> ## Documentation Index > Fetch the complete documentation index at: https://docs.clarvo.ai/llms.txt > Use this file to discover all available pages before exploring further. # Search filters > Every filter in Clarvo, what it does, and whether it narrows your results (hard) or shapes their ranking (soft). Filters refine a search until the results match who you're looking for. Open **Edit Filters** from your search to access them, grouped into the tabs below. ## Hard vs. soft filters Filters work in one of two ways, and knowing which is which is the key to widening or tightening a search: **Narrows** your results. Candidates that don't match are removed from the list. **Shapes ranking**. Matching candidates move up the list, but non-matching candidates can still appear. Clarvo matches these by *meaning*, not exact text. Several criteria you might expect to be strict (job titles, industries, and a candidate's field of study) are deliberately soft. Clarvo matches them semantically and uses them to rank candidates, so a strong match with a differently-worded title or a related industry still shows up instead of being filtered out. ## Job title & role Clarvo matches titles by **meaning**, not exact wording, and uses them to **rank** candidates. Someone whose title is phrased differently but does the same work still appears. Narrows to the seniority levels you select, with a little tolerance for the levels just above and below. Candidates well below your minimum are filtered out; the exact amount of experience also influences ranking. You can set a maximum too. Narrows to candidates whose average time per role meets your minimum. **Limit to** or **exclude** current freelancers, useful when sourcing for contract work. Preferences like remote work shape ranking rather than strictly excluding candidates. ## Industry Industry experience is matched **semantically** and shapes ranking. A candidate from a closely related industry still appears, ranked by how relevant their background is. ## Location Keeps only candidates in the cities or countries you choose. Add several locations for a broader or remote-friendly search. ## Skills Skills are the most flexible filter: they can narrow **or** just rank, depending on the mode and how you group them. In **Loose** mode the skills filter is broad: it splits your skills into two buckets, *required* and *nice to have* (below). In **Strict** mode it becomes a hard filter built around your skill groups (Must / Can groups, further below). ### Required vs. nice to have in Loose mode In Loose mode your skills sit in two buckets, and they behave very differently. Drag a skill between buckets, or use the **Add skill** button in each, to decide which is which. Shown under **"Must have at least one of these skills."** Clarvo keeps only candidates who match **at least one** of these skills, no matter which you list. It's an *any-of* match (OR logic) across the whole bucket, not an *all-of* one, so adding more required skills **widens** the pool rather than narrowing it. Leave this bucket empty to skip skill-based narrowing entirely. Shown under **"Preferably has some of these skills."** These never exclude anyone, they only **lift** matching candidates in the ranking. A search with *only* nice-to-have skills (no required ones) applies no skill filter at all and simply ranks everyone by how well they match. Loose mode's two buckets are the same idea as Strict mode's **Must** and **Can** groups, just simpler: one shared "any of these" requirement instead of several groups that each must be matched. Switch to Strict when you need a skill from *every* group, not just one skill overall. ### Skill groups as boolean logic (Strict mode) Strict mode turns your skills into a boolean expression. You build it by sorting skills into **groups** and marking each group **Must** or **Can**: * **Within a group, skills are OR.** A candidate qualifies for the group by matching *any one* of its skills. Put interchangeable skills together, like `React` OR `Vue` OR `Angular`. * **Between Must groups, the logic is AND.** A candidate must satisfy *every* Must group to appear. Use a separate group per distinct requirement. So three Must groups read like `(React OR Vue) AND (TypeScript) AND (AWS OR GCP)`: a frontend framework, *and* TypeScript, *and* a cloud platform. Drag skills between groups, or use **Add new group**, to shape the expression. Loose mode is the special case where every skill shares one big OR. In Strict mode, a candidate must match at least one skill from **every Must group** to appear (OR within the group, AND across groups). **Can** skills sit outside the AND requirement: they only affect ranking, lifting matching candidates without excluding anyone. Setting a desired minimum experience for a skill (the clock icon) influences ranking. ## Companies Keeps only candidates who have worked at the companies you choose (at any point in their history), great for targeting competitors. Removes candidates who **currently** work at the companies you choose. Switch the recency to also exclude anyone who worked there in the past. Narrows by how large a candidate's company is, or how recently it was founded. ## Education Narrows to candidates at or above a minimum degree level (or, optionally, excludes anyone with a higher degree). Narrows by when candidates finished their most recent education. Matched by meaning and used to **rank** candidates. A related field still appears, it just ranks accordingly. Treated as preferences that inform ranking rather than strict cut-offs. ## Languages Candidates must speak every required language (at a reasonable proficiency). A bonus that improves ranking but never excludes anyone. ## Interaction filters These narrow results based on your team's prior activity, so you don't reach out to the same person twice. Each can be scoped to **this project** or **your whole organization**. **Limit to** or **exclude** candidates you (or your team) have already contacted or viewed, optionally within a recent time window. **Limit to** or **exclude** candidates who are already LinkedIn connections. ## Saving filters as templates Built a filter set you like? Save it as a **template** to apply the same criteria to future searches in one click. Your filters also persist as you work, so you won't lose them when refining a search. Too few candidates? Loosen the hard filters first. Switch skills to Loose, move a Must group to Can, widen seniority or experience, or remove a location, then turn on [deep search](/search/deep-search). Soft filters (titles, industries, field of study) never need loosening; they only affect ordering.