Niche-aware discovery: how the search queries get written

Ask ten people to find link opportunities for a project management tool and you will get ten versions of the same search: “best project management software.” The results are the same twenty articles, every one of them pitched weekly by every competitor in the category, most of them run by publishers with a rate card rather than an editor.

The opportunities worth having are usually one step sideways from the obvious keyword. Query planning is how discovery gets there deliberately rather than by luck.

It starts with your own site, not a keyword

Before any search runs, Dice reads your website and builds a profile of the market you actually operate in. That profile is deliberately structured into five different vocabularies, because they behave differently in search:

A single keyword collapses all five into one. Keeping them separate is what lets discovery reach an article about a workflow problem that never names your category once, and is nonetheless exactly the kind of page your product belongs in.

Four query strategies

The profile is expanded into concrete search queries, and each query is tagged with the strategy it came from. The tag is not cosmetic — it follows the candidate through the entire pipeline.

1. 01

   Category queries

   The direct route: pages about your category, its leading tools, and the choices buyers make within it. These find the obvious roundups. They should be searched — they just should not be all you search.

2. 02

   Blog and resource queries

   Pages that exist to collect useful things: resource lists, curated guides, tool directories, “further reading” sections. These are structurally receptive to an addition, because listing things is what they are for.

3. 03

   Competitor queries

   Pages that name your competitors. This is the strategy that finds the article listing four alternatives to a rival where you are conspicuously missing. You confirm the competitor list first, so the search reflects your real market rather than a guess.

4. 04

   Long-tail queries

   Specific, low-volume phrasings drawn from the problem and audience vocabularies. These reach independent blogs and small publications that keyword-volume tools rank as worthless and editors answer personally.

What gets thrown away immediately

Search results arrive with a lot of noise attached. Before anything is scored, candidates are normalized and filtered on cheap, unambiguous rules:

• Your own domain, which cannot link to you
• Blocked domains
• File downloads rather than pages
• Account, login, and other utility pages
• Search result pages, which are not editorial content
• E-commerce product listings
• Results that are clearly irrelevant to the profile

URLs are also normalized before deduplication, so the same article arriving from three queries with three different tracking parameters becomes one candidate with three pieces of evidence — not three candidates competing with each other.

Evidence is the point

Every surviving candidate carries a record of how it was found: each query that surfaced it, the strategy that query belonged to, its rank in those results, and the competitor seed if a competitor query found it.

That record is used, not merely stored. When enrichment decides which candidates are worth the expense of reading, discovery evidence is a direct input: a page found by several distinct queries earns a bonus, a page found across several different strategies earns another, a page found by a competitor query earns another, and a page that ranked in the top few results earns another still. Converging evidence is a genuinely better signal than a single strong hit, and scoring it explicitly is how that belief becomes reproducible instead of intuitive.

A page found by one query at position 40 and a page found by four queries across three strategies are not the same candidate, and should not be treated as one.

Why smaller publishers are a feature

Long-tail and audience-term queries surface a lot of small sites: one-person blogs, company blogs at ten-person startups, niche newsletters, community resource pages. It is tempting to filter them out on traffic.

Dice does not, because reachability and relevance both run the other way. The independent blogger writing carefully about your exact problem reads their own email, has no rate card, and links because your product is genuinely useful to their reader. The metric-heavy publication has a submissions form and a media kit. Popularity alone is not treated as quality — publisher intelligence classifies the site so you can make that call with the facts in front of you.

What happens to a candidate after discovery — which ones get read, and what reading them produces — is covered in what page enrichment extracts.