Define readiness score formulas that go beyond gut feel

The cost of vibe checks — and the upside of clarity

Gut feel is fast, but it’s expensive. Sending unready learners to panels burns recruiter goodwill and confidence. Holding back ready learners delays offers. Programs that adopt a transparent readiness_score typically see:

• 12–20 point lift in readiness over 4–6 weeks by focusing practice where it matters
• 18–25% improvement in offer rate by green-lighting at the right time
• 30–45% mentor hours saved by triaging feedback to high-impact gaps
• 7–12 days faster time-to-panel through clear go/no-go rules

With SageStepper, the raw signals—Nova interview logs, rubric ratings, streak length, and coverage—flow into a single score you can defend in committee and explain to a learner.

Clarity beats vibe checks. When a candidate passes the threshold, both coach and learner know why.

The inputs: what data belongs in a fair readiness score

A strong score uses signals that are explainable, coachable, and relevant to job performance. In SageStepper, the following are captured automatically or via mentor rubrics:

• Mastery: difficulty-adjusted correct rate and rubric pass rate from Nova interviews and live mocks.
• Consistency: stability across the last 3–5 sessions (lower variance, higher score).
• Transfer: performance on unseen questions and full-panel simulations vs practice sets.
• Communication: structured responses (e.g., STAR), clarity, and completeness per rubric—no penalizing accent or non-native speech.
• Coverage: percent of target competencies attempted with passing performance.
• Coachability: speed-to-improvement after feedback; retry_success_rate within 7 days.

Signals we intentionally avoid:

• Protected attributes (gender, race, age) or proxies.
• Device type, connection speed, or background noise as performance proxies.
• Black-box features you can’t explain (e.g., opaque “likeability scores”).

SageStepper’s progress radar visualizes these inputs so learners see which spoke is holding them back and what to do next.

A simple, auditable formula you can tune

Start with a weighted blend that any coach can compute by hand. Keep scores 0–100 and weights that sum to 1.0.

// typescript
export type Pillars = {
  mastery: number;        // 0-100 difficulty-adjusted pass rate
  consistency: number;    // 0-100 (100 = low variance over last 5 sessions)
  transfer: number;       // 0-100 performance on unseen / panel mocks
  communication: number;  // 0-100 rubric score for structure & clarity
  coverage: number;       // 0-100 % competencies with passing evidence
  coachability: number;   // 0-100 improvement speed after feedback
};

export type Weights = {
  mastery: number; consistency: number; transfer: number;
  communication: number; coverage: number; coachability: number;
};

export const defaultWeights: Weights = {
  mastery: 0.35,
  consistency: 0.15,
  transfer: 0.20,
  communication: 0.10,
  coverage: 0.10,
  coachability: 0.10,
};

export function readinessScore(p: Pillars, w: Weights = defaultWeights): number {
  const score = (
    p.mastery * w.mastery +
    p.consistency * w.consistency +
    p.transfer * w.transfer +
    p.communication * w.communication +
    p.coverage * w.coverage +
    p.coachability * w.coachability
  );
  return Math.round(score);
}

Thresholds and decision rules:

• score >= 78 → Go (green): candidate enters recruiter loops within 7 days.
• 70 ≤ score < 78 → Yellow: targeted Interview Sprint on 1–2 weakest pillars.
• score < 70 → Not yet: 2-week plan plus mentor checkpoint.

Config as data, not code, so you can adjust without redeploying:

# readiness-config.yaml
weights:
  mastery: 0.35
  consistency: 0.15
  transfer: 0.20
  communication: 0.10
  coverage: 0.10
  coachability: 0.10
thresholds:
  go: 78
  yellow: 70
window:
  recency_days: 28  # EMA half-life for recency weighting
policy:
  protected_attributes: [gender, race, age, disability]
  excluded_features: [accent_score, device_type, background_noise]

Calibration, ethics, and explainability

A score is only useful if it predicts outcomes and treats cohorts fairly. Build a quarterly ritual to calibrate weights and check drift.

1. Link to outcomes. Compare readiness_score to live panel results and offer decisions. Target calibration: for candidates within ±3 points of the threshold, predictive accuracy > 70%.
2. Audit fairness. Run difference-in-performance (DIF) by cohort and program. Flag if average score at the decision boundary differs by > 3 points across groups with similar outcomes.
3. Explain every point. Learners should see their pillar breakdown in the progress radar and a plain-language reason, e.g., “Coverage 62: No system design evidence at D3+.”
4. Protect privacy. Use a 28-day rolling window; archive raw audio and video after feature extraction; restrict access via role=mentor|admin.
5. Human-in-the-loop. Edge cases (few data points, big variance) route to a mentor for review before a go decision.

Quick weekly checks you can automate:

# bash (runs in CI or a cron job)
python score_candidates.py --config readiness-config.yaml \
  --since 7d --emit metrics.json

python fairness_check.py --input metrics.json \
  --threshold 78 --max_gap 3

python drift_monitor.py --historical scores_history.csv \
  --alert_slack '#readiness-ops'

Ethical guardrails baked into SageStepper:

• Nova never grades accent or appearance; communication scoring is rubric-based on structure and completeness.
• Learners can request rubric reviews; all overrides are logged and visible in the action boards history.

Operationalize in SageStepper: go/no-go with confidence

Translate the math into a daily workflow that saves mentor time and builds learner confidence.

• Configure thresholds in the workspace settings on the progress radar so green/yellow/red reflect your program.
• Run an Interview Sprint. 5-day sequence with Nova targeting the 2 weakest competencies; retests auto-schedule until coverage ≥ 80.
• Use action boards. Each yellow candidate gets 2–3 atomic tasks tied to their pillar gaps (e.g., “D3 arrays practice set” + “Peer teach-back on STAR”).
• Hold a 20-minute readiness huddle (twice weekly). Pull the “Near Threshold” swimlane and decide: go, sprint, or mentor review.
• Publish clarity. Add a score_card.md to each learner’s board with a one-line rationale and next steps.

Quick links:

• Explore the platform: /product
• See sample workflows: /workflow
• Join community office hours: /community
• Pricing and plans: /pricing

Example: three learners, three decisions

Below is anonymized sample data from Nova and rubric logs. We compute pillars, the readiness_score, and the decision.

{
  "aisha": {
    "pillars": {"mastery": 82, "consistency": 76, "transfer": 79, "communication": 86, "coverage": 74, "coachability": 88},
    "notes": ["D3 arrays and SQL strong", "Coverage low on system design"],
    "streak_days": 9
  },
  "leo": {
    "pillars": {"mastery": 67, "consistency": 61, "transfer": 58, "communication": 72, "coverage": 69, "coachability": 77},
    "notes": ["High variance; misses unseen graph problems"],
    "streak_days": 3
  },
  "marta": {
    "pillars": {"mastery": 80, "consistency": 90, "transfer": 84, "communication": 79, "coverage": 92, "coachability": 70},
    "notes": ["Excellent coverage; slower to adopt feedback"],
    "streak_days": 14
  }
}

Applying the default weights:

• aisha → score 81 → Go. Action: schedule recruiter loop; assign one micro-sprint for system design.
• leo → score 65 → Not yet. Action: 2-week Interview Sprint on graphs + unseen sets; mentor check-in after 5 sessions.
• marta → score 84 → Go. Action: mock panel this week; communication drill on concision.

Observed impact over 4 weeks using this framework at a partner bootcamp:

• Average threshold-crossing time dropped from 21 to 13 days.
• Offer rate improved from 41% to 51% among green-lighted candidates.
• Mentor 1:1 time per learner decreased by 32% while satisfaction scores rose.

Weekly rituals, dashboards, and guardrails

Make the score durable with simple rituals and lightweight automation.

• Monday: Recompute readiness_score for all active learners; post the “Near Threshold” list in Slack.
• Wed/Fri: 20-minute huddles; decide go/sprint/review.
• Daily: Nova practice streaks (aim for 5+); push targeted flashcards for any pillar < 75.

Sample SQL to materialize a scoring view:

-- BigQuery or Postgres SQL (simplified)
WITH recent AS (
  SELECT
    learner_id,
    AVG(difficulty_adjusted_pass) FILTER (WHERE days_ago <= 28) * 100 AS mastery,
    (100 - STDDEV_SAMP(session_score) OVER (PARTITION BY learner_id)) AS consistency,
    AVG(unseen_pass) * 100 AS transfer,
    AVG(rubric_communication) * 100 AS communication,
    (COUNT(DISTINCT competency_passed)::float / COUNT(DISTINCT target_competency)) * 100 AS coverage,
    AVG(retry_success_within_7d) * 100 AS coachability
  FROM practice_events
  WHERE days_ago <= 90
  GROUP BY learner_id
)
SELECT *,
  ROUND(
    mastery*0.35 + consistency*0.15 + transfer*0.20 +
    communication*0.10 + coverage*0.10 + coachability*0.10
  ) AS readiness_score
FROM recent;

Slack template for the huddle:

Readiness Huddle (20m)
- Greens (>=78): {{names}} → Schedule panels, assign final mock.
- Yellows (70–77): {{names}} → 5-day Interview Sprint; gaps: {{pillars}}.
- Reds (<70): {{names}} → 2-week plan; mentor assigned: {{mentor}}.
- Outliers: {{names}} → Review due to low sample size/high variance.

Guardrails to prevent gaming:

• Cap credit from repeated identical questions; require diversity in competency_id.
• Weight recency (EMA) but lock-in a minimum coverage requirement (e.g., 70+ across core competencies) for green.
• Randomize Nova question sets and require at least one full mock panel before go.

Put it to work today

• Define your rubric and pillar weights in one hour; publish thresholds to learners.
• Turn on the progress radar and action boards; map yellow learners to focused work.
• Schedule twice-weekly huddles; measure offer-rate lift and mentor hours saved monthly.

If you want a hand, we’ll share our starter configs and help you calibrate on your data.

• Start for free: /login
• Book a product tour: /product
• Learn from peers in our community: /community