What are the best practices in Calibration?

Hi ​@shragr, Excuse us for the wait here. Below are 9 best practices. 

1. Treat Calibration as an Iterative Alignment Tool

Calibration is not a “one and done” task. Expect criteria to evolve based on the position and shifting employer needs.

• Validate the extremes: After setting criteria, review both the highest-ranked and lowest-ranked leads. Adjust only if you strongly disagree with the results.

• Refresh regularly: Revisit calibrations for roles that have been open for a long time or when the hiring manager’s requirements change.

2. Hit the Completeness Basics

A "complete" calibration provides the engine with enough data points to build a reliable model. Aim for these thresholds:

• Core Components: Location, ideal candidates, skills, titles, and years of experience.

• The "Rule of 3": A strong baseline typically includes 3+ ideal candidates, 3+ skills, and 3+ alternate titles, along with min/max years of experience.

3. Ideal Candidates: Quality Over Quantity

Ideal candidates have the most significant impact on the match score—they define "what good looks like."

• The Sweet Spot: Use at least 3 ideal candidates, but keep the set between 3–5 to avoid introducing too much "noise" into the signal.

• Profile Richness: Select individuals currently in the same or similar roles who have detailed profiles (comprehensive skills and experience).

• Avoid Outdated Data: Do not use ideal candidates who are no longer in relevant roles, as this can create inflated or irrelevant match signals.

4. Skills: Be Selective to Avoid Over-Constraining

Over-filtering can significantly reduce the quality of your lead pool and negatively impact match behavior.

• Preferred vs. Required: Use 3–5 preferred skills. Use no more than one "Required" skill, and only if it is truly non-negotiable.

• Limit Constraints: Avoid excessive "Must Have" or "Must Not Have" rules.

• Audit Adjacencies: Double-check "related skills" to ensure a chosen skill isn't accidentally pulling in a completely different talent pool.

5. Titles: Use Industry-Standard Alternates

Account for the different ways companies label the same work.

• Diversify: Add at least 3 alternate titles to broaden the search.

• Avoid Internal Jargon: Steer clear of company-specific acronyms, abbreviations, or overly generic titles.

• Stay Flexible: Avoid marking alternate titles as "Required," as this can unnecessarily narrow your results.

6. Experience: Set Realistic Ranges

Years of experience impact the match score but should not be used to hard-exclude every candidate outside the range.

• Relevant vs. Total: Depending on the role, decide whether "Relevant Years" (specialized/technical) or "Total Years" is the better indicator.

• Avoid Strict Hard-Stops: Treat the range as a guide for the match engine rather than a binary filter.

7. Keep Calibration Broad, Use Filters to Narrow

Calibrate to cast a reasonably wide net first. Once you have a healthy pool, use pipeline filters and saved views to drill down into specific slices of the talent pool.

8. Leverage the Calibration Assistant Carefully

The Calibration Assistant can speed up the process by suggesting roles to copy, sample candidates, and related titles.

• Standardize: Only mark a calibration as "reusable" or a "reference" once it is 100% complete and performing accurately.

• Verify After Copying: When copying a calibration from another role, always re-check fields like title, location, and experience, as these may not always carry over.

9. Understand the Match Score

• Realistic Expectations: Not every role will yield 5-star matches. Depending on market supply, a 3-star match might be your top-tier candidate.

• Alignment vs. Quality: Remember that a match score indicates how well a candidate aligns with your calibrated requirements, not a guaranteed prediction of their future performance.

Additional resources 

• Calibration FAQs 

• Calibration & Filters Best Practices