Performance Evaluation
Framework

A structured, evidence-backed approach to fair and transparent 6-month performance reviews across the Data Science Team. Seven pillars. One consistent standard. No single category can carry the rest.

1Task Scoring Standard
2Performance Evaluation
01 · Pillars

The 7 Performance Pillars

Each pillar has a defined evidence type. Task-driven pillars flow from work output; independent pillars require separate evidence collection.

Delivery

Consistently delivers end-to-end solutions with clear SLAs and measurable business impact.

Task scoresProject outcomesBusiness impact metricsSLAsDeployments
Task-Driven

Technical Excellence

Advances MLOps, Agentic AI practices, automation, scalability and reduces operational toil.

Design reviewsArchitecture contributionsAutomation & CI/CDReusable frameworks
Task-Driven

Reliability

Ensures model & data quality, monitoring, stability and upholds privacy/ethics-by-design.

MonitoringIncident historyData/model quality reportsRegression metrics
Task-Driven

Collaboration

Translates business goals, communicates clearly, aligns stakeholders and influences decisions.

Peer feedbackStakeholder feedbackCross-team alignmentCommunication clarity
Partially Task-Driven

Leadership / Team Contribution

Raises team standards, creates assets, mentors others and improves team practices.

MentorshipOnboarding contributionsStandards & templatesGuidelines & initiatives
Mostly Independent

Development

Continuously improves skills, builds portfolio, learns, experiments and shares knowledge.

CertificationsTalks & coursesPet projectsExperimentationLearning logs
Independent

Company Culture

Demonstrates growth mindset, respectful behavior, and supports a blameless, psychologically safe culture.

Peer feedbackManager observationsRespectful & accountable behavior examples
Independent
02 · Foundation

How This Framework Fits Together

Two systems — task scoring and pillar evaluation — work in sequence. Understanding how they connect is key to using the framework correctly.

Step 1 · Input

Task Scoring Standard

Rates individual tasks 1–5 stars across 4 weighted dimensions. Produces evidence-backed task scores that feed into pillar calculations.

  • 4 weighted dimensions (Quality, Impact, Timeliness, Engineering)
  • Work type guidelines (Production, POC/Research, Operational)
  • Star rating output (1–5)
  • Evidence & artifacts
View Task Scoring Standard

Step 2 · This

Performance Evaluation

Evaluates overall contribution across 7 pillars over 6 months. Task scores are evidence — but not the only factor.

  • Broader behaviors
  • Sustained impact over the period
  • Growth & influence
  • Balanced contribution across all 7 pillars

How the two systems connect — the 3-stage mechanism

Two separate sets of weights serve two completely different purposes. This is the part most often misunderstood.

1

Score the task

Use the 4-dimension rubric to determine how well the task was executed.

Quality40%
Impact30%
Timeliness20%
Engineering10%

→ Task Score (0–100) → ★ rating

This answers: how good was this task?

2

Assign pillar weights

Separately, decide which performance pillars this task is relevant to — and how much.

Delivery40%
Tech Excellence35%
Reliability25%

(example — must total 100%)

→ Task Score × each % = pillar contribution

This answers: which areas does this task count toward?

3

Aggregate per pillar

Each pillar collects contributions from all tasks that were mapped to it, then averages them by weight.

All Delivery contributions → weighted avg

All Tech contributions → weighted avg

All Reliability contributions → weighted avg

(repeated for all 7 pillars)

→ 7 Pillar Scores (1–5) → Performance Rating

This answers: how did this person perform across the full period?

The critical distinction: The 4-dimension weights (40/30/20/10) determine how good the task was. The pillar weights are a completely separate set of numbers — they determine which performance areas the task is relevant to. A single task can contribute to multiple pillars simultaneously, with different amounts.

Core Principle

No single category can compensate for another. We value balanced, sustainable, and well-rounded contribution.

03 · Mapping

How Tasks Roll Up Into Pillar Scores

Each task is scored independently, then mapped to the performance pillars it influenced — with contribution weights that reflect actual relevance. Pillar scores are the weighted average across all contributing tasks.

1

Input

Evaluate Tasks

Score each task 1–5 on 4 weighted dimensions (Quality, Impact, Timeliness, Engineering Maturity) using the Task Scoring Standard.

2

Distribution

Map to Pillars

Assign what % of each task's impact goes to each performance pillar. Weights must total 100% across all pillars for that task.

3

Aggregation

Aggregate Scores

Each pillar's score is the weighted average of all contributing tasks. Higher weight + higher score = greater pillar impact.

Example · Task-to-Pillar Contribution Map

Each task's weights must total 100% across all assigned pillars.

TaskScoreDeliveryTech Exc.ReliabilityCollaborationLeadershipDevelopmentCulture
ML Pipeline Automation4.5★40%35%25%————
Incident Root Cause Resolution5★20%10%50%20%———
Stakeholder Dashboard Project4★50%20%20%10%———
Pillar Score
4.36/ 5
4.42/ 5
4.66/ 5
4.67/ 5
———

* Only pillars with a weight > 0 are included in that pillar's score calculation.

How it was computed

TaskTask ScoreDelivery WeightScore × Weight
ML Pipeline Automation4.5 ★40%4.5 × 0.40 = 1.80
Incident Root Cause Resolution5.0 ★20%5.0 × 0.20 = 1.00
Stakeholder Dashboard Project4.0 ★50%4.0 × 0.50 = 2.00

Sum of Weighted Scores

1.80 + 1.00 + 2.00 = 4.80

Sum of Weights

0.40 + 0.20 + 0.50 = 1.10

Delivery Pillar Score

4.80 ÷ 1.10 = 4.36 / 5

Why not a simple average? (4.5 + 5.0 + 4.0) ÷ 3 = 4.50. But the Stakeholder Dashboard Project contributed 50% of its impact to Delivery — so it should weigh more. The weighted result (4.36) reflects actual relevance across the period.

04 · Rating Scale

How Performance Is Rated

All 7 pillars use the same 1–5 scale. Scores are assigned through calibrated manager judgment, supported by evidence.

1

Significant Concerns

Performance far below expectations.

2

Partially Meets

Inconsistent. Needs improvement.

3

Meets Expectations

Consistent, solid, dependable.

4

Exceeds Expectations

Clearly above expectations. Strong impact.

5

Exceptional Impact

Outstanding, sustained. Role model.

Threshold Rule

To receive an overall rating of 4 or 5, every critical pillar must score at least 3 (Meets Expectations). Company Culture is an explicit critical pillar — a score below 3 caps the overall rating at 3, regardless of all other scores.

05 · Guardrails

Framework Guardrails

Built-in protections that prevent distortion, gaming, and accidental over-rewarding or under-recognition.

Category Isolation

Strong performance in one pillar cannot offset weak performance in another. Each pillar stands entirely on its own.

Minimum Threshold Rule

An overall rating of 4 or 5 requires every critical pillar to be at least 3 (Meets Expectations). No pillar can be critically low.

Sustained Performance

One-off big projects don't define performance. Consistency, repeatability, and long-term impact across the full period are what matter.

Score Rounding — Manager's Discretion

Computed pillar scores are decimal values (e.g., 4.36, 4.42). Final performance ratings are expressed as whole numbers (1–5). Rounding a decimal score up or down to the nearest integer is at the manager's discretion — applied based on the weight of evidence, overall performance patterns, and sustained behavior across the review period, not mechanically. This discretion is bounded by the framework guardrails above and the Manager Override Clause (±1) in the contingencies below.

Framework Self-Audit — Two Dimensions to Always Distinguish

≠ High Collaboration ≠ High Performance≠ High Delivery ≠ Good Teammate
A

Behavioral Integrity Override

2+ formal manager observations or 2+ independent peer behavioral reports automatically cap the overall rating at 3, regardless of all other scores. Requires written documentation to prevent misuse.

B

Company Culture — Explicit Hard Gate

Company Culture is a critical pillar. A score below 3 prevents any overall rating above 3, regardless of all other pillar scores.

C

Collaboration: Engagement Quality Sub-Dimension (~20%)

Measures whether the person engages respectfully, gives credit, and handles disagreement constructively — not just what they delivered, but how they showed up.

D

Structured Anonymous Pulse Survey

For Company Culture, a 3–4 question anonymous survey at mid-period and end-period supplements manager observation to reduce popularity bias and create consistency.

E

Manager Override Clause (±1)

Managers may adjust the final calculated rating by ±1 with required written justification, handling edge cases the formula cannot capture (e.g., high scores but a serious behavioral incident).

06 · Process

The Review Process

A structured 5-step cycle to ensure consistent, fair, and actionable performance reviews.

1

Self-Reflection

Employee prepares evidence and self-assessment across all 7 pillars. Collect impact metrics, task scores, artifacts, and examples.

2

Manager Review

Manager evaluates using this framework, submitted evidence, peer feedback, and behaviors observed over the 6-month period.

3

Calibration

Managers align scores across all team members to ensure consistency and fairness. Outlier scores in either direction require justification.

4

Feedback Conversation

Discuss strengths, growth areas, and the development plan together. Frame opportunities against the career progression path.

5

Action Plan

Set concrete goals and focus areas for the next 6 months. Tie development directly to Career Paths and team strategic roadmap.

07 · Best Practices

Employee Best Practices

Build evidence continuously throughout the period — not just at review time.

Maintain an Accomplishment Log

Document completed tasks, decisions, and outcomes in real time — not just at review time.

Collect Impact Metrics & Artifacts

Attach dashboards, code links, feedback emails, and adoption numbers to your task records.

Document Learning & Growth

Track certifications, talks, courses, experiments, and any knowledge-sharing activities.

Share Early, Communicate Clearly

Timely status updates and stakeholder alignment are evidence for Collaboration and Delivery.

Our Goal

Grow people. Deliver impact. Build the future together.