Optimizing on-site search using machine learning

Pushed by customer expectations, merchandizers can no longer ignore the need for on-site search optimization. Fortunately, machine learning now enables what was impossible before: genuine optimization of search results on an individual level and, subsequently, a revenue uplift.


The importance of on-site search

In an attempt to avoid information overload, customers have reclaimed power, demanding excellent, seamless online customer experiences. Our partner Bloomreach calls this the classic Amazon- or Google- expectation problem: by constantly improving their online customer experiences, they have set a high standard. Failure to comply will send customers running for the hills. Or rather, to a website that does meet their requirements.

How are these excellent experiences shaped, you ask? By personalizing marketing content. And this is where on-site search comes into play.

To watch this video you need to accept marketing cookies

When people visit a website, they consciously and unconsciously express their purchase intent, thus providing endless amounts of information regarding their interests. For instance, they may enter a term in the search box, click on a product and add or remove items from the shopping basket. Understanding the search query and subsequently presenting the visitor with relevant products is crucial to customer satisfaction.

But on-site search does more than providing a good online customer experience. A merchandiser might decide to push products with a low rate of return or a high margin, placing these higher in the search results. Similarly, they may limit visibility of products that are short in supply.

This makes on-site search a splendid tool to balance customers’ intent and companies’ goals, as optimization will almost definitely result in increased revenue and margins.

On-site search optimization is hard

Unfortunately, the optimization of your e-commerce site search does not come easy. In fact, Internet Retailer surveyed 119 retailers and found that 39 percent of e-commerce companies struggle with relevance and show their customers poor results. Meanwhile, Accenture’s Personalization Pulse Check confirms that “40 percent of consumers have left a business’s website and made a purchase elsewhere because they were overwhelmed by too many options”.

Optimizing on-site search using machine learning - Statistics - by Accenture

No scalability options

This is no surprise. After all, there is only so much you can do manually. Surely, companies will focus on best-selling products and optimize frequently-used search terms. ‘Knee-long skirt‘ should not be a problem. But how about smaller product categories or more detailed search queries? ‘Knee-long red skirt with roses’, for example? On-site search optimization is notoriously time-consuming. As a result, long-tail search queries such as the above are left untouched, rendering a truly relevant experience for all visitors – a.k.a. ‘scaling up’ – impossible.

Suboptimal search engine performance

But limited capacity is not the only problem. Most search engines don’t understand language, but merely process words. The result: a customer searching for a shirt dress will be presented with this type of clothing but might also be bothered by results showing dress shirts – entirely different attire. Needless to say, this failure to understand customer intent will lead to suboptimal customer experiences and to equally disappointing conversion rates, revenue and margins.

Optimizing on-site search using machine learning - In-article image - by Accenture

It’s costly

On top of all this, we see many cases in which the product descriptions in the company’s systems keep engines from producing relevant results. As John Klein of LiveArea said in a Bloomreach blog: “If a car runs on gasoline, a search engine runs on data.” Theoretically, it is possible to do the product information upload, tagging and rule-writing manually. But in reality, there are not enough staff members in the world to make this happen. Unless your pockets are deep.

If you wish to create the optimal shopping experience that customers have come to expect, as well as pursue your company goals, then focus on machine learning.

If a car runs on gasoline, a search engine runs on data.

This is where machine learning comes into play

Machine learning means that the search engine will improve itself continuously by correlating the search term with the on-site behavior and optimizing accordingly.

AI-powered machine learning:

  • Applies a profound understanding of language;
  • Can surface the right product to each visitor, even if the possible visitor/product combinations are countless;
  • Is able to ensure high-quality data in the catalog.

Furthermore, machine learning platforms such as the one offered by Bloomreach have algorithms in place to optimize margin, revenue, stock levels, and much more. All a click away.

By applying AI in on-site search, we can generate a lot of business from the long tail, thus scaling up perpetually. Start from scratch and e-commerce sites may increase their revenue with well over 30 percent. Even retailers who already perform well look at possible increases of 5 to 15 percent.

Doesn’t machine learning make my employees redundant?

But what about human employment? Will your staff be replaced by machines? While you can focus entirely on machine optimization, the best results will be achieved when there is a good balance between AI and human curation.

You see, an algorithm will not be aware that a celebrity was spotted with a certain outfit, unleashing an online buying frenzy. A human response will remain necessary to direct these visitors to the right clothing.

Furthermore, we strongly believe that by allowing machines to do the heavy lifting of manual optimization, online merchandisers can focus on tasks that require human creativity. No, employees are not at risk of losing their jobs, but their job descriptions will change – for the better, we might add!

Optimizing on-site search using machine learning - Statistics - by Accenture

Where to begin with your on-site search optimization?

The first thing to do when you opt for site search optimization is to replace the current, last-generation search engine by an AI-powered one. Also, product descriptions will need to be optimized and systems will have to be put in place to feed product data to the search engine properly. Only then will you be able to optimize your on-site shopping experience and harvest results.

But implementing technology is not all you need to do in order to succeed. Business operations will need to adapt and staff will have to grow into their new roles. Therefore, on-site search optimization is always considered to be part of a larger digital transformation. This is exactly where Bloomreach and Accenture together can make a difference: a partnership that offers a state-of-the-art machine learning platform, as well as the essential services to truly transform your business.

On-site search optimization - Be a leader - by Accenture

Take your future into your own hands

It is time to face the truth: how many people leave your site because of poor search results? How much revenue uplift could you create by implementing modern on-site search tools? And do you choose to lag behind your competition, or to take the lead? Because ultimately, AI-powered on-site search optimization is the only way to improve your customers’ shopping experience and increase revenue.


This article has been produced in close collaboration with Jasper de Vreugt, Director of Channel Sales EMEA at Bloomreach

Author: Mehmet Olmez