Insights / Article / Conversion rate optimisation (CRO)
ArticleSite search that sells
Visitors who use site search often have strong intent to buy. Poor search wastes it. Handle real language, offer useful filters and learn from every failed query.
A visitor who types into your search box is telling you exactly what they want. They are usually further along in their decision than someone browsing category pages, and in many online shops searchers make up a disproportionate share of buyers. Yet site search is often one of the least maintained parts of an ecommerce site, configured at launch and then left alone.
When search fails, it fails silently. A customer searches for “trainers”, your catalogue calls them “sneakers”, the results page is empty and the customer leaves. Nobody complains. The sale is simply lost. This article sets out how to make search work as hard as your best salesperson.
Why search deserves attention
Look at your own analytics. Compare the conversion rate of sessions that use search with those that do not, and check what share of revenue involves a search. For many retailers and distributors the numbers are striking enough to justify sustained investment.
Then look at the failures. How many searches return no results? How many return results that nobody clicks? Each of those is a visitor with intent that your site could not satisfy. For B2B sellers with large technical catalogues, the effect is often greater still, because buyers search by part numbers, specifications and their own internal descriptions.
Handle misspellings and synonyms
People search in their own words, not yours. They misspell, abbreviate, use regional terms and describe products by what they do rather than what they are called. A good search engine bridges that gap.
- Tolerate typing errors, so “sofa bedd” still finds sofa beds.
- Maintain a synonym list that maps customer language to catalogue terms: trainers and sneakers, hoover and vacuum cleaner, settee and sofa.
- Recognise part numbers and codes with or without spaces, dashes and leading zeros.
- Handle singular and plural forms, and common abbreviations.
- Understand descriptive searches such as “waterproof jacket for walking” by matching on product attributes.
Most modern search tools, whether built into your ecommerce platform or provided by a specialist service, support these capabilities. The difference between good and poor search is usually not the technology but whether anyone configures and maintains it. Synonym lists, in particular, need a named owner who updates them regularly from real search data.
Autocomplete is part of the same job. Suggestions that appear as the customer types guide them towards terms your catalogue understands and show popular products before they have finished. Check that suggestions reflect what is in stock and avoid offering phrases that lead to empty results pages, which undermines trust faster than having no suggestions at all.
Show useful filters
Once search returns results, filters help customers narrow them down. The best filters reflect how customers actually choose, which is not always how products are organised internally.
A clothing shopper filters by size, colour and price. A buyer of electrical components filters by voltage, connector type and stock availability. A furniture shopper may care about dimensions and delivery time. Talk to customer service and sales teams about the questions customers ask most, and make those the first filters shown.
Filters depend on clean product data. If half your products have no colour recorded, a colour filter will hide them. Before adding a filter, check that the underlying attribute is filled in consistently across the range, and assign someone to keep it that way as new products are added.
Show only relevant filters
Filters should adapt to the results. A search for paint should offer finish and colour, not shoe size. Showing every possible filter on every search overwhelms customers and suggests the site does not understand what they want.
Learn from searches that fail
Every search that returns nothing, or returns results nobody clicks, is a lesson. Export a list of the most common failed searches each week and work through it. Typically, each falls into one of four groups.
- A synonym or spelling the engine does not recognise: add it to the synonym list.
- A product you sell but describe differently: improve the product title, description or attributes.
- A product you do not sell: pass the demand to buyers, and consider showing a helpful alternative.
- An information request such as delivery, returns or opening hours: link those searches to the right help page.
This weekly routine takes perhaps an hour and consistently produces some of the highest-value improvements available to an ecommerce team. It also generates useful commercial insight. Searches for products you do not stock are a free, continuous market research feed.
Use search as a merchandising tool
The order of search results matters as much as whether they appear. By default, many engines rank purely by text relevance. Better results come from blending relevance with commercial factors: stock availability, popularity, margin and customer ratings. Out-of-stock items should drop down or disappear, depending on whether you can take back orders.
For the most common searches, review the first page of results manually. Are the products shown the ones you would recommend if a customer asked in person? If not, adjust the ranking rules or pin specific products to the top. Treat your top fifty searches with the same care as your home page, because for many customers they are the real entry point to your range.
We reviewed our zero-result searches for the first time and found our customers were searching for a product range we had renamed two years earlier. Fixing that took an afternoon.
Ecommerce manager, home improvement client
A practical starting plan
This week, pull three numbers from your analytics: the share of sessions that use search, the conversion rate of searchers compared with non-searchers, and the rate of searches returning no results. Then export your top fifty searches and your top fifty failed searches. Fix the failures, check the first page of results for the most common queries and appoint an owner for search. Repeat the review weekly, and track the zero-result rate as it falls.
Source: Intelliversal analysis.
