Building a candidate scorecard: must-haves and nice-to-haves in AI sourcing

What a candidate scorecard does in the sourcing phase

A candidate scorecard is a fixed method for assessing candidates early in the process. You translate the vacancy into clear criteria, decide which points are genuinely must-haves, and set out exactly what evidence you want to see for each in a profile. That lets you compare candidates objectively and rank the longlist based on visible relevance rather than gut feeling alone.

This works fundamentally differently from an interview scorecard. You use that during conversations, because at that point you can genuinely assess answers, behaviour and examples. A sourcing scorecard is used much earlier, before you even start outreach. At that stage you're looking exclusively at publicly visible information. That's why the criteria used at this stage need to be very concrete, verifiable and directly work-related.

For recruiters, sourcers and hiring managers this is particularly valuable, because it makes prioritising vacancy criteria a lot easier. You explain clearly upfront what carries real weight and what matters less. That leads to far less discussion afterwards, and also lets you better justify why one candidate gets approached before another.

Why a candidate scorecard works better than loose keywords

In practice, loose keywords often turn out to be too limited. A candidate can be a perfect fit for the role even if their current job title, industry or the terminology used in their profile differs considerably from the literal vacancy text. That's why it's smarter to start from the actual outcome of the role. First decide what someone needs to achieve in this specific role. Then translate that into criteria for assessing candidates that you can trace back through experience gained, responsibilities held and results achieved.

This targeted approach makes a recruitment scorecard far stronger, because you're not stuck on overused labels. You genuinely look at the work someone has done in the past. That helps enormously in assessing candidates both more broadly and more fairly. It also stops knock-out criteria in hiring from becoming unnecessarily strict, simply because you draw a sharper line between what's necessary and what's desirable.

An added benefit is that working with hiring managers goes noticeably more smoothly and calmly. When the criteria are clearly set out in advance, it's far easier to explain why someone ranks high on the list. That's exactly why a fixed candidate scorecard also helps build buy-in, especially when several people are looking critically at the same longlist.

How to build a candidate scorecard with must-haves and nice-to-haves

Start with the outcome of the role

Always start with the core goal of the role. What does this new person actually need to improve, build, manage or speed up in practice? Think more stable workflows, fewer mistakes made, better internal collaboration, or simply more grip on the data. Once this is completely clear, it immediately becomes obvious which experience is genuinely relevant. And that forms the absolute foundation of a usable candidate scorecard.

Translate the outcome into verifiable criteria

Write out every criterion in plain, understandable language. Avoid vague catch-all terms like culture fit, a senior presence or commercial energy. Terms like that are very hard to substantiate in the sourcing phase. Instead, choose visible, demonstrable work, such as building data pipelines, actively running projects, or working closely with internal decision-makers. That makes the criteria you use to assess candidates far easier to defend to the team.

Keep the number of must-haves limited

The must-haves from the vacancy need to stay clear and few. Often four or five points already turn out to be plenty. Ask yourself the critical question for each requirement: could someone reasonably do the job without this specific point too? If the answer is no, it can safely count as a must-have. If the answer is yes, even if it takes some ramp-up time or a slightly different route, it usually belongs among the nice-to-haves instead. That way you avoid tight knock-out criteria in hiring shrinking the available pool too much.

Give nice-to-haves a clear place

Nice-to-haves distinguish between candidates who already meet the baseline anyway. For this you apply clear weighting of your selection criteria. You're awarding extra points for experience that increases the odds of success in that particular role. Think completed migrations, coaching colleagues, valuable sector knowledge or demonstrable stakeholder management experience. For teams that want to apply all of this in a structured way, a fixed approach to AI sourcing for corporate recruiters helps maintain both quality and transparency early in the process.

Must-haves in a candidate scorecard: what counts and what doesn't

When is something genuinely necessary

A criterion is only genuinely a must-have if someone can't reasonably do the job without it. That may sound simple, but this exact question makes a huge difference in practice. For instance, experience with a specific type of work is generally more important than specific experience at one particular kind of company. The ultimate goal is to keep the must-haves in the vacancy sharp, logical and strictly functional.

Which exclusion grounds are often too strict

A degree isn't automatically a must-have by definition; years of hands-on experience can demonstrate exactly the same level. Formal job titles are also often fairly unreliable, since different companies label exactly the same work differently. The same often applies to the most recent employer or specific sector. Instead, always look at the concrete tasks carried out, the responsibility held and the demonstrable results achieved. That way your assessment stays genuinely relevant to the role itself.

What counts as evidence in a profile

Preferably work with three very simple levels per criterion: sufficient evidence, partial evidence, and finally missing evidence. Sufficient evidence shows beyond doubt that someone has genuinely done this work. Partial evidence gives clear pointers in the right direction but still calls for an extra check. Missing evidence just means you simply can't see it yet. It's never an automatic rejection. And that realisation is exactly what makes comparing candidates objectively far more achievable in your day-to-day routine.

Weighting the candidate scorecard: making nice-to-haves fair and explainable

A good points scale needs, first and foremost, to stay workable and simple. The must-haves function as gate criteria: candidates either meet them or they don't. Nice-to-haves are then only assigned points based on their expected impact on success in the role. That makes the final score highly explainable for recruiters, sourcers and hiring managers. You don't need to build a complicated mathematical model to justify ranking the longlist clearly and logically.

A practically designed model simply tends to work best. For instance, give each nice-to-have present a light, medium or high value. Then apply exactly the same logic consistently every time. A nice-to-have that directly and concretely helps achieve the role's main objective generally carries more weight than one that's purely 'handy to have'. That way the weighting applied to your selection criteria stays consistent at all times.

Always flag any uncertainty separately. If a profile, for example, says surprisingly little about giving coaching or maintaining stakeholder contact, that is absolutely not proof that someone is naturally weak at it. For situations like that, prepare a clear check-in question for outreach or the first intake. That keeps the overall score fairer at all times. An AI candidate score also only genuinely works well if the underlying input is very clear and the outcome stays properly verifiable.

Example candidate scorecard for a senior data engineer role

Below is a fully worked-out example for a senior data engineer at a mid-sized organisation. The role calls for someone who builds and manages data pipelines independently, thrives in production environments, and can also bring colleagues along on the technical choices made. This example shows nicely how must-haves, nice-to-haves, evidence and ranking come together smoothly in one practical, efficient method.

The four must-haves

• 1. Relevant programming experience: the profile clearly shows the candidate has recently and actively worked with programming languages well suited to data engineering.
• 2. Experience with data pipelines: the profile convincingly shows the candidate has independently built, actively managed or significantly improved data flows.
• 3. Working in production: this particular candidate has directly applicable experience with systems that are genuinely used widely within an organisation.
• 4. Practical availability: the reviewed profile makes it plausible that a long-term working arrangement in the desired setup is geographically and practically feasible.

The nice-to-haves and their weight

• Cloud migration: gets a high weighting, because this specific experience will directly help with the planned modernisation of the current platform.
• Mentoring: a medium weighting, because in practice you often see that a senior colleague needs to be able to actively guide less experienced team members.
• Sector knowledge: a somewhat light weighting, simply because while it's useful and handy, it generally counts for less than core technical experience.
• Working with decision-makers: medium weighting, because this role requires a lot of alignment with both other parts of the organisation and the management team.

What the scoring logic looks like

Use the must-haves first and foremost as the absolute gate. If a candidate falls short on a genuine must-have, that person doesn't end up at the top of the final outreach order. If someone easily meets all the must-haves, you then add up the nice-to-haves they've scored. In this example, cloud migration carries the heaviest, highest value, closely followed by mentoring and stakeholder contact, with sector knowledge counting only after that. Missing or non-visible information is simply noted as an open check-in question, not treated as an unfair mark against the candidate.

Candidate A

Candidate A easily meets every must-have. The profile reviewed shows very clear, concrete experience with the programming languages, data pipelines, production and collaboration in the preferred working setup. Among the nice-to-haves, there's strong evidence for both the cloud migration and mentoring experience. Constructive collaboration with internal decision-makers also comes through credibly. As a result, this particular candidate lands on a fairly high total score, say 92 points, and firmly takes first place in the resulting outreach order.

Candidate B

Candidate B also meets every must-have comfortably. At first glance there's more than enough evidence for extensive programming work, setting up data pipelines, production and availability too. Among the chosen nice-to-haves, however, there's considerably less visible evidence for both mentoring and active stakeholder contact. As a result the final score comes out somewhat lower, say 78 points in total. This candidate remains interesting enough, but logically lands just in a second tier, flagged with at least one very targeted check-in question for the first intake.

How to order the longlist logically

With this pragmatic approach you can rank the longlist entirely on the basis of public, visible relevance. You see far more quickly who should be approached directly and proactively, who genuinely needs extra checking in the process, and who falls out straight away based on the must-haves. That brings an enormous amount of calm to the sometimes hectic first selection round, and also makes the substantive conversation with all stakeholders much clearer, because every step taken can be traced back flawlessly afterwards.

Checking a candidate scorecard for bias before going live

Always have a second reader critically review your scorecard before putting it to use, just to be safe. That person checks thoroughly whether every chosen criterion genuinely relates directly to the day-to-day work. Look especially critically and sharply for possible unintended preferences around, for example, age, background, gender, a previous employer or an education route completed. If it turns out that such a specific preference isn't directly necessary and essential for carrying out the role, rewrite it into concrete, visible work behaviour, or simply remove it altogether.

Always keep any uncertainty visible. Once again, missing information doesn't count as negative evidence. That realisation matters enormously, simply because some good candidates happen to describe their experience a little more briefly or less fully on paper than others. A solid bias check successfully prevents writing style, the size of someone's network, or the particular shape of a career from suddenly counting for more than genuine, deep-rooted suitability. That way your recruitment scorecard stays considerably more reliable, and also a good deal easier to explain.

How a candidate scorecard fits into AI sourcing with human oversight

A scorecard like this is, above all, a fantastic tool within a much broader, mapped-out process. The recruiter remains ultimately responsible for the criteria, the interpretation and, of course, the final choice. In our approach, a scorecard per candidate within AI sourcing can nonetheless help enormously by making every must-have, nice-to-have and its supporting evidence directly visible. That means you consistently see faster why someone ranks high or somewhat lower on the list, while still keeping human oversight firmly in the lead.

That step is, of course, never fully separate from the rest. The entire hiring process naturally runs from the first vacancy through to searching, assessing and effectively approaching candidates. If you'd like to see exactly how these parts logically connect, feel free to read how we work at heart, from vacancy to personal outreach. That keeps it very clear in practice that scoring is purely a valuable intermediate step, and certainly never becomes a fully automatic final decision.

Careful, manual checking also remains essential, especially with tricky borderline cases or simply too little information available online. If you'd also like to see exactly how this methodical approach is applied successfully in day-to-day practice, feel free to take a look at our current real-world results with AI-supported recruitment. Those concrete examples give immediate, clear context for the sometimes tricky choices in the process and its practical application.

Fillable template for your own vacancy

Feel free to use this format to set up your own scorecard quickly and easily. To do that, first write out every section clearly in plain, ordinary language. That makes the final setup widely usable for recruiters, sourcers and, of course, hiring managers. It also helps to discuss the prioritising of vacancy criteria together at an early stage, so the bar stays exactly the same for everyone right from the start.

• Step 1: outcome of the role Describe crystal-clearly what this person needs to concretely achieve in the role, within the team, a process, or the organisation as a whole. Think specifically about the result, the ownership held, and the overall impact.
• Step 2: must-haves from the vacancy Carefully fill in which unique characteristics are genuinely needed to reasonably do the day-to-day work at all. Keep this section strictly limited to the points that are truly gate criteria.
• Step 3: nice-to-haves with weighting Note clearly which additional experience strongly increases the odds of long-term success, and quickly assign each point a light, medium or possibly high value.
• Step 4: evidence per criterion Record unambiguously what exactly counts as convincingly sufficient evidence, still somewhat partial evidence, or plainly missing evidence in the profile reviewed.
• Step 5: uncertainty and check-in questions Note down which points you'd rather test more deeply a little later in the process, of course without accidentally weighing them negatively already at this stage.
• Step 6: bias check with a second reader Check, or have someone check with certainty, whether every criterion written down is genuinely work-related, and whether any unnecessary preferences have quietly crept into the text.

If you'd like to explore this clarifying setup in your own workflow personally, you can try Elvatix yourself and see at your leisure how these scorecards fit brilliantly within smart, personal recruitment AI.

Frequently asked questions about scorecards in recruitment

What's the difference between an interview scorecard and a sourcing scorecard?

An interview scorecard is naturally used mainly during actual conversations to assess the answers given, behaviour and the practical examples mentioned. A sourcing scorecard, by contrast, is used much earlier in the process to rank potential candidates in advance, based entirely on visible criteria and the evidence that comes through in their online profile.

How many must-haves should I include?

Keep the total number as limited as possible. In surprisingly many cases, just four or five genuine must-haves turn out to be plenty. Any stricter requirements unnecessarily and dangerously shrink the talent pool, and considerably increase the risk that you rule out genuinely good, suitable candidates far too early in the process.

How do I prevent bias in criteria for assessing candidates?

Always check thoroughly whether every criterion mentioned genuinely relates directly to the work to be done. Ideally have a second person with a fresh perspective review it, and be incredibly critical of underlying, unintended preferences around age, background, gender, current employer or completed education route. Also set any uncertainty that comes up neatly aside in a list, rather than immediately scoring and weighing it unfairly as negative.

What do I do with missing information in a profile?

Never treat plainly missing information as direct grounds for rejection. Simply and efficiently note it as an open question. You can then easily test this in a very targeted, sharp way during outreach or in a good first intake. That way you ensure the assessment made stays considerably fairer for everyone involved.

When do you use an AI candidate score as a tool?

That's undoubtedly especially useful when you're dealing with the fact that you want to assess quite a lot of candidates against fixed, tight criteria, while still wanting to keep firm, clear control over the logic used in that process. The recruiter deployed always remains personally responsible for the criteria set, the weighting applied, and ultimately the absolute final choice too. That's exactly why this handy support also works best by far as an integral part of a very clear process, where adequate human oversight always remains central.