A search that returns nothing — or skips someone you know is in your system — almost always has a plain cause. This page is the short list of them, with the fix for each. Work down it in order; the first two cover most cases.
The resume never parsed
Search matches the text on a candidate's record. That text comes from resume parsing, which extracts structured fields and body text from each resume on upload.
When parsing has nothing to work with, search has nothing to match. The usual culprit is an image-only PDF — a resume that is a scan or an exported image rather than real text. Parsing can't read it, the extracted text is mostly empty, and every search mode will miss that candidate.
Fix: open the candidate and check whether their fields are populated. If the record is blank, the resume didn't parse. Re-upload a text-based version of the resume, or fill the key fields in by hand so search has something to match.
The query is too narrow
A boolean query with several AND clauses stacked together is the most common way to get zero results. Each AND is another condition every candidate must satisfy at once.
"staff engineer" AND rust AND kubernetes AND fintech AND remote
That asks for one person who satisfies all five. You may not have them.
Fix: loosen it. Turn the must-haves into nice-to-haves with OR, or drop the weakest clause and add it back only if the result set is too big. Boolean is also literal — kubernetes will not match a resume that says "K8s". The boolean syntax page covers this.
You're looking at the wrong scope
Search runs inside whatever you have open. If you are scoped to a single job, the search only covers that job's candidates — not everyone in your system. A candidate who applied to a different role won't appear.
Fix: check the scope. If you meant to search everyone, move out of the single-job view before running the query.
You expected a saved search to be frozen
If you re-opened a saved search and the people changed, that is the design, not a bug. A saved search is a query, not a fixed list — re-running it scores your current pool, so results move as candidates come and go.
Fix: if you need an unchanging set of people, use a list instead. Saved searches explains the difference between the two.
In deep search, the candidate missed the recall set
A deep search scores at most 150 candidates. Before scoring, the recall phase narrows your pool down to that cap. A candidate who exists in your system but didn't rank into the recall set is never scored — so they won't appear in the results, even though they match.
This happens on broad queries against a large pool. Recall keeps the strongest 150 and drops the rest.
Fix: narrow the query. Add a location, a seniority, a tool, a domain — anything that pushes the candidate you want higher in recall. How a deep search run works covers the cap in detail.
Still nothing?
If the candidate's record looks complete, the query is loose, the scope is right, and they still don't surface, the catch-all is wording. Search responds to how you describe the role — try a plainer phrasing, or switch modes. Smart search reads intent; boolean matches exact text; deep search reads the resume body. One of the three usually sees what the others missed.