ParticleSearch2026-08-0314 min read

ParticleSearch Ranking Rules: When to Pin, Boost, Move Lower, or Hide Products

A search can contain the right products and still make the shopper work too hard to find the best answer. Ranking is the process that decides the order of eligible results. A ranking rule is a merchant’s deliberate instruction to change part of that order for a defined reason.

The goal is not simply to put a preferred product first. It is to allocate the shopper’s limited attention more usefully while preserving the meaning of the query. That can help a buyer notice a legitimate launch item, find an exact model, or see an available option before a weaker one. It can also damage trust when commercial preference outruns relevance.

That boundary is the whole decision. If waterproof jacket returns eligible jackets but hides a launch set, ranking may be appropriate. If the waterproof field is missing and the right products never enter the result, a rule would only make the evidence gap harder to see. By the end of this guide, you will understand how the base order is formed, when a rule is appropriate, what each action communicates to a shopper, how rules interact, and how to judge whether the change earned its place.

Chapter 1 · Understand what ranking changes

Ranking allocates a shopper’s limited attention inside an eligible result set

A result position is not only a technical number. Early products are seen sooner, become the shopper’s first comparison points, and can be interpreted as the store’s judgement of what best answers the query. A ranking rule deliberately changes that distribution of attention.

This is why the same intervention is not equally safe for every query. A broad query such as rain jacket leaves room for discovery. An exact model query carries a much narrower promise: the shopper has already named the thing they want. Commercial preference should become more restrained as query specificity increases.

Reserves a specific position, even when the underlying order would place the product elsewhere.

Shopper meaning: This is the store’s strongest visible recommendation, so the product must withstand that level of scrutiny.

Boost

Raises the probability that an eligible product appears earlier without guaranteeing one fixed slot.

Shopper meaning: The store is expressing a preference while leaving room for query evidence and other signals.

Move lower

Preserves discoverability but spends less of the first screen on the product.

Shopper meaning: The item may still answer the query, but it should not lead the shopper’s comparison.

Hide

Removes the product from a defined result context rather than merely changing its position.

Shopper meaning: The store is saying the item is not a useful answer here; that claim needs a narrow, reviewable scope.

Chapter 2 · Follow the result order

A ranking rule changes one stage in a longer search decision

Before a product can be moved up or down, it must first be a valid candidate. The catalogue and request context decide whether the product is published, available under the relevant policy, represented by the correct product or variant, and supported by searchable evidence. The search system then evaluates those candidates and creates a normal, or base, order.

The rule sits on top of that base order. It can change how much attention an eligible candidate receives; it does not create the missing evidence that made a product eligible or relevant. That is why a ranking preview must show both the original order and the final order. Without the baseline, the merchant cannot tell whether the rule corrected a small preference or overrode a large relevance disagreement.

Eligible candidates

Which products are allowed to answer this request?

Publication, market, stock policy, product identity, variant state, and searchable evidence determine what can enter the result.

Relevance evidence

How well does each candidate answer the query?

Exact terms, fields, attributes, category fit, phrase evidence, and other retrieval signals create the base judgement.

Base order

Which valid answers would normally appear first?

The search system orders the eligible candidates before any deliberate merchant preference is applied.

Ranking rule

Should a defined merchant preference change that order?

A pin, boost, move-lower, or hide action changes visibility inside a stated query and business context.

Shopper experience

Does the final order still make sense as an answer?

The cards, variant, price, availability, and handoff must support the preference the new position implies.

The practical boundary: when the expected product is absent, investigate the candidate set. When it is present and valid but receives the wrong amount of attention, ranking becomes a sensible layer to inspect.

Chapter 3 · Decide whether order is the problem

Use a ranking rule only when the desired candidate already deserves to answer

A legitimate rule has three ingredients: a valid candidate, a defined shopper or commercial reason, and a scope narrow enough to protect neighbouring intent. The examples below begin with the observed situation because similar-looking result pages can require completely different repairs.

Situation	First diagnosis	Decision	Reason
A relevant launch product is present but buried on a broad category query	The product is eligible, the query evidence is real, and the issue is the amount of early visibility it receives.	A scoped boost may fit.	The rule changes order without inventing eligibility or claiming an exact identity match.
The expected product does not appear anywhere in the result	The candidate set may be missing source data, field coverage, market eligibility, publication, or freshness.	Do not start with ranking.	A rule cannot reorder a candidate the search did not retrieve.
An exact model query returns the correct item below loose matches	Confirm the exact identifier is indexed and the intended variant is eligible before judging order.	A narrow exact-query rule may fit.	The merchant is protecting a precise shopper promise, not applying a broad commercial preference.
A product is relevant but temporarily low in stock	Decide whether stock changes eligibility, only presentation, or the preferred order in this buying job.	A temporary move-lower rule may fit.	The item remains discoverable while more dependable options receive the first attention.

Chapter 4 · Choose the action

Pin, boost, move lower, and hide answer different questions

A ranking control should make a business judgement explicit without pretending it is a relevance repair. Use the smallest action whose side effect you can accept.

Action	Use it when	Boundary
Pin	A small, deliberate set should occupy a defined position for a scoped query.	Do not use it to conceal a missing product or override an exact identifier.
Boost	An eligible product or group deserves more visibility for a business reason.	Do not make a weak candidate look relevant when the query evidence is absent.
Move lower	Keep a product discoverable while putting better-fit or available options first.	Do not treat a temporary stock issue as a permanent quality judgement.
Hide	A product should not appear for a tightly defined context and has no useful next action there.	Do not hide a product globally because one query is poorly configured.

Chapter 5 · Read the query before the rule

A good ranking change starts with a before-and-after question

Suppose the query is waterproof jacket. If eligible waterproof jackets are present but a seasonal launch set is too difficult to find, a scoped boost may be reasonable. If the product is absent because its waterproof attribute is not searchable, ranking cannot solve the evidence gap. If the query is AB-120-BLK, a campaign should not displace the exact variant simply because it is commercially important.

Live

Query → returned products → positions

Record what a shopper sees before the rule.

Draft

Same query → changed order → stated reason

Change one ranking behaviour at a time.

Verify

Exact + nearby query → safe handoff

Check side effects before trusting the rule.

When the question is vocabulary rather than order, use the synonyms and redirects guide. When it is a seasonal business moment, use the search campaigns guide.

Chapter 6 · Read the combined state

Rules interact through scope, product identity, timing, and priority

A preview of one rule in isolation can look correct while the live result is wrong. The storefront evaluates the conditions that are active together: the query, eligible product or variant, market, audience, date, and any competing action. The combined state is the behaviour the shopper receives.

For example, a November boost for a Canadian launch and a permanent move-lower rule for low-stock products may both match one jacket. The final position depends on how those actions combine. If an operator cannot explain the winner before publishing, priority is part of the unfinished work.

Layer	Question	Failure if ignored
Query scope	Which exact query, phrase family, or collection context activates the rule?	A broad phrase rule can unexpectedly catch an exact model or a different shopper job.
Product scope	Does the action target a product, variant, collection, or attribute-defined group?	A product-level preference can surface the wrong variant or an unavailable market offer.
Audience and time	Who receives the preference, and during which valid window?	A local or temporary judgement can become global ranking debt.
Priority and order	Which action wins when more than one eligible rule applies?	Two individually sensible rules can produce a combined result that nobody intended.

Chapter 7 · Record the boundary

The rule should tell the next operator what it is allowed to change

Expected set

Name the product, variant, category, or acceptable alternatives the query should return.

Disqualifiers

Record incompatible models, unavailable variants, unrelated categories, and exact identifiers that must stay precise.

Scope

Write the query, collection, market, audience, device, or experiment context where the rule applies.

Lifecycle

Record owner, priority, effective dates, review date, and the condition that ends the rule.

Do not confuse a successful publish with a successful rule. The storefront still needs a runtime check, and the query still needs to satisfy the intended product and variant promise.

Chapter 8 · Judge the outcome

Measure the intended movement and the relevance cost around it

First record the counterfactual: the result order, availability, and relevant shopper behaviour before the rule. After publishing, confirm that the intended products moved, then inspect protected exact queries and nearby phrases that should remain unchanged. A target-query improvement is incomplete if it creates a wider relevance regression.

Clicks and orders are useful observations, but they do not prove that position alone caused the change. Price, stock, imagery, seasonality, traffic mix, and promotion may move at the same time. Use a nearby untreated query, a prior comparable period, or an experiment when the commercial risk justifies stronger causal evidence.

Target

Did the intended eligible product move by the amount and in the context described?

Guardrail

Did exact, nearby, mobile, market, and variant checks preserve the shopper promise?

Decision

Does the evidence support keeping, narrowing, pausing, or retiring the preference?

Chapter 9 · Publish and verify

A ranking rule is ready when the changed query and protected queries both pass

1
Capture the live result for the target query and a nearby query.
2
Choose one action and write the intended before-and-after difference.
3
Preview the rule with exact identifiers, availability, and variant handoff in the test set.
4
Check overlapping rules and publish only with a clear scope and priority.
5
Re-run the same queries after the storefront has loaded the published state and record the outcome.

ParticleSearch’s publish-safety workflow gives this process a merchant-visible place for drafts, previews, runtime state, protected queries, and history. It does not decide whether the business judgement is right; it keeps the evidence attached to the change.

Run the acceptance test

The smallest safe ranking change is the one you can explain and reverse

Pass

The intended eligible products move as described, exact queries remain precise, variant handoff is correct, and the rule has a clear owner and end condition.

Pause

The rule changes unrelated queries, promotes unavailable products, conflicts with another rule, or cannot be tied to a defined shopper or commercial decision.

Use ranking rules to express a scoped preference after the result is eligible and understandable. Use the merchandising and campaigns operating guide when the change also needs audience, schedule, priority, and expiry.