Every recruiting team has them: the search that's been open a quarter, the hiring manager who's rejected forty profiles, the role that's burned through two sourcers. Hard-to-fill roles rarely stall for lack of effort — they stall because the search is misdesigned. Diagnose before you source harder.
Step 1: size the market before arguing about it. Most stalled searches die here: the brief describes 200 people on earth and demands 5 in your metro at your comp band. Run the actual numbers — SeekOut and Findem produce talent-market maps (how many people match, where they are, what they earn) that turn "we can't find anyone" into "there are 340 matches, 60 in-range, and we've already contacted 45." That reframes the conversation from sourcing effort to search design.
Step 2: brief surgery. Separate the requirements that predict success from the ones that describe the last person in the seat. The highest-yield cuts, in order: years-of-experience thresholds (weak signal, big filter), specific-tool requirements where adjacent tools transfer, degree requirements, and location constraints that predate remote reality. Each cut multiplies the candidate universe; three cuts often 10x it.
Step 3: switch evidence channels. Stalled searches are usually LinkedIn-exhausted — everyone findable there has been InMailed twice. Go where the skill leaves evidence: GitHub and conference talks for engineers (AmazingHiring aggregates this), patents and publications for research roles, community forums for niche skills. Declared-intent pools (Wellfound for startups, Dice for tech) reach people who are actually moving.
Step 4: change the outreach economics. Response rates on over-fished talent pools collapse — the fix is warmer paths, not more volume. Referral mapping (Teamable mines employee networks for warm introductions) and multi-quarter nurture (Gem) out-convert cold sequences by wide margins for scarce profiles. For volume coverage of the searchable market meanwhile, an agent (Noon) keeps the pipeline moving while humans work the warm paths.
Step 5: when all else fails, change the job. If the market map says the person you want earns 30% more than your band, the sourcing problem is a compensation problem. If nobody has all five skills, the role wants to be two roles or the team should train for one of them. The recruiter who brings market data to that conversation stops being the person who "can't fill the role" and becomes the person who fixed it.
The meta-lesson: hard-to-fill roles are search-design problems wearing a sourcing costume. Teams that institutionalize the diagnostic — market map first, brief surgery second, channels third — stop accumulating 90-day searches at all.