Hiring

Pre-employment assessment tools that predict performance

A pre-employment assessment is a forecast. The only question worth asking about one is whether the forecast beats what you would have done without it — and for most hiring processes, the honest answer today is that a résumé screen plus an unstructured interview is barely better than chance.

r ≈ .42

Structured interviews, revised meta-analytic estimate (Sackett et al., 2022)

r ≈ .19

Unstructured interviews, same analysis

Every hire

Gets a structured guide generated automatically

Rank your selection methods by what they actually predict

Sackett, Zhang, Berry and Lievens (2022) re-estimated the validity of common selection methods after correcting the range-restriction errors that inflated the widely-cited 1998 figures. The ordering survived; the magnitudes shrank. Structured interviews and work samples lead, unstructured interviews and years of experience trail badly.

The practical implication is unglamorous. Most organisations could improve hiring accuracy more by writing down their interview questions in advance than by buying any software, including this one. TalentSpark's position is that the assessment should feed the structured interview, not replace it.

  • Structured interview — the highest-leverage change available, and procedural rather than technical
  • Work sample — strong, expensive, and hard to standardise across candidates
  • Psychometric profile — modest on its own, valuable as the thing that tells the interview where to look
  • Unstructured interview — near-worthless as a predictor, excellent at transmitting your biases

The workflow, end to end

Invite the candidate. They complete an adaptive battery in around fifteen minutes on any device. Scoring runs immediately: percentile scores against the job family benchmark, validity flags, and a role-fit estimate. Claude drafts a narrative report and a structured interview guide anchored to that candidate's specific risk areas.

The hiring manager opens one document that says what to probe and why, with a scoring rubric attached. Interviewers score independently before conferring — the ordering matters, because discussion before scoring collapses independent judgments into the loudest one in the room.

Candidate experience is a selection variable

Assessment abandonment is not evenly distributed. Candidates with competing offers leave first, which means a long assessment silently selects for the candidates with fewer options. Adaptive delivery, mobile-first design and a visible progress indicator exist to protect the top of your funnel, not to be nice.

Every candidate receives their own profile summary after completion. This is a fairness commitment, and under the EU AI Act's transparency obligations for high-risk employment systems it is increasingly a legal one.

Frequently asked questions

What is the most predictive pre-employment assessment?

Structured interviews and work samples lead, at around r = .42 and r = .33 respectively on revised meta-analytic estimates. Personality measures are more modest. The strongest practical approach is a composite: a psychometric profile that tells a structured interview where to probe, scored against a rubric.

When in the funnel should assessment happen?

After an initial qualification screen and before the interview loop, so the profile can inform the interview. Assessing every inbound applicant is expensive and irritates candidates; assessing after the interview wastes the profile's main use.

Does pre-employment testing hurt candidate experience?

Long, irrelevant and opaque testing does. Assessment that is short, obviously job-relevant, and returns something to the candidate generally improves perceptions of process fairness. Returning the profile to the candidate is the part most vendors skip.

How do we prove our assessment is job-related if challenged?

Content validity evidence linking assessed constructs to a job analysis, adverse impact monitoring on live selection rates, and per-decision audit trails. TalentSpark produces all three as a by-product of normal use.

Predict performance before day one.

Validated psychometrics and Claude AI, in one platform. No credit card required.