Audit your store's search in 20 minutes: 15 checks

An online store's search is one of the few things you can evaluate yourself, today, without installing anything and without booking an hour of anyone's time. You just need to know what to test.

This is the list we use when someone shows us their store. Fifteen checks, twenty minutes. It works for any store, with any search engine and with none.

Grab a piece of paper and note the ones that fail. At the end I'll tell you what to do with the number.

Before you start

Two rules, or the result is worthless.

Do it on a phone. That's where most of your customers search, and where everything the desktop hides becomes visible.

Don't search the way you search. You know the exact names of your products. Your customers don't. That bias is the reason almost everyone believes their search works fine.

Finding the search box (1-2)

1. Can you see it without looking for it? Open the home page on a phone. Is there a visible search box, or a magnifier that opens a menu that opens something else? Every extra tap between the customer and search costs you customers who already knew what they wanted.

2. Is it everywhere? Open a product page and a category page. Search has to still be there. The moment people search most is right after looking at something that wasn't it.

How people actually type (3-8)

3. Misspell a word on purpose. Change a letter, drop one, double another. If the results vanish, you have a big problem: on a phone, with a small keyboard and in a hurry, typos aren't the exception.

4. Drop the accents. Search your best seller without any accents. And the other way round: add one where there isn't one. Plenty of people skip accents deliberately because it's faster.

5. Singular and plural. Search "t-shirt" and "t-shirts". They should return the same thing. If they don't, your search is comparing strings of text, not words.

6. Two words. Search an adjective and a product together, something like "folding table". See whether it respects both or hands you every table you sell. Lots of results is not a good sign: it's the polite way of finding nothing.

7. Put the important bit last. Search three words where the last one is the most specific: a size, a measurement, an unusual colour. If the results ignore the last word, your search is silently discarding exactly what mattered most to the customer.

8. Use the everyday name. The one your customers say on the phone, not the one in your catalogue. "For back pain" instead of the brand name. "Cordless vacuum" if you list it as "stick vacuum cleaner". This is where search engines that look for words part company with the ones that look for meaning.

Your catalogue from the inside (9-11)

9. Search by SKU. Take a code from your catalogue and type it. Then take out the hyphens. Then take out the leading zeros. If your customers buy by code — and in B2B they always do — each of those three forms has to land in the same place.

10. Search a brand you carry. Its products should come up, and not the competitor's products that mention it in their description.

11. Search for something you don't stock. A product you know you don't sell. The right behaviour is to tell you there's nothing and offer a way out. What it must not do is hand you anything at all just to cover the gap.

The results (12-13)

12. See whether you can filter. Search something broad, the kind that returns a hundred results. Are there filters for price, brand or size right there? Do the numbers on each filter match what you get when you click? A counter that promises twelve and delivers four is worse than no counter.

13. Look at what each result shows. Photo, name, price and availability, without opening anything. If you have to open the product page to learn the price, you're asking for one click per product from someone comparing five.

What you can't see by looking (14-15)

These two are the ones almost nobody does, and the ones that move the most money.

14. Does the URL change when you search? Run a search and look at the address bar. If the address doesn't change — because results open in a layer over the page — there's a good chance your analytics is recording no searches at all. You'll see an empty search report and conclude nobody uses your search, when in fact you aren't measuring it. It's a measurement failure, not a behaviour one, and it leads to expensive decisions.

15. Open your analytics and find two numbers. How many sessions use search, and what happens next. In a normal store, people who search buy considerably more than people who browse: that's the real reason any of this matters. If you can't answer, check 14 tells you why.

What to do with your score

0-3 failures: your search is fine. Spend the effort elsewhere, genuinely.

4-7 failures: most of them are fixed by configuring what you already have. Synonyms for point 8, reviewing the minimum word length and field weights for 3 and 9, and switching on search tracking for 14. You don't need to change tools.

8 or more: it isn't configuration, it's that your platform's search doesn't stretch to your catalogue. That usually starts at a few thousand SKUs, with several languages, or when your customers search by code.

And when the problem isn't search

This is the part nobody selling search engines will tell you.

If your product pages are empty, no search engine will save you. A product with a name and two lines can't be found by what it does, because that information doesn't exist anywhere. Before changing search engine, write the product pages.

If the problem is that you don't stock what they're asking for, search is only showing you that. You don't fix it by buying software: you fix it with your catalogue, and the list of what people search for and you don't have is the best buying guide you'll ever get.

If you have fewer than two hundred products and your customers know them, search adds little. Good category navigation will do more for you.

And if your traffic is low, no change to search will move your sales noticeably. Traffic first.

If you want a second opinion

Send us your store and we'll run these fifteen checks ourselves, against your real catalogue, and tell you what fails and what can be fixed without paying us anything. No need to buy: write to us with the URL.

If you then want to try it with your own products, there's a free 1-month trial.