'%20clip-path='url(%23canvas-clip)'/%3e%3cg%20clip-path='url(%23canvas-clip)'%3e%3cg%20transform='rotate(0,%20256,%20256)'%3e%3cg%20transform='translate(87%2087)%20scale(14.083333333333334%2014.083333333333334)%20translate(0%200)'%3e%3cpath%20fill='none'%20stroke='%23FFFFFF'%20stroke-linecap='round'%20stroke-linejoin='round'%20stroke-width='2'%20d='m18.064%2010.877l-4.89-7.26c-.42-.625-1.287-.803-1.936-.397a1.4%201.4%200%200%200-.41.397l-4.893%207.26C4.24%2013.715%204.9%2017.318%207.502%2019.423a7.13%207.13%200%200%200%204.168%201.572M15%2018a3%203%200%201%200%206%200a3%203%200%201%200-6%200m5.2%202.2L22%2022'/%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
ParticleSearch Search Profiles: Choose the Right Starting Strategy for Your Catalog
A search profile is a coordinated starting strategy. It changes several connected choices together so a merchant does not have to assemble relevance, typo, identifier, recovery, availability, and presentation behavior one switch at a time.
Profiles exist because those choices interact. A more generous typo and discovery posture can help a shopper describe a need, but it can also make a near-collision identifier less safe. A strict exact-match posture can protect parts and SKUs while giving an exploratory shopper too little help. The profile makes the starting trade-off coherent and visible.
The right profile is therefore determined by the queries your store must answer and the mistakes it cannot afford, not by the industry name that sounds most familiar. This guide explains what a profile changes, why the settings move together, how the six ParticleSearch postures differ, and how to test and publish one without trading a visible improvement for a hidden regression.
Profile behaviour checked August 4, 2026. The profile names and merchant-visible behaviour were verified across the ParticleSearch dashboard, published settings contract, and storefront consumer. Private scoring logic is intentionally omitted.
Chapter 1 · The profile's job
A profile aligns related choices around one search posture
Strict identifier retrieval and broad discovery make different tradeoffs. A catalog with model numbers should not recover from a near match as freely as a visual home-decor store responding to “warm minimalist lamp.” One default cannot express both priorities equally.
ParticleSearch profiles coordinate the starting posture. They can affect relevance, typo behavior, identifier strictness, recovery, suggestions, dynamic filtering, diversity, stock and discontinued handling, new arrivals, product presentation, and the semantic quality floor that decides how close a meaning-led candidate must remain before it can participate. The settings remain merchant-controlled after the profile is applied.
Matching posture
How strongly exact identity, wording, meaning, and recovery should influence retrieval, and how close a meaning-led candidate must remain before it is eligible.
Catalog posture
How variants, availability, structured fields, diversity, and new products should behave.
Storefront posture
Which product details help a shopper judge the result without losing product identity.
A semantic profile controls both influence and admission
A strategy can give meaning-led retrieval more or less influence after candidates are admitted. It also needs a boundary for which candidates are close enough to consider at all. Without that second boundary, reducing keyword influence can allow increasingly remote products into the set.
ParticleSearch profiles coordinate that eligibility floor with the rest of the posture. Exact-match and technical strategies remain more conservative. Discovery-led strategies can admit a wider set, but they still use a quality boundary. Merchants do not need to tune or memorise internal distance values.
This makes a profile change more consequential, not less. Replay exact identifiers, descriptive queries, known empty searches, and near-neighbour products so broader discovery does not become false relevance.
A profile is a starting posture, not a promise
The name of a profile should help a team discuss tradeoffs, not replace testing. “Exact-match catalogue” can protect identifiers while making broad discovery too narrow. “Discovery-led catalogue” can recover useful language while making a near-collision part number less safe. That is why the decision belongs to the query fixture set and the observed result, not the label alone.
Decision principle
Choose the profile that protects your most expensive mistake
There is no universally best profile. A parts merchant should prefer precision over generous recovery. A visual catalogue should prefer useful discovery over a rigid identifier posture. Start with the mistake that would cost the store the most, then verify that the profile also supports the secondary queries shoppers use every day.
Balanced commerce
Use this when your store has mixed intent and you need a dependable baseline before specializing.
Exact-match catalogue
Use this when a wrong identifier result is more expensive than a broader discovery miss.
Variant-rich retail
Use this when option identity, availability, and product-card context determine whether a result is usable.
Discovery-led catalogue
Use this when shoppers describe a look, use case, or need instead of naming one known product.
Technical specs
Use this when dimensions, compatibility, standards, and structured fields must remain precise.
Content-rich products
Use this when descriptions and guidance are part of the path to choosing a product.
Chapter 2 · The six profiles
Choose by query risk, not by industry label
Balanced commerce
Mixed catalogs that need a dependable starting point across exact and descriptive queries.
Prioritizes
Balanced relevance, practical typo handling, recovery, and conventional commerce display.
Test first
A brand query, a category query, one identifier, one misspelling, and one descriptive query.
Exact-match catalogue
Parts, wholesale, replenishment, or technical catalogs where identifiers carry strong intent.
Prioritizes
Precision, strict identifier behavior, conservative recovery, and stable product identity.
Test first
Full and partial SKUs, punctuation variants, model numbers, barcodes, and near-collision identifiers.
Variant-rich retail
Apparel, beauty, footwear, and other catalogs where options decide whether a result is useful.
Prioritizes
Variant-aware presentation, swatches, availability, and option-level product handoff.
Test first
Color, size, material, style, unavailable options, and product-versus-variant identity.
Discovery-led catalogue
Visual or lifestyle catalogs where shoppers describe a need rather than type a known item.
Prioritizes
Meaning-led discovery, broader recovery, visual browsing, and useful result diversity.
Test first
Use-case phrases, aesthetic language, long queries, broad categories, and adjacent intent.
Technical specs
Catalogs where dimensions, compatibility, standards, materials, or structured attributes drive fit.
Prioritizes
Specific field evidence, attribute precision, filters, and conservative compatibility claims.
Test first
Measurements, units, compatibility codes, specifications, and category-specific filters.
Content-rich products
Products supported by detailed descriptions, ingredients, use cases, manuals, or editorial context.
Prioritizes
Richer descriptive retrieval while preserving product identity and commerce constraints.
Test first
Benefits, ingredients, care, application, audience, and detailed problem statements.
See the choice in a real query, not an industry label
The profile name is only a starting hypothesis. The same store can need a different posture for a known identifier and a descriptive discovery query, so use a small fixture set to test the trade-off before publishing.
Catalogue
Replacement parts and wholesale
Fixture query
“AB-120 or a compatibility code”
Starting profile
Exact-match catalogue
Why
The expensive mistake is a plausible but incompatible result. Start strict, then test real punctuation, variant and near-collision cases before allowing broader recovery.
Catalogue
Fashion with many options
Fixture query
“black linen wrap dress, size 12”
Starting profile
Variant-rich retail
Why
The product is not useful unless colour, size, availability and the selected variant survive the result and product handoff.
Catalogue
Home and lifestyle discovery
Fixture query
“warm minimalist lamp for a small room”
Starting profile
Discovery-led catalogue
Why
The shopper has expressed a use case rather than a known item. Test useful breadth, diversity and the ability to narrow without losing the original intent.
Catalogue
Mixed catalogue with unclear risk
Fixture query
“a brand term, a SKU, a typo and a broad category”
Starting profile
Balanced commerce
Why
When evidence is mixed, a balanced baseline is safer than choosing an aggressive posture and discovering later that one important query family was sacrificed.
Balanced commerce is the responsible fallback when the evidence is mixed. Do not choose a more aggressive posture because its name sounds sophisticated. Choose it when your fixture queries show a consistent need and the failure cost is understood.
Chapter 3 · Select and verify
Build the test set before changing the profile
Collect real query fixtures
Use buyer language, exact identifiers, variants, misspellings, broad discovery, and a query that should stay empty.
Describe success before choosing
Record the expected product identity, acceptable alternatives, maximum useful position, and required variant state.
Choose the closest operating posture
Select the profile that matches the costly mistakes your store must avoid, not the profile name that sounds like your industry.
Preview the full fixture set
A profile changes a bundle of behavior. Judge the whole set before publishing any single improvement.
Adjust only the proven edge cases
Keep the profile as a coherent baseline, then use focused controls where catalog evidence justifies them.
The fixture set should include known-good searches, not only failures. A change that repairs a broad descriptive query but pushes a high-value exact SKU down the page is not a clean win.
Use ParticleSearch previews and protected searches to compare expected product identity and position before publishing. Replay the same protected set so the profile change is judged against equivalent inputs instead of a remembered result. For query-specific changes after the profile is stable, use the merchant query tools guide.
The profile governs search behavior, while the storefront still determines how products, variants, filters, and actions are presented. Use the ParticleSearch storefront search guide to review that separate layer.
Chapter 4 · Keep the boundaries clear
Do not use a profile to solve a different layer
Profile vs layout
A profile changes search behavior. The storefront experience controls density, visual presentation, filters, and product-card decisions.
Profile vs catalog repair
A profile cannot create missing product fields, repair inconsistent identifiers, or infer compatibility your source data does not contain.
Profile vs ranking rule
A profile is a store-level starting strategy. A ranking rule is a focused intervention for a specific query or business need.
Profile vs synonym
A profile governs matching posture. A synonym says two reviewed terms can act as true substitutes in a defined context.
Profile vs experiment
A profile selection is a configuration decision. An experiment compares controlled variants when the better option is uncertain.
Chapter 5 · Trial acceptance checklist
A profile is ready when the critical query set is explainable
Exact identifiers return the intended product or variant without loose substitutions.
Descriptive queries have useful breadth without drowning precise matches.
Misspellings recover only when the correction is commercially reasonable.
Out-of-stock and discontinued products follow the store policy.
Variant-rich results preserve options and availability through product handoff.
Filters match the fields shoppers need for the selected catalog posture.
A query that should remain empty does not receive an invented recovery.
Meaning-led candidates outside the profile quality boundary do not enter merely because keyword evidence is weak.
Protected searches still pass after the profile is published.
Operating example · exact identifier profile
Choose the profile that protects the expensive failure
For a technical catalogue, test “AB-120” as an exact identifier and “replacement bracket for AB-120” as a descriptive query. The profile should be judged across both jobs, not by one attractive aggregate.
Expected evidence
Exact AB-120 returns the intended product or variant first, near-collision identifiers do not substitute loosely, and the descriptive query has useful breadth without displacing the exact answer.
Weak evidence
The profile raises clicks for descriptive queries, but the card or product page hides the option, availability, or identifier. The posture may be plausible while the storefront handoff remains unresolved.
Failure evidence
AB-120 returns AB-121, a no-product query is given an invented answer, or the profile broadens exact matches to reduce empty results. This is a profile failure even if aggregate coverage improves.
Next decision
Choose Exact-match catalogue when precision is the dominant risk, keep the fixture set protected, and send vocabulary or single-query issues to query tools after the profile is stable.
Strongest alternative: use Balanced commerce when the catalogue has mixed evidence or a real discovery requirement. Do not choose an aggressive profile merely because it promises broader recovery.
Acceptance judgement
Accept a profile only when the critical exact, descriptive, typo, availability, variant, and empty-query fixtures are explainable and the storefront preserves their intended evidence. A profile that improves one metric by sacrificing an expensive query class is not accepted.