Here is a scenario every staffing agency recruiter in India has lived through: the AI shortlisting tool surfaces a candidate with a 78% match score. Strong enough to send to the client. The client comes back and says the candidate does not know Java — and Java was listed right at the top of the JD as a mandatory requirement.
The recruiter is embarrassed. The client loses confidence. The AI tool gets the blame. But the tool was not wrong exactly — it was solving the wrong equation. It scored the candidate on how many skills they had out of the full list, treating every skill as equally important. It never asked which skills were non-negotiable and which were just nice to have.
This is the structural flaw in how almost every ATS and AI recruitment tool handles skill evaluation — and it is directly responsible for a large proportion of client shortlisting rejections in Indian staffing agencies.
An ATS that treats must-have and good-to-have skills identically is not doing skills matching. It is doing skills counting. These are not the same thing.
What Must-Have and Good-to-Have Actually Mean
The distinction sounds simple. In practice, most recruitment systems ignore it entirely.
Non-negotiable. Candidate without these cannot do the job. Missing even one is a hard disqualifier.
Valuable but not blocking. A candidate without these can still do the job — they add value if present.
A Java developer role might list Java, Spring Boot, and MySQL as must-have skills, with AWS, Docker, and Kubernetes as good-to-have. A candidate who knows AWS and Docker but not Java is not a 60% match — they are a zero. They cannot do the job. No amount of AWS experience compensates for the absence of the core language the entire role is built around.
How the Broken Scoring Actually Works
When a standard ATS receives a JD with eight skills listed, here is what it does internally. It assigns a weight to each skill — typically equal weight across all eight. It checks the candidate resume for evidence of each skill. It scores the candidate proportionally on how many skills are present.
Five skills present out of eight means roughly 62.5%. Six out of eight means 75%. The score looks meaningful. But what if the three missing skills include the two most critical ones?
In a real agency dataset, approximately 18% of recruiter call rejections were caused by missing must-have skills — candidates who passed AI screening at 70%+ but failed immediately when the recruiter asked them about a core required competency. Every one of those calls was preventable with a hard gate on must-have skills.
Why ATS Vendors Build It This Way
The equal-weight approach is not incompetence. It is a deliberate design choice — and it makes sense from a software perspective. Requiring a recruiter to classify every skill as must-have or good-to-have adds friction to JD input. Most ATS vendors optimise for ease of use over precision of results.
The consequence is that the system becomes a volume tool rather than a quality tool. It passes more candidates through — which feels like it is working — but it passes the wrong ones, which is discovered only at the client rejection stage.
There is also a subtler problem: most JDs written by clients do not clearly distinguish must-have from good-to-have. They list ten skills in a bullet format with no indication of which are dealbreakers. The ATS has no way to know — so it treats all ten equally and hopes for the best.
The responsibility for must-have classification cannot sit entirely with the JD. The recruitment system needs to prompt for it, enforce it, and use it as a hard gate — not an optional field.
The Hard Gate Model: How It Should Work
A hard gate means must-have skills are evaluated first and act as a binary pass/fail filter — before any other scoring happens. Here is the correct architecture:
Gate Check
+ experience
Gate Check
Filtered out.
The cap at 55% is important. It does not score the candidate at zero — that would discard candidates who might be genuinely close to the requirement. It signals clearly that the candidate is a moderate match at best, and that a must-have gap exists, without making the system binary in a way that misses edge cases.
The Four-Component Scoring Model
With must-have skills correctly separated, a well-designed matching model allocates weight across four components:
| Component | Weight | Logic |
|---|---|---|
| Must-Have Skills Gate | 20% | Binary pass/fail first. Score = (must-haves evidenced ÷ total must-haves) × 20. If score is below 10 (less than 50% must-haves present), cap total matching % at 55 regardless of other components. |
| JD Responsibilities Alignment | 40% | Scored as (responsibilities evidenced ÷ total responsibilities) × 40. Counts concrete work evidence only — not claimed skills. |
| Experience Quality and Relevance | 25% | Years of relevant experience, depth in required domain, career trajectory. Recency weighted — last 3 years at full value, older experience in different domains at 50%. |
| Good-to-Have Skills | 15% | Skills that add value but are not blocking. Scored proportionally. Cannot compensate for must-have gaps due to the cap rule above. |
| Cap rule: If must-have component scores below 10 points, total matching_percentage is capped at 55% regardless of other component totals. | ||
A candidate who scores 20/40 on responsibilities, 18/25 on experience, and 12/15 on good-to-have skills — but only 8/20 on must-haves — gets a total of 58 points. Under a standard ATS, they would proceed. Under the cap rule, they are capped at 55% and filtered out. The recruiter never wastes a call on them.
How to Identify Must-Have Skills in a JD
This is where the system needs human input — or smart AI extraction. Not every JD clearly labels skills as mandatory. Here are the signals to look for when classifying skills from a JD:
Keywords like "must have", "mandatory", "required", "essential", "non-negotiable", "only candidates with" directly preceding a skill name are clear signals. If the JD says "must have Java experience", Java is a hard must-have — no interpretation needed.
Skills listed at the top of a requirements section, before any "preferred" or "plus" qualifier, are typically must-haves. Skills listed under a section titled "Nice to Have", "Preferred", or "Additional Skills" are good-to-have by definition.
For technical roles, the primary programming language, core framework, or primary platform the role operates on is always a must-have, even if the JD does not explicitly say so. A React developer role without React experience is not a developer for that role — period.
The most reliable source is the client. When a new JD comes in, a two-minute conversation to ask "which skills are absolute dealbreakers?" transforms JD intake from a guessing exercise into a precise filter. Most clients know exactly which skills they cannot train for and will not compromise on — they just do not always write it clearly in the JD.
The Compounding Problem: Skills Inflation in Indian Resumes
In India specifically, resume skills inflation makes the equal-weight problem worse than it would be elsewhere. A large proportion of Indian job applicants on platforms like Naukri and Shine list skills they have limited or outdated exposure to — because they know ATS systems scan for keywords.
A candidate who attended a three-day workshop on AWS five years ago lists AWS on their resume. An ATS that treats that keyword as a confirmed skill match, at equal weight to five years of daily AWS usage, is producing a meaningless score.
The must-have gate partially addresses this by requiring evidence of the skill in actual work experience — not just presence of the keyword on the resume. A skill that appears only in a certifications section or a training list, without any evidence of it being used in a real work context, should carry significantly lower weight than a skill embedded in a project description or job responsibility.
There is a critical difference between a skill being present on a resume and a skill being evidenced in work experience. ATS systems that keyword-match against the resume as a whole cannot distinguish between "I attended a Docker training in 2019" and "I architected and deployed containerised microservices using Docker across 12 production systems for 3 years." Both trigger a positive match. Only one is real evidence.
What Recruiters Should Change Right Now
Stop accepting JDs as a flat skill list. Every JD that comes in from a client should be processed through a must-have classification step before it is loaded into any system. Ask the client, or review the JD carefully, and tag each skill as must-have or good-to-have before configuring the job in your ATS. This takes three to five minutes per JD and changes the quality of every candidate that flows through it.
Use the pre-qualification questionnaire to cross-verify must-have skills. Even after must-have gates are set in the AI scoring model, ask candidates to self-rate their proficiency on each must-have skill as part of the pre-qualification form. A candidate who scores 9/10 on Java but has no Java mentioned in any work experience is giving you a false signal — the AI re-scoring step should cross-reference the self-rating against resume evidence and weight accordingly.
Track must-have skill failures at the client rejection stage. When a client rejects a profile, ask them to specify which requirement the candidate failed on. If the answer is a must-have skill three times in a row, your must-have gate for that JD is not calibrated correctly. Feed that information back into the system.
One Change That Outperforms Everything Else
Of all the improvements a staffing agency can make to its AI screening system, separating must-have from good-to-have skills and implementing a hard gate is the single change with the highest impact-to-effort ratio. It does not require a new platform, a major integration, or weeks of configuration. It requires classifying skills correctly when a JD is set up — and configuring the matching threshold to enforce that classification.
The agencies that do this consistently will send fewer profiles to clients, with dramatically higher shortlisting rates, and will build a reputation for quality that compounds over every placement cycle.
TeamOB - SafalHires separates must-have and good-to-have at the JD intake stage
The platform prompts recruiters to classify every skill when loading a JD, applies a hard cap when must-have skills are missing, and cross-references candidate self-ratings against resume evidence in the pre-qualification re-scoring step. Fewer profiles to clients — higher shortlisting rate on every one.
Free trial