Talent Matching

01The keyword-matching problem in practice

A Boolean search for "project manager" AND "Agile" AND "Scrum" will silently exclude a candidate who wrote "led sprint planning and delivery for a cross-functional product team" — genuinely describing the same experience in different words, which happens constantly with candidates from non-traditional backgrounds, career changers, and non-native English speakers translating their experience into a second language. The keyword filter has no way to know these are the same thing; it can only match strings.

The failure runs in both directions. A candidate who lists "Agile," "Scrum," and "stakeholder management" in a skills section without ever describing what they actually did with those methodologies will pass the same filter a genuinely experienced candidate might fail. Keyword matching optimises for resume vocabulary, not job-relevant evidence — the two frequently diverge.

02What changes with dimension-based matching

Expertini extracts competency dimensions directly from each job description — not a generic template, the specific requirements of that specific posting — and evaluates a CV against each dimension for documented evidence, regardless of the exact words used. "Reduced page load time by 40% through backend query optimisation" and "improved application performance through database tuning" score similarly against a dimension like "performance optimisation experience," because both describe the same underlying evidence.

In practice this means a recruiter's shortlist changes shape. Fewer strong candidates get filtered out for vocabulary mismatch; fewer weak candidates get through purely on keyword density. The shortlist that results tends to be smaller and stronger, not just differently ordered.

03Where this matters most

The gap between keyword and evidence-based matching widens with three specific candidate populations: people transitioning between industries (whose relevant experience is described in a different industry's vocabulary entirely), candidates writing in a second language, and candidates from educational or professional backgrounds that don't map onto a recruiter's mental template for "what this role's resume looks like." None of these groups are less qualified on average — they're simply less legible to a keyword filter, which is a property of the filter, not the candidate.

04How this connects to the underlying score

Talent matching and the Candidate Match Score are the same mechanism described from two angles: this page from the recruiter's outcome, the CMS page from the formula's internals. Every matching decision Expertini makes is backed by the same dimension-level evidence and the same deterministic arithmetic — nothing described here happens "automatically" in some separate, less-explainable layer.

05A concrete before-and-after

Consider a job description requiring "experience leading data migration projects." A keyword filter searching for "data migration" and "project lead" would miss a candidate whose CV says "oversaw the transition of a legacy customer database to a new CRM platform across three regional offices" — genuinely the same experience, described operationally rather than in the exact phrase a recruiter guessed to search. Dimension-based matching reads the second phrase as strong evidence for the same underlying competency, because it evaluates what was actually done rather than whether specific words appear.

This pattern repeats across industries with different specifics: a healthcare candidate describing "coordinated post-operative care across a 12-bed unit" against a JD asking for "patient management experience," or a finance candidate writing "built variance models flagging forecasting drift" against a requirement for "financial modelling skills." In each case the vocabulary gap is real and the underlying evidence gap is not — which is precisely the distinction keyword matching cannot make and evidence-based matching is built to make.