Imagine an online store that sells kitchen equipment to cafés and home cooks. A visitor types "expresso machine" into the search box. The page says: "No products found." The store has eleven espresso machines, three of them on sale. The visitor does not try again with a different spelling. They go back to the search engine they came from and open the next shop.
Nobody at the store sees this happen. The analytics dashboard shows a visit that ended after one page. The product team sees healthy stock and a working search box, because when they test it, they type the product name exactly as it is written in the catalogue.
That gap is the whole problem. People who use the search box usually know what they want, which makes them some of the most valuable visitors a store has. When search fails them, the loss is quiet and it repeats every day. The checks below are about closing it.
1. Does search forgive spelling mistakes?
Customers type on phones, in a hurry, often in a second language. "Expresso", "stainles", "blendr" and a missing space are normal input, not edge cases. If the search only matches exact words, each of these ends on an empty page.
Most search engines used by online stores can tolerate one or two wrong letters per word. It is often switched off, or set so strictly that it never helps. Turn it on, then test it with short words as well: a one-letter tolerance on a three-letter word can match the wrong thing, so short words usually need stricter rules than long ones.
2. Does it know the words your customers use?
The catalogue says "French press". The customer types "coffee plunger". The catalogue says "sneakers"; half the visitors write "trainers". These are not spelling errors, so no amount of typo tolerance will fix them.
A synonym list solves this, and it does not need to be large. Start with twenty or thirty pairs taken from what people actually typed (see point 7), plus the obvious regional differences and plural forms. Decide for each pair whether it works both ways. "Mug" should find "cup", but "cup" may not need to show every mug.
3. Can people search by code, size and brand?
Business customers and repeat buyers often search the way they remember the product: by model number, part code, barcode or a size like "28 cm". A search that only reads the product name misses all of these. Codes cause their own trouble: "AB-1200", "AB1200" and "ab 1200" should all lead to the same product.
Make sure the search index includes the fields people really use, and that codes are normalised (case, spaces and dashes removed) both when products are indexed and when a query comes in. For stores selling spare parts or consumables, this one change can matter more than everything else in this list.
4. What does the search actually read?
Search can only find what has been put into its index. It is common to find that only the product title is indexed, while the category name, brand, colour, material and short description are left out. Then "red kettle" finds nothing, even though there are four kettles whose colour field says red.
The opposite mistake also happens: the full long description is indexed with the same weight as the title, so a search for "glass" returns a frying pan because its description mentions a glass lid. Index the useful fields, and give them different weights: title and code first, then brand and category, then attributes, then description.
5. What does the empty result page say?
Sometimes there really is no match. Even then, the page should not be a dead end. Compare two versions:
- "No products found."
- "We couldn't find 'cast iron wok'. Here are our woks and cast iron pans," followed by a few products, the main categories, and a short line on how to ask the store directly.
The second page keeps the visitor in the store. It can also check whether the query matches a category name and show that category instead of an empty list. If the store has a phone or messaging line for orders, the empty result page is one of the best places to show it.
6. Who decides what comes first?
When a search does return results, the order matters almost as much as the match. If a search for "knife" shows a knife sharpener, a knife block and a children's butter knife before any chef's knives, the visitor scrolls, gets tired, and leaves.
Agree on simple rules for ranking: exact title matches before partial matches, products in stock before products that are out of stock, and a small, deliberate boost for products the store wants to promote. Keep promotional rules few and visible to the team. Hidden rules added over the years are the most common reason nobody can explain why a search returns what it returns.
Also check what happens while people type. Suggestions under the search box ("espresso machines", "espresso cups") let visitors reach the right category in two taps, and they quietly fix spelling before the search even runs.
7. Does anyone read the search log?
This is the cheapest check and the one most often skipped. Every store with a search box is collecting a list of exactly what its customers want, in their own words. Most never look at it.
Once a week, export the searches from the last seven days and sort them two ways: the most common queries, and the most common queries with zero results. The second list is a work plan. Each line is either a missing synonym, a missing product field, a spelling the search did not forgive, or a product the store does not sell yet. The last group is useful for purchasing as well.
A short routine is enough:
- Fix the top ten zero-result queries each week.
- Search for each fix yourself on a phone and check the first five results.
- Keep a note of what was changed, so the next person knows why a synonym exists.
A quick test you can run today
Take your phone, open the store as a new visitor and search for ten things: five product names misspelled the way a hurried customer would write them, two product codes written without dashes, two everyday words your catalogue does not use, and one product you do not sell. Count how many searches lead to a useful page. The result is usually a clear picture of where to start.
In short
Site search fails quietly. Make it forgive spelling, teach it your customers' words, let it read codes and attributes, index the right fields with the right weights, turn the empty result page into a way forward, keep ranking rules simple, and read the search log every week.
When we build e-commerce sites, we set up the search index, synonym list and zero-result page before launch, and we review the first weeks of search queries with the store team.
Planning a project like this?
Plan it in 10 steps