Alternatives and Compatibility: Structuring the 'Dupe' Relationship

Consider a composite example: a Rotterdam DTC team sells robot-vacuum filters from a New Jersey 3PL. Its page says the part “fits most popular robot vacuums.” Asked which filter fits a 2022 unit, an assistant returned two marketplace listings and a forum thread; the team appeared once, as “other compatible filters are also sold.” Support answers it forty times a week.

A dupe is not an adjective but a relationship record — what replaces what, which models it fits, under which conditions, on what evidence — and a page missing any part gets quoted as “other compatible options.”

Three relationships, three field sets

Substitute, compatible, and accessory are three questions, and “fits most models” answers none of them.

Relationship	Buyer asks	Fields that must exist	Fails as
Substitute	What else works?	Object replaced, generation, dimensions of difference, exclusions, warranty position	“Third-party alternative,” no comparison
Compatible	Does it fit my X?	Fit list by year, version, region; interface and dimensions; explicit does-not-fit list	Claim skipped; answer routes to a retailer with a list
Accessory	What else do I need?	Host product, replacement interval, box contents, substitutable parts	“May need accessories,” never bought

One field set leaves half the cells wrong: a substitute answers worse where, compatibility answers yes, no, or yes if, an accessory does the host work without it.

Why “fits most models” is dropped

A model does not process “most”; it processes enumerations and conditions, so a fuzzy quantifier is downgraded to a disclaimer.

The buyer’s question has three valid answers: yes, no, yes if. A page offering none leaves one sentence it can stand behind — “claims broad compatibility; confirm before purchase” — true and worthless.

The follow-up carries more value: buyers ask whether a part will damage the machine, and the model hunts for boundaries — which models are excluded, under what condition. “Not compatible with 2023+ housings; the lid geometry changed” gets lifted intact and blocks the return before the order exists.

Model names break it too: one unit sells under different names in the US, Germany, and Japan, and a record keyed on one spelling fails to resolve. Aliases are a column — see the alias table for model numbers.

Example: one record, before and after

Example — brands, models, and figures are invented; no real product is described.

Before: “Brand A replacement filter, compatible with most robot vacuums, H13 grade, the ideal OEM alternative.”

After:

Field	Value
Relationship	Substitute, compatible
Subject	Brand A HEPA filter, SKU AF-3H
Replaces	OEM filter, 2020–2022 generation
Fits	2020–2022 standard and pet editions; 2021 serials with prefix B need an adapter ring
Does not fit	2023 and later housings; slim edition
Difference	11 oz vs 9 oz; 6-month interval vs 12; $24 vs $38
Evidence	Fit-test method and date, housing drawing, independent teardown

What matters is not “better” but decidable: fits, does not fit, fits at a cost.

Build the record first, the page second

The relationship table is the asset; pages, feeds, and markup are views of it. Three copies, three answers to maintain.

Step	Do this	Owner	Passes when
1 Collect phrasings	Thirty real fit questions from support, site search, and assistants, sorted by type	Support, content	Each maps to one type and one URL
2 Define fields	Five per type, plus an alias column	Category, engineering	Name hidden, fields still say who fits what
3 Build the master table	One table as source of truth; pages, feed, markup generated from it	Engineering	All three agree value for value
4 Render	Fit and exclusion lists on the PDP; model filters on category pages	Content, front end	Reads with JavaScript off
5 Verify	Table alone into a model, three fixed questions	Content	Correct twice in a row

Step 3 decides whether this survives: relationships scattered across copy, FAQ, and support macros expire; a master table means a model change edits one row. Use native schema.org properties where they exist — isAccessoryOrSparePartFor, isConsumableFor, isSimilarTo — and put conditions in additionalProperty under one site-wide key.

Step 5 is the only falsifiable one: will it fit my X, are there versions it does not fit, how does it differ? A missing answer is a missing field; a confident wrong one, an unaligned alias.

Naming a competitor’s model without borrowing it

Model names belong in fit and exclusion statements, not headlines; the closer the wording gets to “official” or “equivalent,” the more the record becomes a liability.

• Descriptive use only: the name in “fits” is necessary; in a title it trades on someone else’s reputation.
• No equivalence claims: “just like OEM” is unverifiable and becomes evidence in a complaint; compare capacity, weight, interval.
• Print identity and disclaimer on the page: independent third party, not the manufacturer; trademarks belong to their owners.
• Have counsel read it: rules differ across the US, Germany, and France, and a rival’s page may just never have been challenged.

Silence forfeits the question; a wrong detail costs returns, and the “does not fit” line works as a brake.

Where this breaks

• Safety-critical categories. Batteries, chargers, gas fittings, child seats, medical parts: publish official fit and exclusions, delete “universal.”
• No accessory ecosystem. One SKU, no consumable, no host: the table is pure cost.
• Revenue on the marketplace. Platform fields and reviews do the work; the site table comes second.
• No named owner. Discontinued models and batch revisions expire the table; no maintainer, no launch.
• Undecidable fit. Custom and modified parts, fragrance “dupes” where similarity is opinion: write “confirm with us,” not compatible.

A record counts only once the entity resolves: check whether your brand entity exists. The gate into the shortlist is covered by this issue’s lead, AI shortlists only 3–5 names; the layer sits on the technical foundation hub; the volume traces back to the four AI search entry points.

This month, open the page behind your most-asked compatibility question, delete “fits most models,” and replace it with five lines: three models that fit, one that does not, one precondition. Re-ask in two weeks.