Manual Overrides: The Merchandising Log Nobody Reads

Somebody pinned a product to the top of search results six months ago, and nobody remembers why. The override is still there, the product is out of stock, and the ranking it broke has been broken ever since.

Search merchandising is the practice of applying business intent on top of relevance ranking: instead of letting the algorithm decide what appears in slot one, you override specific results for specific reasons. Four controls cover most cases. A boost surfaces a product higher. A bury sinks it. A pin locks it to a specific position. A hide removes it from results entirely. Each rule fires only on the queries you target, and the goal is to merchandise as little as the business will allow, reaching for a control only when there is a reason the ranking could not have known.

Four controls, and the one that causes the most damage

The controls are simple, which is why they get misused. A boost raises a product's effective rank on matching queries without locking it to a specific slot, so it still competes on relevance. A bury lowers it. A pin fixes a SKU at an exact position, and a hide removes it from results entirely. Pinning is the dangerous one. Kibo's documentation warns against pinning more than 10 or 20 products, because boosting is more efficient and has lower impact on system load when you want to promote a larger set. A pin also interacts with manual ranking in a way that surprises people: changing the ranking of one result locks all results above it, while pinning does not lock the results above the pinned item, so a pinned product can survive a re-ranking that moves everything else.

The damage comes from pins that outlive their reason. A pin set for a seasonal campaign stays through the season, into the next one, and onto a product that is now out of stock or discontinued. The ranking underneath has moved on, the inventory has moved on, and the pin has not, because nobody owns the log. A store with a hundred pins is a store whose search ranking is a museum of past campaigns, and the shoppers are the ones paying admission.

Related reading on this site: RelevanceRail and Search Ranking for Long-Tail Queries: When the Best Match Has No Sales History cover the neighbouring problems in more depth.

Write the rule for the reason, not for the product

The triggers that justify a rule are business events, not preferences. Seasonal campaigns, new-arrival launches, overstock clearance, margin prioritization, and out-of-stock suppression are the standard list, and each one has a natural expiry. A clearance push ends when the stock clears. A launch ends when the product has its own ranking. A margin push ends when the margin changes. A rule written for a reason that has an end date can be retired on that date. A rule written because someone liked a product has no end date, and it will still be there in a year.

Every rule should carry its reason in the open. The merchandising log should record who set the rule, which queries it fires on, what it does, and when it expires. A rule without an expiry date is a rule that will be forgotten, and a forgotten rule is a ranking defect with a name on it. The audit is a monthly habit: list every active override, check whether its reason still holds, and retire the ones that do not. A store that audits its overrides monthly has a search ranking that reflects the business today. A store that never audits has a ranking that reflects the business six months ago.

When the override is the symptom, not the fix

Merchandising rules sit on top of the ranking layer rather than replacing it, and they are there for the genuine business exceptions, not to prop up search that is not surfacing the right products on its own. If the team is writing overrides every week to fix the same query, the ranking is broken, and the overrides are hiding it. A boosted product still needs some relevance to the query to surface at all, so boosting a completely off-topic SKU will not force it into results, and a pin on a query the product does not match is a confession that the catalog or the ranking needs work.

The test is the count. A healthy store has a small number of active overrides, each with a named reason and an expiry date, and the number goes down over time as the ranking improves. A store with a growing override count has a ranking problem wearing a merchandising costume, and the fix is in the ranking, not in the next override. The log is the diagnostic: read it, count it, and ask whether each entry is a business decision or a workaround. The workarounds are the ones to fix first.

Frequently asked questions

What are the four search merchandising controls?

Boost, bury, pin, and hide. A boost surfaces a product higher, a bury sinks it, a pin locks it to a specific position, and a hide removes it from results entirely. Each rule fires only on the queries you target.

How many products should I pin in search results?

Kibo's documentation recommends against pinning more than 10 or 20 products. Boosting is more efficient and has lower impact on system load when you want to promote a larger set.

What triggers justify a merchandising rule?

Business events with a natural end: seasonal campaigns, new-arrival launches, overstock clearance, margin prioritization, and out-of-stock suppression. A rule written for a reason that has an end date can be retired on that date.

How do I know if overrides are hiding a ranking problem?

Count them. A healthy store has a small number of active overrides, each with a named reason and an expiry date, and the number goes down over time. A growing override count means the ranking is broken and the overrides are covering for it.