Search best practices

Independent research from the Baymard Institute found that 56% of ecommerce sites have mediocre or worse search. We use their published findings as our checklist, and Algolia AI Search gives you the tools to meet them.

Source: Baymard Institute, ecommerce search UX benchmark 2026

12%

of sites have issues with this

“Keurig K45 Elite”

With Algolia: Exact matches rank first, and typo tolerance still finds the product when the model number is mistyped.

20%

of sites have issues with this

“sandals”

With Algolia: Query categorisation recognises the category and sends shoppers straight to the right range.

39%

of sites have issues with this

“leather jacket”

With Algolia: Attributes like colour, size, material and brand are matched to filters, so results only show what fits.

43%

of sites have issues with this

“wedding gift”

With Algolia: Semantic AI search understands intent, and merchandising rules let you curate results for occasions.

44%

of sites have issues with this

“Dell laptop adapter”

With Algolia: Compatibility data is indexed and searchable, so shoppers see parts and accessories that fit.

37%

of sites have issues with this

“stained rug”

With Algolia: Semantic search and synonyms connect a problem to the products that solve it.

54%

of sites have issues with this

“13in laptop sleeve”

With Algolia: Synonyms (in = inch = ″) and AI-suggested synonyms from your own search data.

66%

of sites have issues with this

“return policy”

With Algolia: One search box across products, help pages and policies, with each result type clearly labelled.

Autocomplete

80% of ecommerce sites offer autocomplete, but only 19% get all the details right. And 69% don’t give useful suggestions when a shopper makes a small typo. (Baymard, Baymard benchmark)

✓  Keep the list short: no more than 10 suggestions on desktop and 4–8 on mobile.

✓  Tell suggestions apart: category suggestions styled differently from search suggestions.

✓  Highlight what’s new: emphasise the predicted part, not what the shopper already typed.

✓  Forgive typos: Algolia’s typo tolerance keeps suggestions relevant when a word is misspelled.

✓  Work with the keyboard: arrow keys and Enter to pick a suggestion, with the choice copied into the search box.

✓  Suggest real searches: Query Suggestions built from what your shoppers actually search for.

Results pages

Baymard’s research also covers results logic, “no results” pages, layout and filtering. These are the areas we check and configure for you. (Baymard ecommerce search research)

A “no results” page is a dead end that sends shoppers elsewhere. Algolia analytics show every search that returns nothing, so we can fix it with synonyms, rules or redirects.

Filters change with the search: “sofa” shows size and material, “TV” shows screen size. Algolia’s dynamic faceting does this automatically.

Results ranked by relevance and what sells, with your merchandising rules on top, so the best match is in the first row.

The search term stays in the box and the results stay in place when shoppers go back, so they never start over.

1

We test your search against the Baymard query types and autocomplete practices above, using real searches on your store.

2

You get a simple scorecard: what passes, what fails, and which failures cost you the most sales.

3

We show how Algolia would fix each gap, and what that could be worth to you.