PrestaShop search: how to improve it and what to measure

The search that comes with PrestaShop works well in small catalogues and with customers who type the exact product name, but it fails with typos, synonyms, partial references and large catalogues. It is improved in two steps: tuning the native search (index, weights, aliases) and, when that is not enough, moving to a search engine with its own index, self-hosted or in the cloud. The way to know whether you need it is to measure the searches that return no results.

Whoever uses a store's search already knows what they want. They are the visitor with the strongest purchase intent, and every search with no results is often a customer leaving for another store.

How the native search works

PrestaShop builds a word index from each product's name, reference, description, brand, category, tags and features. Each field has a weight, and when searching, the weights of the matching words are added up. From the back office (Shop Parameters, Search) you can:

• Rebuild the index, in full or only for missing products.
• Change the weight of each field, for example giving more importance to the reference than to the description.
• Create aliases: so that a misspelt word or a synonym leads to the right one.
• Enable fuzzy search in the versions that include it, with limited tolerance for typos.
• Adjust the minimum length of words and the list of words to ignore.

Reviewing these settings comes first, and in many stores just reindexing and raising the weight of the reference makes a noticeable difference.

Where it falls short

Typos

Customers type quickly and on their phones. If the search does not tolerate one swapped letter or two misspelt words, it returns zero results for products you do have. Aliases fix it case by case, but you cannot foresee every mistake.

Synonyms and the way customers talk

You call it «sports shoe» and the customer searches for «trainers»; you say «compatible cartridge» and they type «ink». Without synonyms per language, the search only finds your vocabulary.

References and codes

In spare parts, consumables or technical equipment, customers search by reference, sometimes with different hyphens or spaces, or only part of it. A word index handles these codes poorly.

Speed with large catalogues

With tens of thousands of products and many combinations, each search costs the database more, and suggestions as you type arrive late or feel slow. A store that is slow to respond to each letter loses the feel of a responsive search.

Filters and results

The native search returns a list. It does not filter by category, brand or price within the results themselves, and in large catalogues that forces the customer to search several times.

What to measure: searches with no results

Before changing anything, look at what is happening. The most revealing data:

The list of searches with no results is also a purchasing plan: the products people ask you for and you do not have. And an SEO clue, because those are the words your customers use and they may not appear on your product pages.

The alternatives when the native search is not enough

Cloud search engines with a monthly fee

There are external services that index your catalogue on their servers and return very fast results with typo tolerance, synonyms and filters. They work well, but there are two downsides: a monthly fee that never ends, often tied to the number of searches or products, and your catalogue and your customers' searches leave your server.

Self-hosted search engines

The other route is a search engine that runs inside your store, on your server. It can rely on its own index built in the database or on a self-hosted search engine such as Meilisearch, which is open source and designed to respond in milliseconds and tolerate typos. What to bear in mind:

• The server has to be able to handle it. An engine like Meilisearch is one more service running on the server; on basic shared hosting it cannot always be installed.
• What happens if the engine goes down. A good setup falls back automatically to database search so the store is never left without search.
• No fee per search, and the catalogue never leaves your machine.

Common mistakes when improving search

• Changing search engine without looking at the data first. If you do not know what is searched for and not found, you will not be able to check whether the change worked. Keep a couple of weeks of searches before touching anything.
• Filling the aliases with one-off cases. Aliases are for frequent mistakes. If you need hundreds, the problem is typo tolerance and synonyms, not aliases.
• Forgetting the product pages. A search engine only finds what is written. Product pages without the manufacturer's reference, without a brand or with generic names give poor results with any engine.
• Not reindexing after an import. After loading or changing many products at once, the index has to be rebuilt; otherwise the search shows yesterday's catalogue.
• Testing only on desktop. Most searches come from mobile, with a small keyboard and more typos. That is where speed and tolerance need to be checked.
• Leaving PrestaShop and its modules out of date. A new search engine on an outdated store inherits its performance and compatibility problems.

What to ask of a store search engine

• Results from the first letters, with image, price and availability.
• Configurable typo tolerance and synonyms per language.
• Search by reference and by code, full or partial.
• Filters built from what has been found: category, brand, features, price.
• Adjustable weights: the name, the reference or the brand takes priority depending on your catalogue.
• Statistics on what is searched for and not found.
• Compatibility with your PrestaShop version, multistore and multilingual if you use them, and no overrides.

How to solve it in PrestaShop

The self-hosted Visual Search No SaaS search engine with Meilisearch option opens full screen over the page, shows results as the customer types and builds the filters (categories, brands, features, attributes and price) from what it has found. It tolerates typos in a configurable way, supports synonyms per language with CSV import and relevance weights by name, reference or brand, and lets customers add to cart straight from the result. It reads its own index with a results cache and, if you enable it, uses Meilisearch as the engine with automatic fallback to MySQL if it does not respond. It logs what is searched for, what is clicked and what does not appear, and can email whoever searched for something when it comes into the catalogue. You pay for it once, it does not call external servers or use overrides, and it is compatible with PrestaShop 1.7, 8 and 9, multistore and multilingual.

Frequently asked questions

Why does PrestaShop search not find products that exist?

Most often the index is not up to date or the customer types it differently from how it is on the product page. Rebuild the index and review the aliases and weights.

How do I reindex PrestaShop search?

From Shop Parameters, Search, with the option to rebuild the entire index. In large catalogues it is best done at low-traffic times.

Does Meilisearch work with PrestaShop?

Yes, through a module that uses it as the engine. You need a server where you can install the Meilisearch service.

Does a better search engine help SEO?

Indirectly: it improves conversion and shows you which words your customers search with, which you can then use in categories and product pages.