Interpretar resultados de SJT: scores, patterns y qué hacer

Score sin contexto es ruido

Score 72%. ¿Strong? ¿Average? Depende de:

1. Metodología (most-effective vs distance)
2. Comparison group
3. Specificidad del rol
4. Master ranking

Metodologías de scoring

Most-effective (MD)

Punto solo si top match expert. Binario, defensible. Fácil explicar.

Distance-based

Distance metric (sum abs diffs). Recompensa partial alignment, granular.

Usa MD para: managers; consistency. Usa distance para: individual contributors; nuance.

Internal benchmarks vs external norms

Internal (recomendado)

Top performers toman SJT. Promedio = comparison.

Por qué: mide alignment con tu definition.

Cómo crear:

1. 5-10 high performers, 2+ años
2. Toman SJT
3. Calcula promedio + variability

High variability = scenario ambiguo (bueno).

External norms (cautiously)

Comerciales tienen norms. Tricky:

• Generic, no custom
• Sanity-check, no decisión

Patterns importan más que score

Dos candidatos 76%. Pattern diferente:

A: matches en technical, deviates en people. Hipótesis: weak en people judgment. B: less consistent overall. Hipótesis: doesn't understand context.

Ambos 76%. A weakness coachable. B harder to fix.

Track por dominio: technical, people, execution, risk.

SJT + interview coherence

Strong SJT + strong interview: high confidence hire. Strong SJT + weak interview: know in abstract, no examples. Probe. Weak SJT + strong interview: no entendieron contexto. Investiga. Weak + weak: consistent signal.

Cuándo SJTs no predicen

Predicen: decision quality, problem-solving, escalation, people judgment.

No predicen: execution speed, persistencia, learning velocity, communication, technical skill.

Pareja con coding, behavioral, structured.

Red flags

1. Todos scoreas igual → assessment no diferencia
2. Variance no correlate con seniority → midiendo lo wrong
3. Demographic groups score significativamente diferente → bias

Comunicar

No "score 72%." Sí: "Ranked top consistently con top performers en 3/5. Judgment en X aligned. Approach en Y differs."

Usar en decisión

1. Screening filter
2. Interview probe
3. Tiebreaker
4. Onboarding data

No knockout para borderline. Plataforma ClarityHire.