Sourcing offers a more flexible and powerful alternative to traditional candidate search. Instead of translating a role into rigid keywords, dropdowns, and filters, you can describe what a strong candidate looks like in natural language—including their experience, responsibilities, company context, and transferable skills.
This approach may take a little getting used to, especially if you are familiar with building highly structured searches in other tools. Rather than trying to define the perfect search upfront, you will get the best results by starting with the core requirements, reviewing candidates, and using feedback to guide the agent.
Once you become comfortable with this iterative approach, you can search in a much more nuanced way than traditional filters allow. The agent can interpret context, explore adjacent backgrounds, and improve its understanding of the role over time.
This guide will help you adjust your search strategy and get the most from that process: start a search → rate candidates → refine → shortlist.
- Kick off search by providing one of the following: Job Description, Job Post link or a description of your ideal candidate
- The sourcing agent generates an Ideal Candidate Profile (ICP) for you. Review it and edit before you start rating candidates.
- The agent searches databases and finds candidates that match your ICP.
- Give feedback on at least 5 candidates using
Yes/ Maybe/ No. This feedback is how the system learns.
- Click
Refine search to improve results. Expect 2–3 refinement cycles to get dialled in.
- Saved candidates appear in the
Candidates tab (top right).
See our help documentation for more details on this process.
Improve quality faster
Ensure you have a strong ICP
Treat your ICP as a living document — refine it as you learn. Here is a helpful checklist:
- Must-have skills are specific (tools, systems, domain knowledge)
- Seniority is clearly defined (what “senior” means for this role)
- Company context is included (stage, size, pace, constraints)
- Nice-to-haves are separated from must-haves
- Exclusions are explicit (who you don’t want to see)
Rule of thumb: If you’d still interview someone without a requirement, it probably belongs in Nice to have, not Must have.
Give feedback that’s easy to learn from
The fastest way to improve results is to make “No” feedback specific.
Advanced prompting techniques
Sourcing works best when you describe the context of the role, not just the job title. Think about where this person is coming from, what they’ve done before, and why they’re a fit for your company.
Upload a resume of a top performer and ask: “Find candidates similar to this person, but with more fintech experience and leadership background.”
Use strategic exclusions
Exclusions help the AI avoid “close but wrong” profiles. Call out the backgrounds you don’t want to see to tighten results faster.
- Marketing example: “Growth marketing manager with B2B SaaS experience, but not pure brand marketers”
- Data example: “ML engineer with production deployment experience, but not purely academic researchers”
You can also exclude candidates who are unlikely to be useful because your team has already encountered them. Open the Exclusions menu or tell the agent directly to exclude candidates who:
- Were seen in a previous search
- Are already in your connected ATS
- Were contacted in a sequence
- Recently changed jobs
For example:
“Exclude candidates already in our ATS or contacted in a sequence.”
These settings apply only to the current search, so you can use different exclusions for each role. The agent can still access excluded candidates and explain who was filtered out and why.
Great candidates don’t always come from identical backgrounds. If someone could succeed because they have similar responsibilities, domain knowledge, or growth-stage experience—even from a different industry—include those adjacent profiles in your search.
Add company context
Company context explains what it’s like to work at your company (industry, size, stage, and pace) so the AI can prioritise candidates with relevant experience.
- Stage: Define “early-stage startup” vs “scale-up” vs “enterprise”
- Work style: “Remote-first culture requiring async communication skills”
- Team: “Small team where everyone wears multiple hats”
- Growth: “Rapid scaling phase, need someone comfortable with ambiguity”
Prompt templates
These example templates show how to combine role requirements, company context, and exclusions into a clear sourcing prompt.
Technical role template
[Role] for a [company stage] [industry] company, with [X years] experience in [technologies/domains]. Must have: [must-have requirements]. Nice to have: [nice-to-haves]. Exclude: [clear exclusions]. Must be comfortable with: [work environment / culture].
Leadership role template
[Role] who has scaled [function/team] from [X to Y], with experience in [industry or domain]. Must have: [leadership scope, change management, hiring/structure]. Exclude: [wrong background]. Looking for someone who leads with: [leadership style].
Advanced features you can use
Once you’ve calibrated your search, AI Sourcing can help you analyse and share your candidate set. Try asking:
- “Show me a breakdown of candidates by experience level”
- “Create a chart of candidates by previous company size”
- “Export this candidate list as a CSV for my hiring manager”
- “Visualize the geographic distribution of these candidates”
Advanced workflow: use Sourcing from Slack
For a more efficient, asynchronous workflow, connect Slack and continue your sourcing searches without repeatedly checking Metaview. Slack is especially useful when you are running several searches at once or want to respond quickly when the sourcing agent needs input.
You can use Slack to:
- Start a search by messaging the Metaview Slack app with a role or sourcing brief
- Receive updates when the Ideal Candidate Profile needs review
- Reply in the same thread to answer questions or refine the search
- Attach job descriptions or other relevant files
- Receive a notification when the Candidate Pack is ready
- Jump back into Metaview for detailed candidate review
Keep each search in its original Slack thread so the agent retains the right context.
Quick troubleshooting
If your search returns too few candidates
If you’re seeing fewer candidates than expected, your search may be more constrained than intended. Small requirements can combine to narrow the pool significantly. Try the following:
- Review the generated ICP. Check whether any preferences have been treated as must-haves when they should be nice-to-haves.
- Relax one constraint at a time. Start with location, seniority, exact job titles, target companies, or exclusions so you can see what is limiting the search.
- Focus on skills and responsibilities, not only titles. Strong candidates may use different titles or come from adjacent industries, company stages, or functions.
- Use an example candidate. Share a representative profile or resume to help the agent understand the kinds of adjacent matches you would consider.
- Refine the existing search. This preserves your ratings and feedback, helping the agent improve with each iteration.
Our candidate data is broad, but it is not exhaustive, and coverage can vary across locations, industries, and highly specialised roles. We are continuously working to improve our data sources and search coverage. If you believe relevant profiles are consistently missing for a particular role or market, please share the search and examples with our team. This feedback helps us identify gaps and prioritise improvements for real sourcing use cases.
If the candidates aren’t relevant yet
It can take a few rounds of feedback for the search to understand exactly what a strong candidate looks like for your role. Rather than starting over, use this refinement loop:
- Rate at least 5 candidates. A mix of
Yes, Maybe, and No ratings gives the agent enough signal to identify patterns.
- Explain every
No. Be specific about what is wrong—for example, “too junior,” “missing enterprise sales experience,” or “wrong type of company background.”
- Update the ICP. Add any missing must-haves and clarify the backgrounds, skills, or profiles you want to exclude.
- Click
Refine search. The agent will use your ratings, feedback, and updated ICP to adjust the results.
- Repeat for 2–3 cycles. Search quality should improve as the agent learns what matters most for the role.
Please keep in mind that Sourcing relies on the information available in public professional profiles, which can sometimes be incomplete or ambiguous. Some details—such as a candidate’s current interests, exact scope of responsibility, or openness to a new role—may only become clear once you reach out and speak with them.
Benchmarks
Use these benchmarks to know when your search is calibrated:
- Hit rate: 50–75% of candidates are “Yes” or “Maybe”
- Calibration speed: 2–3 refinement cycles to reach quality matches
- Outreach-ready: ~80% of “Yes” candidates are worth contacting
- Time to quality: find strong candidates within ~15 minutes
