Marketplace Alert Keyword and Exclusion Rules

The most useful marketplace rule is one you can explain with an actual listing. Before excluding a word or requiring a model number, ask which unwanted post it removes and which good post it might hide.

This guide separates marketplace keywords from Classifindr rules, explains the matching choices that often cause surprises, and gives you a practical tuning sequence. The goal is fewer unusable matches without turning a live buying search into a silent one.

Separate discovery from the buying decision

Start with a buying brief: the item you need, acceptable versions, complete budget, usable location, and condition you can assess. Keep optional preferences outside the non-negotiables.

For a used camera, the brief might be a working body with a particular lens mount, a battery and charger, and collection within your available time. Sellers may omit the charger from the title even when it is included. Requiring every detail in title rules would reject candidates before you could ask a sensible question.

Use the main marketplace keyword to discover relevant stock. Use supported price, category, condition, and location controls where they fit. Add Classifindr rules for wording that reliably separates useful titles from noise. Check details that are absent or ambiguous on the original listing.

Use seller phrases, not a specification pasted into keywords

Marketplace keywords and Classifindr rule patterns are different inputs. Do not assume comma syntax in a rule has the same meaning in every marketplace’s search box.

Choose a recognisable discovery phrase such as aeron chair or makita drill. If sellers regularly use another name, try it as a separate search rather than assuming the source will interpret a long list of aliases as intended. A sofa search and a lounge search may find different candidates even when both answer the same buying brief.

Searches should be split because seller language or buying intent differs, not merely because you want more alerts. The search organization guide helps keep those purposes visible.

Read comma-separated rules as alternatives

In a Classifindr rule, comma-separated terms match as alternatives. Matching is case-insensitive. An include pattern of sofa, lounge can pass a title containing either word; it does not require both.

Rule choice	Example pattern	What the rule means	What it does not prove
Include title terms	sofa, lounge	At least one listed alternative appears in the title	The furniture is the right size or condition
Exclude title terms	case only, empty box	A matching alternative can reject the title	All accessories have been removed from the source results
Plain term	pro	The text can match inside a longer word	Only the standalone model word Pro matches
Quoted term	"pro"	The term is matched with word boundaries	The advertised item belongs to the correct generation

Plain patterns use substring matching, which can surprise buyers with short model words. For example, unquoted pro can also match projector. A quoted term narrows that text match, but punctuation, aliases, omitted words, and misleading titles still need review.

Do not use unlocked, 256 as if it means an item must be both unlocked and 256 GB. If both are essential, use suitable source filters where available and confirm both on the listing. A rule is not a substitute for complete-item verification.

Prefer specific exclusions to broad accessory words

Broad blocks are quick to add and expensive to misunderstand. Excluding case can remove a camera bundle whose title says it comes with a case. Excluding parts can hide an item advertised with spare parts. Even broken can match a seller saying that nothing is broken.

Use a phrase that describes the unwanted offer, then test it against useful titles.

Observed noise	More precise exclusion to investigate	Good candidate to check before keeping the rule
Packaging sold without the device	empty box, box only	Complete item described as boxed
Protective cover offered alone	case only	Device supplied with a case
A buyer requesting the item	wanted to buy	Seller wording that happens to mention wanted
Equipment offered for rental	hire only, rental only	An item previously used in a hire business but now for sale
A dismantled vehicle	wrecking, breaking for parts	A running car supplied with spare parts

These are patterns to evaluate, not a universal exclusion pack. If unwanted posts do not use the phrase, the rule will not remove them. If useful posts do use it, the rule can remove those too.

New rule creation supports title and location targets. Older description-based rules may still appear on existing searches, so inspect the target as well as the text when reviewing a surprising rejection. Do not assume an old rule operates on the title alone.

Keep title overrides narrow and deliberate

A matching title override allows the listing through the other text rules. It can rescue valuable wording, but it can also admit a title that your exclusions were intended to reject.

Suppose Aeron is an override and the title is Aeron chair parts only. The model word can bypass your text exclusion for parts only. That is not a working chair, even though the brand is right.

An override does not bypass the entire search or certify the item. Other filters still apply. Prefer repairing an overbroad exclusion before adding a popular brand as an override. Keep an override only when you understand the rejected examples it is meant to rescue and the unwanted examples it might admit.

Treat location rules as wording checks

Location rules inspect the listing’s location text. They do not calculate whether the item is inside your collection radius. Listings with no location text pass that rule scope, so missing location wording is not evidence of a nearby seller.

Use the marketplace’s supported geographic controls and verify the source page. Facebook searches use the source’s anonymous default area rather than a custom Classifindr radius. An include term naming your suburb cannot make every returned listing local.

The pickup-radius editorial explains why usable collection area matters more than a larger result count.

Reserve AI relevance filtering for context

AI relevance filtering can express a buying intention that is awkward to capture with short title terms. For example: look for a complete coffee machine offered for sale, not replacement accessories or a servicing business.

Use it to help judge relevance, not to authenticate products, establish ownership, diagnose hidden faults, or prove that every accessory is present. Your buying brief still determines which source details need checking. Missing evidence remains missing even when a match looks promising.

Tune with accepted and filtered examples

Use this sequence when repairing a noisy search:

1. Keep several useful and unwanted examples long enough to compare their wording.
2. Check whether the mismatch comes from source keywords, price, location, a rule target, or AI relevance.
3. Make one meaningful change, such as narrowing an exclusion from box to box only.
4. Review accepted matches and filtered listings with their match reasons.
5. Confirm the change removes the observed noise without hiding a useful example.
6. Split the search when different aliases or buying goals cannot sensibly share the same rules.

A quiet period alone does not prove a successful rule edit. Stock may be scarce, source access may vary, or the search may be paused. Check its status before repeatedly loosening rules. The search experiment article offers a more structured way to compare changes.

Choose check speed and alert channels after the search becomes useful. A faster check cannot repair ambiguous seller wording. If alerts appear to show the same physical item again, investigate the relisting problem before adding exclusions that might hide genuinely useful stock.