To fully recognize the promise of generative AI, companies will have to invest in enriching its data and training the Large Language Models (LLM) that power genAI. This investment takes time and resources, and it directly affects how effective your AI will be. However, there is a more immediate and impactful opportunity that genAI could address today, with very minimal investment. An issue that could be costing you as much as 3-5% of sales revenue.
What is the issue?
The technology behind the search bar that drives your e-commerce site is likely returning unhelpful search results almost a third of the time, even if you may have the very items in inventory that your customers are looking for. Your search may be returning zero, low, or irrelevant (ZLI) results. For many product categories, searches using brand, model, names, and keywords make up the majority of your customers’ initial queries. One, single character–a misspelling in one, single character—will fail to return useful results 34% of the time on the top ecommerce sites.
Any time your customer searches for something and you return a “sorry, we can’t find that,” or the very unhelpful, “0 results for your search,” you’ve missed that revenue opportunity from their original intent to buy. Even though you may have dozens of items that match their original request, your search might be returning just a handful of results. Or worse yet, you could be showing your customers results that have nothing to do with their original query. Sorry, we couldn’t find that insulated tent you were looking for, but how about this sleeping pad or this thermal underwear instead? Yes, that was an actual search experience I recently had. And, yes, they actually did have insulated tents if you knew the exact brand and model name or browsed through every tent they had to find the ones that were insulated.
How much is this really impacting my revenue?
A lot. You can use Engenai’s ZLI Calculator to see how much revenue your business might be losing. By default, the calculator uses industry averages and statistics from the top e-commerce sites, so all you need to do to get started is enter your monthly website traffic. The Engenai ZLI calculator does the rest.
You can refine the calculations by using the optional advanced search fields with numbers specific to your business. If you don’t have all the numbers, that’s okay. Use the ones you have and the calculator will use the default averages.
Are you sure you’re not exaggerating these numbers?
Nope. In fact, we’ve used the most conservative industry figures. The problem is, your Key Performance Indicators (KPIs) and standard business insights reports don’t usually capture these metrics because this is customer attrition that is happening as a result of poor search results. Your customers don’t tell you they’re leaving because they couldn’t find that item that you actually had, but they didn’t know how to get to it. They just leave, and most never come back. These numbers get lost in your overall abandoned carts, bounce, and exit rates.
When you lose customers who have a poor search experience, you lose a critical and extremely valuable customer segment. Customers who use site search are 2-3x more likely to make a purchase and spend 2.6x more than customers who don’t. Over $300B in US sales are lost each year because customers can’t find what they’re looking for. Some other important stats to consider:
- 43% of retail site visitors go straight to the search box
- 31% of top commerce site searches end in unhelpful search results
- 68% of customers who leave because of a poor search experience will not return
- 12% of customers who leave will go to a competitor’s site
So you aren’t alone in this. Even market leaders face these challenges with search intelligence. After you factor in search conversion rate, average order value, and exit rates, the average e-commerce site may be leaving 3-5% in direct and indirect revenue on the table. With global e-commerce CAGR projected to be between 12-15% , this means that you could be cutting your potential growth by almost one-third. And losing this revenue will continue until you improve your search. What’s worse, you may be handing customers to your competitors with better search intelligence.
How can I fix it?
There are practical, actionable steps you can take right now. Even a basic genAI model can address some of the main causes of ZLI that result from a single character, word, or phrase. These are things like: pluralization, spaces, alternative spellings, dialect differences, and synonyms. We’ll be discussing these issues in more detail in a future post. Without requiring a massive amount of data or extensive training of the AI, you can get started with your existing product data. The risk for you is minimal because you already show most of this product info on your website. Building a solution for yourself with a specific LLM and expensive GPU hardware will probably take you months and require a significant investment in engineering and infrastructure costs. Or, you can leverage Engenai’s LLM Ops to begin improving your site search in days with minimal upfront infrastructure costs that you can scale up as your business grows.
