Predictive search suggestions, the dropdown that appears while a shopper is still typing, are the shortest path from intent to product that exists on your store. A shopper who taps a suggestion skips your navigation, your collection grid, your filters, your pagination, all of it: three keystrokes and they are standing on a product page. Every other journey on your site is a hallway. This one is a teleporter, and most merchants have never once looked at where it teleports people to.

The dropdown is a shop window with three slots

Type two letters into your own search box on your phone and look hard at what comes back. That dropdown shows perhaps three to five products. Are they your bestsellers, in stock, at full size runs? Or an alphabetical accident: a discontinued colorway, a $6 accessory, a gift card? The shopper tapping a suggestion carries the highest intent on your entire site, and the dropdown is choosing their destination. Default suggestion logic matches text; it does not ask whether the match deserves the slot. This is the same lesson as tip 2’s collection shelf, compressed into three rows of pixels: whatever appears there is what your store is recommending, whether you chose it or not.

Make the teleporter land somewhere good

• Test your top ten queries (you have them from your search reports, tip 17). Two or three typed characters each: do the right products surface before the query is even finished?
• Never suggest a dead end. A sold-out product in a three-slot dropdown is a wasted third of your best real estate.
• Mind the partials. Suggestions fire on fragments, so “gl” decides between gloves and glitter. Your bestselling category should win its own fragments.
• Automate the ordering. Suggestion slots deserve the same merchandising rules as your collections: bestsellers forward, out-of-stock demoted, new arrivals given their shot. This is exactly what Dynasort does to the predictive dropdown, the same automated ranking that runs your collections applied to those three little slots, so the teleporter always lands on a product worth selling.

Shoppers who search convert at a multiple of browsers; shoppers who take the suggestion shortcut are the sharpest end of that group. Sharpen the shortcut back. It is three slots. Curate them like the window of the store.

Zero-result searches are the highest-intent failures on your entire store. Think about who types into a search box: someone who knows roughly what they want, described it in their own words, and asked you directly. Tip 40 had you measure how much better those people convert. And your store’s response to some of them, right now, is a blank page and a shrug: “0 results for ‘jogers’.” That is a salesperson watching a customer walk in, ask for joggers, and walking away without a word.

The infuriating part is that the store usually has the product. Zero-result pages are mostly vocabulary accidents: a typo, a synonym you never mapped (tip 30), a singular where your title is plural, a brand name spelled from memory. The shopper did nothing wrong. The catalog did nothing wrong. The translation layer failed, and the default failure mode is a dead end.

Anatomy of a rescue

The rule: the results page may say “no exact matches,” but it must never be empty. Something relevant, something clickable, always.

Build the net, then shrink the hole

• Layer the fallbacks. Spelling suggestion first, then relaxed matching on partial terms, then bestsellers with an honest “no exact matches, but shoppers love these.” Each layer catches what the one above missed. On our own platform this is codified as a rule: when a query in Dynasort-served search comes back empty, we automatically widen the match rather than serve the blank page, because a dead end is the one result we refuse to return.
• Add a human exit. A “can’t find it? Ask us” link on the zero-results page turns your worst page into a support lead with purchase intent attached.
• Fix the causes upstream. Your zero-result query log is a to-do list: every recurring failed term is either a synonym to add (tip 30), a product to stock, or a category you should stop assuming people can spell.

A shopper who searches is asking to be sold to. Answer every time, even when the answer starts with “not exactly, but.”

Search synonyms exist because your customers search in their words while your catalog is written in yours, and the gap between those two vocabularies is a quiet daily leak. The shopper types “hoodie”. Your product is a “pullover fleece”. The search returns nothing, the shopper concludes you do not sell hoodies, and they leave to buy one from someone whose copywriter was less creative. Nobody logs a complaint. The sale just does not happen.

This is a naming-culture problem, and it is universal. Merchants name products the way parents name children: with love, aspiration, and zero regard for searchability. Customers name products the way they talk. “Sneakers” vs your “trainers”. “Couch” vs your “sofa”. “Swimsuit” vs your “bathing suit”. “Phone case” vs your “device sleeve”. Regional English alone, US vs UK vs Australia, will torpedo a search box that only knows one dialect.

Build the dictionary from evidence

Do not brainstorm synonyms in a meeting; harvest them. You read your search logs in tip 17, and the raw material is sitting right there: every query with zero or weak results that should have matched something you sell is a synonym candidate with proof of demand attached. Add the misspellings you see repeatedly, the brand-name genericisms of your niche, and the regional variants for every country you ship to.

On Shopify, the free Search & Discovery app lets you define synonym groups for storefront search, and most dedicated search tools have an equivalent screen. The mechanics take an afternoon. The discipline is the monthly loop: new zero-result queries in, new synonym rows out, forever, because your customers’ vocabulary keeps evolving even when your catalog does not.

One more source hiding in plain sight: your product copy itself. If the words customers use appear in your descriptions and tags (“a hoodie you will live in all winter”), many search engines match without any synonym rule at all. Writing product copy in customer language is the synonym list you never have to maintain, and it converts better for the humans reading it too.

Searchers are your highest-intent visitors, you measured that multiple back in tip 5. A synonym list is twenty minutes of translation work that stops telling them “no results” when the answer was “aisle three, under a different name”.

Your site search logs are voice-of-customer research that runs every day, costs nothing, and gets read by almost nobody. Every query is a shopper telling you, in their own words and with purchase intent, what they came for. Read them monthly and you get a product roadmap, a merchandising to-do list, and a dictionary of customer language, written by the only focus group that pays you.

Surveys tell you what customers think they want. Search logs tell you what they tried to buy thirty seconds ago. When someone types “gift box” into your store and you do not sell gift boxes, that is not a data point, that is a purchase order you declined. Fifty people typing it is a product decision made for you.

Pull last month’s queries from wherever yours live, your search app’s analytics or GA4’s view_search_results events with the search term parameter, and sort every query into four buckets.

Make it a ritual, not a rescue

Thirty minutes, first Monday of the month, recurring calendar slot. Skim the top 50 queries, bucket them, and leave with three actions: one ranking to fix, one synonym to add, one product gap to consider. The discipline matters more than the depth; a shallow read every month beats a heroic audit once a year, because the logs go stale as your catalog and season change.

Between now and Cyber Monday, this ritual has a bonus payoff: September and October queries are a preview of what your Q4 shoppers will hunt for, spelled exactly the way they will spell it. The stores that read the logs get to stock the shelf before the rush finds it empty. And those zero-result queries in bucket two? They are such a costly, fixable leak that they get an entire tip of their own down the road.

What Does Shopify’s Native Search Already Do?

Shopify Search & Discovery is free and better than its reputation: synonym groups, product boosts, filters, and query handling that improves every year. For a small catalog with clean product data it is often enough. Its limits are ranking limits: results cannot be ordered by your sales, margins or inventory, boosts are hand-picked per product rather than data-driven, and sold-out products keep their positions. [VERIFY current S&D feature list before publish]

Replace or Re-Rank: Which Kind of Search App Do You Need?

The diagnosis is measurable. Check your zero-result rate: if a meaningful share of searches return nothing (or garbage), matching is your problem and a replacement earns its complexity. If results are relevant but badly ordered, sold-out items up top, bestsellers on page two, that is a ranking problem, and measuring your search conversion against browse conversion tells you what the fix is worth.

The Best Shopify Search Apps, Reviewed

1. Shopify Search & Discovery (free, native)

The baseline. Synonyms, boosts, filters, zero cost, zero added scripts. Start here, and only spend money once you can name the problem it is not solving.

2. Dynasort (re-ranker)

Keeps Shopify’s matching and re-ranks every result set live on each request by your recipe: sales velocity, inventory (sold-out sinks or hides, restocks return automatically), margin, reviews, newness. Same rules run your collections and recommendations, so one logic governs the store. Flat pricing, 30-day trial, Built for Shopify certified. To be fair: if your matching is broken (heavy typos, cross-language queries, zero-result floods), re-ranking cannot fix what never matched; that is replacement territory, and we say so in our search comparison.

3. Boost AI Search & Filter

The most established replacement in the App Store: instant search, typo tolerance, deep filter building, merchandising rules inside results. To be fair: it replaces your theme’s search UI, pricing scales with product count [VERIFY], and its collection merchandising lives inside the search-and-filter context. Head-to-head: Dynasort vs Boost.

4. Searchanise

A long-running replacement with smart search pages and custom filters at aggressive pricing. [VERIFY tiers] To be fair: the widget look takes theme work to feel native, and merchandising controls are lighter than Boost’s.

5. Algolia

Search infrastructure with genuine speed and typo tolerance, popular with development teams that want API-level control. To be fair: it is a platform you build with, not an app you switch on; budget for implementation, and query-volume pricing needs watching. Our Dynasort vs Algolia page covers when each makes sense.

6. Klevu

Replacement with strong natural-language handling and self-learning ranking, positioned upmarket. [VERIFY current Shopify offering and pricing] To be fair: enterprise-leaning pricing for what many stores need.

7. Fast Simon

Replacement bundling search with merchandising and personalization features. [VERIFY] To be fair: the bundle is broad rather than deep in places; compare the specific features you need. See Dynasort vs Fast Simon.

8. The enterprise tier

Searchspring, Nosto and friends sell search inside a managed discovery platform with services attached. Right for enterprise volume and teams; more platform than most Shopify stores need.

How Do You Choose a Search App?

Measure before you shop: your zero-result rate names the problem, and your search-vs-browse conversion gap prices the fix. Matching problem: pick a replacement (Boost first look, Searchanise on budget, Algolia with a dev team). Ranking problem: pick a re-ranker and keep your theme untouched. Both problems: fix matching first; ranking polish on results nobody gets is wasted. And whatever you install, re-test the zero-result rate and the conversion gap after four full weeks.

Frequently Asked Questions

Is Shopify’s built-in search good enough?

For small catalogs with clean product data, often yes. Spend money only when you can name the failure: searches returning nothing, or good matches badly ordered.

Do search apps slow down the storefront?

Replacements add their own scripts and UI; the good ones are fast, but it is real added surface area. Re-rankers change the order server-side with no storefront scripts, so speed stays exactly as it was.

What is search relevance, actually?

Two separate things bundled in one word: whether the right products match the query at all, and whether the best of those matches appear first. Apps that improve one do not automatically improve the other, which is exactly the replacement-versus-re-ranker split.

Data referenced: 30 days of storefront search behavior (188k clicks). Competitor facts from public listings, August 2026.

Dynasort re-ranks your Shopify search results by what sells, live on every request. Install it from the Shopify App Store or see search re-ranking in action.

Choosing your first recipe

Every store gets two ready-made default search recipes. Relevant & Popular is a balanced blend of popularity signals and is the safe pick for most catalogs. In-Stock Bestsellers leans on sales velocity and inventory depth, and demotes sold-out products to the bottom of the results. Activating one of them on the Search page is step one, and for most stores it should stay step one for at least a week. Do not build a custom recipe on day one. The defaults exist so you can start collecting real serve data before you start expressing opinions.

If you sell anything that goes out of stock, In-Stock Bestsellers is worth a hard look. Across our fleet data, recipes that demote sold-out products convert 56 percent above the sorted-fleet average. That is a correlation across many stores, not a guarantee for yours, but it points at something real: shoppers who search have high intent, and a sold-out product in slot one is the fastest way to waste it.

Note that demote means demote, not hide. Dynasort pushes sold-out products to the bottom rather than removing them, because hiding them creates dead ends. A shopper searching for a product they saw last month wants to know whether it is ever coming back, and an empty results page answers the wrong question.

One setting deserves a special mention: the tie-breaker. Search recipes can use Shopify relevance as the tie-breaker, and for search we recommend leaving it there. When two products score the same under your recipe, they fall back to Shopify’s relevance order. On a young store with thin sales data, lots of products tie, so results quietly degrade to plain relevance instead of arbitrary order. As data accumulates, your recipe takes over product by product.

Tuning weights: one at a time, then A/B it

After a week on a default, you will have serve stats and probably a hunch. Maybe new arrivals deserve more lift, maybe margin should count for something. Here is the workflow we recommend, because it is the one that produces conclusions instead of vibes:

1. Clone the default recipe. Cloning keeps the original untouched as your baseline.
2. Change exactly one weight on the clone. One. If you move three sliders and results improve, you have learned nothing about which slider did it.
3. Run an A/B test with the clone against the original on live traffic. Both recipes serve real shoppers, and the recipe page shows serves, clicks, add-to-carts, and checkouts for each.
4. Keep the winner, then repeat with the next hunch.

Custom attributes are where tuning gets interesting. If your business has a signal Dynasort cannot see, seasonal priority, supplier lead time, a merchandiser’s picks, you can build it from tags, metafields, or an uploaded spreadsheet, and then weight it in the blend like any built-in signal. If you would rather work through this on a screen share, book a free setup session and we will tune your first recipe together.

Making result cards look native

Ranking is half the job. The other half is a results page that looks like your store built it. Card display settings live in the Dynasort admin and apply per store: badges for sale, new, bestseller, and low stock, each with its own toggle and color; color swatches with shape and style options; compare-at pricing; ratings; a second image on hover; quick add to cart; column counts per device; and results per page.

Then there are option display rules, the setting that fixes messy variant data. They control how variant options render on cards: hide a named option group entirely, split combined option values with a delimiter, and choose which option acts as the visual axis. If your products carry a combined option like “Color / Size”, one rule splits it, hides the size half, and shows clean color swatches, no product data cleanup required.

CSS overrides for designers

The storefront widgets deliberately avoid !important, so your theme CSS wins any fight it picks. For the common adjustments you do not even need a selector, because the widgets expose CSS custom properties you can override from your theme: --ds-cols-mobile, --ds-cols-tablet, --ds-cols-desktop, --ds-aspect, --ds-gap, --ds-radius, --ds-swatch-size, and --ds-max-width. For everything else, the markup uses stable namespaced classes like .dynasort-search__card that will not shift under you between updates.

One gotcha worth pinning to the wall: Online Store 2.0 themes store their colors as bare RGB triples. If you want a card to pick up a theme color, write rgb(var(--color-background)), not var(--color-background). The bare variable is three numbers, not a color, and the browser will silently ignore the rule.

Measuring what changed

Every serve is tracked. The recipe page shows serves, clicks, add-to-carts, and checkouts per recipe, so a tuning change is never a matter of opinion for long. Combined with A/B testing, that turns search merchandising into a loop: watch a week of stats, form a hunch, clone, change one weight, let the test decide. And your existing Shopify setup keeps doing its part, since synonyms configured in Shopify’s free Search & Discovery app flow through automatically, and filters configured there render on the results page and keep working.

The setup itself is still three steps: activate a recipe, connect your storefront, and add the app block plus the live search embed in your theme editor. Everything in this guide layers on top of that without touching code, except the CSS section, which is there for the designers who want it.

Install Dynasort free for 30 days or read the search docs. Already running Dynasort Search? Book a free training session and we will review your recipe, cards, and stats with you on a screen share.

How to use this comparison: these are good products with different architectures, and the right choice depends on what you need. If you need typo-tolerant instant search, visual search, or per-shopper personalization, some of the apps below do those things and we do not. If you want merchandising control over the order of results without migrating your search stack, that is what we built. We sell a specific architecture, not a claim that the alternatives are bad.

The one structural difference

Almost every dedicated search app for Shopify works the same way at the foundation: it syncs your catalog into its own search index and serves results from its own engine. That is a deliberate and reasonable design. Owning the index is what buys these vendors true typo-tolerant instant search, semantic matching, and features Shopify’s engine does not offer.

It also has costs, documented in the vendors’ own materials: a second index that has to stay in sync (Searchanise re-indexes products on a 10-minute cycle and asks for manual re-indexation of collections and pages after edits; Klevu’s delta sync runs roughly every 10 minutes; Fast Simon’s Starter plan documents hourly synchronization), synonyms managed in the vendor’s dashboard rather than in Shopify Search & Discovery, and in Algolia’s case, setup instructions that explicitly say to turn off the default Shopify search results in the theme.

Dynasort takes the opposite bet. Shopify’s engine keeps deciding which products match a query, so your synonyms, Shopify’s typo handling, and your Search & Discovery filters all keep working, and there is no second index to sync. Dynasort re-ranks the matched set on the results page and in the predictive dropdown, using the same merchandising recipes that already run your collections: sales velocity, inventory, newness, margin, and your own custom attributes. One recipe engine covers four surfaces (collections, search, recommendations, and post-purchase upsells), with built-in A/B testing and per-serve tracking to prove what a change did.

The comparison table

The apps, one at a time

Algolia is the enterprise search platform of the group, with its own relevance engine, AI ranking, synonym groups, custom ranking rules, and A/B testing. Its Shopify setup syncs products into Algolia’s index and has you turn off the native search results section, with synonyms and ranking managed in the Algolia dashboard. Pricing is usage-based rather than flat tiers, with a free tier up to 10K searches a month. If you want fine-grained relevance engineering, it is the deepest toolset here.

Boost AI Search & Filter is one of the most-installed options, with 1,498 App Store reviews at 4.8 stars. It pairs its own AI search with merchandising campaigns (pin, boost, demote, visual rules) and configurable filter trees, plus per-shopper personalized search on its top tier. Pricing scales with store GMV, starting at $29/mo. A/B testing is not an advertised feature.

Searchanise has been around since 2015 and holds 4.8 stars across 1,071 reviews. It runs search from its own servers with instant-search widgets, unlimited filters, and merchandising tools, priced from free under 25 products and $19/mo after that. Products re-index automatically every 10 minutes; collections, pages, and blog posts need manual re-indexation after edits.

Fast Simon is the broadest own-index platform in this set: AI search across text and image, per-shopper personalization that adapts results to each visitor, visual merchandising tools, filters, recommendations, and upsell, plus A/B testing for merchandising rules. It carries the Built for Shopify badge. Pricing is session-metered from a free 100-session tier, and the Starter plan syncs data to its index hourly.

Klevu sells modular products: Site Search at $649/mo, Category Merchandising at $549/mo, and Recommendations at $449/mo, each usage-metered. In exchange you get an AI and NLP engine with a personalization layer, a visual merchandising dashboard, and A/B testing for recommendations. Its breadth is the closest to Dynasort’s four surfaces, assembled from separately priced modules on Klevu’s own index.

Kimonix is the closest architectural relative to Dynasort: a merchandising-first app that sorts Shopify collections natively rather than replacing search, ranking on 100-plus parameters like revenue, conversion, margin, and inventory, with collection A/B testing built in. Its plans cap re-sorts at 1 to 4 updates per day and cap collection counts per tier, and its search product is a conversational shopping-assistant layer rather than a re-rank of Shopify’s search matches. At $19/mo entry and a 5.0-star rating over 198 reviews, it is a strong choice for collection-focused merchandising.

Beyond the majors: Sparq ($19/mo) and Doofinder ($61/mo, 741 reviews at 4.7 stars) are own-index search and filter apps, with Doofinder adding semantic search, voice and image search, and recommendation carousels. Layers is a semantic search and merchandising platform built exclusively for Shopify Plus, sold direct on custom quotes with near-real-time indexing and visual discovery. Searchmind is an agentic semantic search platform, also quote-priced and integrated with Shopify directly rather than through an App Store listing. All four replace the Shopify search index with their own.

Where the replace-the-engine approach is stronger

Honesty section. Because own-index vendors control matching, they can offer things we structurally cannot:

• True typo-tolerant instant search. Dynasort inherits Shopify’s matching behavior, whatever it is. Own-index engines tune their own.
• Semantic and visual search. Layers, Doofinder, Fast Simon, and Searchmind offer image or intent-based search. We have nothing comparable.
• Per-shopper personalization. Boost, Fast Simon, Klevu, Doofinder, and Algolia offer 1:1 personalized results. Dynasort recipes rank for the store, not per individual shopper.
• Filter depth and visual editors. Boost, Searchanise, Sparq, and Fast Simon are as much filter products as search products, and several vendors ship drag-and-pin visual merchandising dashboards. We lean on Shopify Search & Discovery for filters by design.
• Track record. Boost, Searchanise, and Doofinder each have hundreds to thousands of reviews. Dynasort Search shipped this week.

If any of those is the thing you need, choose accordingly. They are real capabilities and the vendors above have earned their reputations.

Where Dynasort fits

Dynasort is for merchants who like how Shopify’s search matches products and want control over the order it puts them in, without adopting a second index, a second synonym dashboard, or a migration project. You get one automated ranking brain across collections, search results, the predictive dropdown, recommendations, and post-purchase upsells, driven by business signals you choose, measurable with built-in A/B testing and per-serve tracking. Setup is theme-editor only, no code, and pricing is flat: $99, $249, or $499 per month with a 30-day free trial.

Two architectures, two bets. The own-index apps bet that better matching wins. We bet that for most stores, Shopify’s matching is already good, and the ordering is where the money is.

Install Dynasort free for 30 days or read the search docs. Want a walkthrough instead? Book a free setup session and we will compare notes on your current search stack too.

We also published a broader, vendor-neutral version of this comparison: the best Shopify search apps, replace or re-rank.

Two datasets. First, the fleet: every storefront search that flowed through stores on the Dynasort platform over the last 30 days, bots excluded. Second, a deep dive into one high-traffic apparel and jewelry store’s full year of search tracking. Everything is aggregated or anonymized, and as always we will tell you exactly how far each number can honestly be stretched.

The fleet, in two numbers

Just over 1.34 million searches in a month is a lot of shoppers typing exactly what they want into a box. Search users are widely known to be the highest-intent visitors on a store, which is what makes both of those tiles matter: the surface is busy, and one in every 22 of those high-intent moments ends in a blank page.

Where clicks actually land: one store, one year

Fleet totals tell you how big the surface is. To see how shoppers behave on it, we went deep on a single high-traffic apparel and jewelry store: 365 days of tracking, 188,323 clicks on search results. To be clear up front, this is one store we studied closely, not a fleet average.

Only 28.5 percent of clicks landed on the first result. Nearly half of all clicks, 49.4 percent, landed past the top three, and the average click sat at position 4.51. The detail that surprised us most: positions 7 through 12, the bottom half of a typical first screen, collected exactly as many clicks as position 1.

That says two things at once. Shoppers on this store scan well past the top, so a good product at position 9 is not lost. But position still carries enormous weight per slot: one slot got 28.5 percent of clicks, and it took six slots further down to match it. Whatever sits at the top gets a structural advantage, and default text relevance hands that advantage out with no idea what is in stock, what is selling, or what just launched. That is the whole argument for putting merchandising signals in charge of the order, and it is why we built Dynasort Search.

Zero-result searches are shoppers telling you what they wanted

The same store gave us a year of dead ends to study: 602,891 searches, of which 37,183 returned nothing at all. That is a 6.2 percent zero-result rate, a bit above the 4.5 percent fleet figure for the last 30 days.

A zero-result search is the most honest feedback a store ever gets. Nobody types into a search box by accident. This store’s zero-result queries clustered into three buckets: catalog gaps, synonym gaps, and typos. The single largest source of unmet demand was a product category the store simply did not carry, followed by obvious typo variants of things it did.

Dynasort tracks every storefront search on a connected store, including the dead ends, so this list is not a research exercise. It is sitting in your search report: the exact words shoppers used, ranked by how often they hit nothing. Some entries are a merchandising fix, some are a synonym you can add in five minutes, and some are a product line your shoppers keep asking for by name.

How to read these numbers honestly

This is a data post in a series whose whole point is honest math, so here are the limits.

• The click position data comes from one store. It is a large, clean, year-long sample, but a store with a different catalog, theme, or grid layout will scan differently. Treat the shape as a well-documented example, not a law.
• Zero-result rates vary widely by catalog breadth. A store with 40 products will see far more dead ends than a store with 40,000, and neither number is a report card.
• Re-ranking does not fix zero-result searches. If nothing matches, there is nothing to reorder. The fix for dead ends is usually catalog or synonym work. Re-ranking helps the other 95 percent of searches, the ones that do match and then come back in an order chosen by text relevance alone.
• None of this is a causal claim about conversion. Search users convert at multiples of browsing shoppers across the industry because high-intent people search, not because a search box works magic. What the data supports is narrower and still valuable: search is a busy, high-intent surface, position on it carries real weight, and the default order is decided by a signal that knows nothing about your business.

What to do with it

If half the clicks on a busy store land past the top three, the order of the whole first page is merchandising real estate, not just slot one. Dynasort Search puts your recipe in charge of that order: recent sales, inventory depth, newness, margin, or any custom signal you build, on both the results page and the predictive dropdown. Shopify still decides what matches. Your business decides what comes first. Monday’s launch announcement has the full picture, and tomorrow we close the week with a practical guide to search recipes and result cards.

Install Dynasort free for 30 days or read the search docs. Already a Dynasort merchant? Your search report is already collecting this data for your store, and a free setup session gets search live with you on a screen share.

Site search users convert at a multiple of your browsing visitors, on almost every store we have ever looked at. The exact multiple varies, which is why the first half of this tip is: go measure yours. The second half is what that number should change about how you run your store.

The logic is not mysterious. A visitor who types “linen midi dress” into your search box has told you exactly what they want, in their own words, with money in hand. A visitor drifting through your navigation is still deciding whether they want anything at all. One of these people is a hand raised in a crowded room. Most stores treat both hands the same.

Measure your multiple this week

In GA4, compare conversion rate for sessions that include a view_search_results event against sessions that do not. If your search app has its own analytics, cross-check there. Divide one number by the other and write the multiple next to the baseline you set in tip 1. That single division is the business case for every search improvement you will ever consider, and we will lean on it repeatedly later in this series.

Then treat the lane like a VIP lane

• Make the box findable. If search hides behind a tiny icon on mobile, your highest-intent visitors are playing hide and seek with your best feature.
• Try the top ten queries yourself. Type them on your phone. Every dud result set is a shopper with a raised hand being handed a shrug.
• Rank the results like you mean it. Which products come back first for a query is merchandising, exactly like a collection page, and the default order rarely matches what you would show a hand-raised shopper: in-stock first, proven sellers up top, stragglers down. This is why we built Dynasort to re-rank search results automatically with the same signals that sort your collections.
• Stop dead ends. A zero-results page with no suggestions is the single worst page a high-intent shopper can land on. We will spend a whole tip on rescuing it.

Nobody accidentally types “waterproof trail runners size 11” into a search box. Find out what those people are worth to you, then act like it.

Every search is really two jobs

When a shopper types “linen shirt”, two separate questions need answering. First: which products match? That is membership, and it is genuinely hard. It means handling typos, applying synonyms, and respecting any filters the shopper has applied. Second: in what order should the matches appear? That is ranking, and it is a business question, not a text question.

We split the two jobs cleanly. Shopify’s search engine keeps the membership job: it decides which products match a query, including synonyms, typo handling, and the filters shoppers use to narrow down. Dynasort takes that matched set, scores every product with your active search recipe, and re-orders it. Then we render the results grid and the predictive dropdown in your theme. That is search re-ranking, and it is all we do.

What a search recipe actually scores

A search recipe scores every matching product on the merchandising signals you choose and weight. Out of the box that includes recent sales over the last 7 and 30 days, all-time sales, inventory depth measured in variants in stock, newness measured by the age of the newest variant and days since published, on-sale status, sold-out status, margin, and price.

Then there are custom attributes: any product data you can express, you can rank by. Tags, metafields, even a spreadsheet upload. A common recipe choice is sold-out demotion, which pushes sold-out items to the bottom of the results instead of hiding them, so shoppers still find the product but never hit a dead end in slot one.

What we deliberately did not build

We could have built our own search engine with its own index of your catalog. We chose not to, and we consider that the best decision in the whole product. Here is what not replacing Shopify’s engine buys you:

• No index drift. There is no second copy of your catalog to sync, so a product can never exist in one engine and be missing from the other. No sync jobs, no migration, nothing to babysit.
• Your synonyms keep working. Anything you configured in Shopify’s free Search & Discovery app flows through automatically.
• Your filters keep working. The filter definitions you set up in Search & Discovery render on the Dynasort results page and narrow results exactly as before.
• Nothing to unlearn, and nothing risky to undo. Shopify keeps answering “what matches” the way it always has, so removing Dynasort simply returns you to the default order. No re-indexing, no rebuild.

Fast, and never stale

Both surfaces are built for speed. The full results page re-ranks the complete matched set. The predictive dropdown re-ranks a smaller candidate set, because as-you-type has to feel instant.

Ranked results are cached briefly, about 15 minutes per query and recipe, so warm queries serve in tens of milliseconds. The product cards themselves, price and stock included, are always built fresh, so a shopper never sees a stale price or a phantom “in stock”. And because the cache is keyed to your recipe version, editing a recipe takes effect immediately. Change a weight, run the search again, see the new order.

Honest caveat: new stores start plain

Re-ranking is only as sharp as the data behind it. On a brand-new store with little sales history, many products score the same, and tied scores need a tie-breaker. You can choose “Shopify relevance” as that tie-breaker, so on thin data the order gracefully degrades toward Shopify’s default instead of something arbitrary. The system is designed to degrade gracefully, and it gets sharper as your sales and inventory data accumulate.

Every serve is measured

Every search Dynasort serves is logged: which recipe served it, what was shown, what got clicked, and what got added to cart. That powers per-recipe performance stats, and it powers A/B testing, built into Dynasort, where two recipes run head-to-head on live traffic so you can see which order actually sells.

Setup is three clicks, not a project

Because there is no index to build, setup is short: activate a search recipe, connect your storefront with one click, and add the app block plus the live search embed in your theme editor. No code. Rather do it on a screen share? Book a free setup session and we will configure it with you.

Install Dynasort free for 30 days or read the search docs. Tomorrow: the data, what fleet-scale search behavior says about default ordering, and how Dynasort Search compares with the other search platforms you have heard of.