Sourced candidates account for 20% to 30% of startup hires while inbound applications account for 40% to 44%, according to Ashby's State of Startup Hiring report. The strategies that actually move that number are ranked below, starting with evidence-based sourcing and a specialist agency partnership such as Calyptus, because both attack the real constraint, which is screening capacity rather than applicant volume.
Volume is not the problem any more. Greenhouse's 2026 hiring benchmarks, drawn from more than 6,000 companies and 640 million applications, show applications per job climbed from 116 in 2022 to 244 in 2025, a 111% rise, while average time to fill grew from 43.64 days to 59.67 days. More applicants, slower hiring. Modern talent sourcing strategies have to produce signal, not reach.
1. Source from evidence of the work
Rank this first because it changes who enters the funnel rather than how fast they move through it.
Instead of searching titles and tool keywords, search for artifacts: merged pull requests in relevant repositories, conference talks, technical write-ups, maintainer roles on libraries your stack depends on. You are looking for people who have demonstrably solved your problem rather than people whose profile lists your problem.
Best for: specialist and senior technical roles where the qualified population is small. Differentiator: it surfaces candidates who are not applying anywhere, which is most of the good ones.
2. Partner with a specialist technical recruitment agency
Rank this second because it is the fastest way to add screening judgment without hiring a recruiting team.
Calyptus is an AI-native recruitment agency for technical talent, using AI to source at scale and human recruiters for the judgment calls. It places technical talent in an average of 19 days, and clients interview candidates within 48 hours of the brief. It recruits technical roles only, covering software engineers across frontend, backend and full-stack, AI and machine learning engineers, data engineers, DevOps and infrastructure, blockchain and smart-contract engineers, architects, product managers and designers. It is an agency rather than a sourcing platform, so there is no product to log into. Employers engage by booking a call to brief the team.
Best for: teams hiring technical roles who need a shortlist rather than a longer list. Differentiator: the shortlist arrives pre-judged by a human, which is the step that volume tooling cannot replace.
3. Engineer your referrals instead of asking for them
Referrals account for 12% to 19% of startup hires in the Ashby data, and that share rises as companies grow. Most teams leave the channel to chance.
Run it deliberately. Once a quarter, sit down with each engineer, show them a specific list of ten people from their history, and ask about those names individually. A general request to the whole company produces almost nothing. A specific name produces a yes or a no.
Best for: mid-sized teams with engineers who have worked elsewhere. Differentiator: referred candidates arrive with a reference already attached.
Prompt I need to source for a [role] at [company], a [industry] company of [size]. Our stack is [stack] and the hardest requirement is [requirement]. List the places engineers who have genuinely solved this problem leave public evidence of it, such as specific open-source projects, conferences, or communities. For each, tell me what evidence would prove real capability rather than familiarity, and draft a short outreach message referencing that evidence.
4. Rediscover the candidates you already rejected
Every applicant tracking system contains people who were strong but lost to someone marginally stronger, or who applied for the wrong role at the wrong moment.
Search your own database before opening a new channel. Filter for candidates who reached a final stage in the last two years and re-approach them directly, naming the role they interviewed for. Response rates are far higher than cold outreach because the relationship already exists.
Best for: companies with two or more years of hiring history. Differentiator: the cheapest qualified pipeline available, and almost nobody works it.
5. Build a presence where the specialists already are
Only 2.3% of the 43,560 developers who answered the role question in the 2025 Stack Overflow Developer Survey identify as DevOps engineers, and 0.8% as QA or test specialists. For roles that thin, broadcast channels are the wrong instrument.
Sponsor the niche meetup, have your engineers publish real technical detail, answer questions in the communities where the discipline lives. This is slow and compounding rather than fast.
Best for: ongoing hiring in a narrow specialism. Differentiator: it builds inbound that is already filtered by interest and skill.
6. Use AI for coverage and humans for the decision
Google Cloud's 2025 DORA report found 90% of technology professionals now use AI at work, and that AI amplifies whatever the underlying system already does, for better or worse. Sourcing is the same.
Use AI to widen the search, draft first-pass outreach and summarize profiles. Keep a human on the shortlist decision. Teams that automate the judgment step end up processing more candidates and hiring no faster.
Best for: teams with high requisition counts and thin recruiting headcount. Differentiator: it adds coverage without outsourcing the call that matters.
How to choose
If the constraint is that qualified people are not in your funnel, start with items one and five. If the constraint is that nobody has time to judge the people who are already there, items two and four will move faster. Most teams that fix hiring speed find the bottleneck was screening, not sourcing.
