Retail Week 2021

How AI will increase your search ROI

(Source: Supplied)

Your on-site search bar is one of the most popular ways for customers to find what they are looking for on your site. It also offers conversion rates 1.8x higher than browsing alone! However, most e-commerce businesses treat search as an infrastructure cost like web hosting. 

Friends, retailers, countrymen… lend me your ears… if there’s one myth to dispel, it’s that search is “just another website feature” or “sunk cost.” The search box may seem innocuous, but it can be a sales and conversion powerhouse! increased search-related conversions by more than 40 per cent using AI-powered search and search A/B testing. For a business that’s heavily search-focused, better on-site discovery can be a massive revenue win.

Perhaps one of the reasons that search has been labelled an “infrastructure cost” is due to its seeming complexity. It’s true that search engines contain complex technologies. However, modern search-as-a-service companies have lowered the barriers to search adoption and optimisation. Non-engineers can now adjust, test, and improve results themselves.

In this article, I’m going to touch on three ways AI can boost on-site search ROI without requiring a massive investment in IT. But first, let’s briefly look at how search works.

The challenge for retailers

If someone searches for “bank” do they mean a financial institution, how a plane turns, or the side of a river? It’s likely they mean the financial institution. Search engine developers have added features – ones that most retailers are probably familiar with – to handle these kinds of ambiguities. 

Some of these features include:

  • Rules, or if/else statements to tell search engines how to prioritise results; for example, you might create a rule that a search for “fleece” should return jackets, sweaters, hoodies, and blankets in that order.
  • Synonyms to explain that “parka,” “jacket,” and “coat” mean the same thing.
  • Keywords and metadata such as tags to enrich the data.
  • Natural language processing to parse longer and more complex queries like “queen sized cotton fitted mattress sheets”.

The difficulty with all the above – as you probably already experienced – is that retailers must invest days upon days writing hundreds of rules, synonyms, keywords, and hacks to get the right results boosted at the top of the page. Plus, they need to spend endless hours optimising and maintaining these attributes.

No wonder site search is considered an expensive infrastructure cost!

It’s also not uncommon for us to speak with companies that have invested huge amounts of time and money in open source search projects that required teams of engineers and data scientists to optimise. Smaller companies don’t have the resources to afford this kind of investment, and larger companies shouldn’t need to anymore.

Fortunately, these kinds of complexities are soon to become a thing of the past.

Machine learning is ready

Up to now, you’d be right to be sceptical of “artificial intelligence” claims. AI has been touted as being able to help with these kinds of problems, but in reality, it hasn’t greatly affected e-commerce site search.

However, newer machine learning technologies being released today can greatly improve results in several ways while also massively reducing the work you need to put in.

Here are just three ways this newer generation of AI will help:

No more synonyms: When it comes to relevance, modern AI understands the meaning and context of a keyword search to deliver great results – without the need for rules, synonyms, or other search engine “hacks.” Whether a customer searches for a “jacket” or “overcoat”, they’ll get great results. You won’t need to keep adding or changing synonyms or writing rules to remind your search engine exactly what you need.

Self-optimising results: Secondly, as a retailer, you likely want to rank higher converting products (or higher-margin items, higher inventory products, or other kinds of results at the top). Newer AI, such as reinforcement learning, can optimise results automatically for whatever criteria is most important to your business. This means you’re not just displaying the most relevant results, but also the ones that are more likely to turn visitors into buyers.

No more keywords: Lastly – and perhaps most startlingly – with modern AI models, you don’t even need keywords! For example, if you sell “Apple AirPods” but customers are searching for “wireless headphones.” The keywords “wireless headphones” don’t need to be on your product landing pages for the search to work. Seems like magic, but it’s really just AI.

Baymard Research has noted time and again the poor on-site search performance on some of the world’s largest online retailers due to poor query matching. We feel confident that these kinds of problems will soon be relegated to the past, generating search ROI faster and more easily, while freeing up retailers to spend people resources wherever they most need them.

About the author: Jon Silvers is head of marketing for which offers e-commerce site search powered by Neuralsearch