Online store search starts with the words customers type, not the search icon in the header. A shopper who knows a model number needs a different kind of result from someone looking for a broad product use. Your development brief should explain both situations using examples from your own catalogue.
For a business planning an ecommerce website in Kenya, search requirements are worth discussing before choosing an extension or approving a theme. A feature called smart search does not tell you whether it understands your product names, displays suitable results or helps someone recover from an unsuccessful query.
This guide helps a store owner prepare a query-and-result worksheet. It focuses on typed search, distinct from category navigation, product imports and browsing filters.
Collect a small set of genuine customer terms
Ask sales and support staff which words people use when requesting products. A supplier's formal description may differ from the phrase customers use in an email or phone call. Record those differences without copying personal details from customer conversations into the project document.
Include precise and broad requests. An exact item code, a common product name and a phrase describing intended use can reveal different requirements. Select examples your business can interpret confidently rather than inventing a large keyword list that nobody has reviewed.
For each term, record the intended item or product family and why it should match. Add any important exclusions. An illustrative search for a replacement charger should not return every cable merely because both descriptions contain the word charging. Product expertise matters when judging relevance.
Keep website search terms separate from Google keyword research. One concerns finding information inside your catalogue; the other concerns reaching the website from an external search engine. They can inform each other, but adding phrases to a page does not automatically configure the store's search behaviour.
Decide which content search should return
Write down whether the main search is intended to find products only or also help pages, collections and articles. A query about delivery may be better served by policy information than by an unrelated product. Decide whether that belongs in the same search experience or a separate help route.
Shopify's storefront search documentation explains that result types can include products, pages and blog posts depending on settings. It also distinguishes built-in features from additional Search & Discovery customisation. Use that as a prompt to verify the chosen platform's actual setup rather than assuming that every result type appears automatically.
Consider how different result types will be labelled. If an article and a product share a similar name, the customer should understand what will open. Ask the designer to show a mixed-result example if that is part of the agreed scope.
Keep unpublished and internal content outside the public search design. The developer should explain which catalogue states are searchable and demonstrate the result with controlled records. A convenient search box must not become an unintended route to draft prices or private operational information.
Agree names, aliases and exact-code handling
Prepare a small vocabulary sheet linking approved customer terms to catalogue terms. A common abbreviation may be useful, but a technically different specification should not be treated as a synonym simply because it sounds similar. Have a product owner approve the mapping.
Define what should happen to exact item references. Ask whether punctuation, spaces and letter case affect matching and whether a complete exact code should take priority. Demonstrate the behaviour with representative codes rather than relying on a sales description of the search tool.
Discuss spelling mistakes proportionately. You may want obvious typing errors to return a helpful match, but an overly broad correction can send someone to the wrong model. The acceptance worksheet should include cases where a near-match is unsuitable as well as cases where it helps.
Avoid changing public product names solely to manipulate the search results. First ask whether the catalogue data is accurate and whether the search configuration can represent an approved alias separately. The product spreadsheet guide helps organise the underlying names and references before import.
Review the first few results as a buyer
For each sample query, identify which results should appear near the beginning and which should not appear at all. You do not need to prescribe a universal ranking formula. You do need to explain the commercial meaning of a relevant result for the products you sell.
Distinguish relevance from promotion. If the business wants to feature a product, record that decision and its conditions. Do not silently replace an exact requested item with a higher-margin alternative while implying that the search found the same thing.
Review unavailable and discontinued items separately. An existing product reference may still help a customer identify a replacement, but the status and next action must be clear. Ask whether that result should remain, link to an approved alternative or lead to a support explanation.
Check the result card itself. A useful match can still disappoint if its title is unclear, the picture belongs to another option or the displayed price does not explain the item being offered. Search acceptance should include what the buyer sees before opening the product page.
Specify the phone search interaction
Ask the designer to show opening search, typing, waiting, selecting a suggestion and viewing full results on a phone. A search field that looks generous on desktop can become cramped when the phone keyboard is visible. Review the actual journey rather than a static header mockup.
If predictive suggestions are included, decide what they represent. Suggested terms, product matches and category shortcuts are different types of action. Make their presentation understandable and ask how someone reaches the full result list without being forced to choose a suggestion.
Review how the query can be corrected. A customer should be able to see the current text, edit it and understand whether they are viewing the old or new result. Agree the feedback shown while results update, especially if the network response is delayed.
Include keyboard and assistive-technology requirements in the development discussion. Ask the team to demonstrate the implemented suggestion control using the agreed evaluation methods. A visually attractive dropdown is only one part of an operable search experience.
Make unsuccessful searches useful and honest
Choose a query that should produce no match and review the whole screen. State whether no products match or whether the service is temporarily unavailable. Those are different situations and should not be hidden behind the same empty page.
Provide an understandable recovery route. Depending on the store, that could mean editing the query, browsing a relevant category or asking for help. If alternatives are presented, label them as alternatives rather than implying that they satisfy the original specification.
Avoid placing an empty-result product enquiry into an unrelated marketing sequence. If search assistance collects contact details, agree its service-specific destination, attribution and response owner with the CRM team. Keep that separate from the site's general newsletter or shop purchase workflow.
Where filters can narrow a result further, define how their selections interact with the typed query. The product filters requirements guide covers matching combinations and empty-result recovery in more detail. Keep both briefs consistent without treating typed search and filters as the same feature.
Use a repeatable search acceptance sheet
Create columns for query, starting catalogue state, expected leading matches, unacceptable matches, result types and observed outcome. Add the build or configuration version reviewed. This gives the developer an actionable example when the business says a result is irrelevant.
Test with a representative controlled catalogue first. Include exact references, an approved alias, a deliberate typo, a broad request and a query that should return nothing. Do not create live orders or customer enquiries merely to inspect search results.
When an example fails, identify whether the cause is missing catalogue information, a disputed vocabulary mapping or implementation behaviour. Correct the underlying cause and repeat related queries. Fixing one search term should not quietly damage another important match.
Maintain the query list after launch
Assign an owner for vocabulary and search configuration changes. Staff adding products should know which fields influence discovery and where approved aliases are recorded. Otherwise, useful search behaviour can degrade as new naming habits enter the catalogue.
If search reporting is available, review it with appropriate privacy controls and use aggregate patterns. A query can contain information someone typed unexpectedly. Do not copy raw search logs or personal information into a widely shared marketing report simply because the reporting tool permits export.
Treat zero-result counts as investigation prompts. They may reflect missing products, unfamiliar wording, deliberate testing or requests the business does not serve. Verify the explanation before changing stock plans or claiming that a search update increased sales.
Frequently asked questions
Is typed search the same as a product filter?
No. Typed search interprets a query, while filters narrow results using selected properties. A store may combine them, but both need clear rules and appropriate examples in the acceptance plan.
Must we buy an advanced search extension?
Review your requirements against the platform's actual capabilities first. Ask the developer to demonstrate representative queries and explain any additional tool, recurring cost or maintenance responsibility before you approve it.
Can a popular search term justify adding stock?
It can prompt investigation, but it is not proof of demand or a completed purchase. Review the term's meaning, repeated patterns and relevant business evidence before making a stock decision.
What should we send for an ecommerce proposal?
Share sample products, customer vocabulary, important item codes and a handful of expected search results. Get a free quote and include search configuration, mobile interaction and acceptance evidence in the scope.
