Merchandising Rules

Boost and Bury Rules · Merchandising Overrides

Definition

Merchandising rules are manual business rules that adjust algorithmic sorting and recommendations by promoting, excluding or pinning specific items.

Why merchandising rules exist

Algorithmic recommendations optimise for a metric, usually conversion or revenue. But a business
has goals the metric cannot see: launching a new arrival, pulling a defective item, promoting a
partner’s exclusive collection.

Merchandising rules are the control surface between business logic and the algorithm. They let
category managers and marketers govern the output without code.

The main rule types

Rule type	Action	Example
Boost	Lift in the ranking	New arrivals always in the top three
Bury	Push down	Items rated below 3.5 go to the end
Pin	Fix a position	The deal of the day is always first
Exclude	Remove	Drop out-of-stock items and items with no photo
Include	Force in	A promotional item is always shown

How rules interact with the algorithm

The typical architecture: the algorithm generates a ranked candidate list → the rules adjust the
order or the composition → the final output reaches the shopper.

Algorithm: [A, B, C, D, E, F...]
Rule: pin item X to position 1
Rule: exclude out-of-stock (removes C)
Result: [X, A, B, D, E, F...]

Important: the more hard rules there are, the less room personalization has. If the first five
positions are pinned, the algorithm only influences what is left.

When to use them and when to hold back

Rules are justified for:

Promotions with fixed positions (Black Friday, a collection launch)
Technical constraints (removing items with no photo or no price)
Partner commitments (a guaranteed share of impressions for a brand)

Rules create risk when:

There are too many of them and they conflict with each other
They are permanent rather than time-bound, blocking the algorithm from adapting
No A/B test confirms the business effect

Frequently asked questions

What is the difference between merchandising rules and algorithmic recommendations?

The algorithm optimises for a metric — clicks, conversion, revenue. Merchandising rules add business logic on top: promotional positions, partner quotas, availability constraints. The rules are where a human overrides the machine.

How do you avoid a conflict between rules and the algorithm?

Apply rules narrowly — only where the business requirement cannot be expressed as a metric. The more rules there are, the less the algorithm has left to optimise. A rule that says the top five are always new arrivals effectively switches personalization off for those slots.

Should merchandising rules be A/B tested?

Ideally yes. A rule that lifts high-margin products can raise margin and depress conversion at the same time, leaving total profit lower. A test shows the real effect on revenue and gross margin together.

Gravity Field